1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
2 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-linux -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-linux -fexceptions -fcxx-exceptions -emit-pch -o %t %s
4 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-unknown-linux -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK2
5
6 // RUN: %clang_cc1 -triple x86_64-unknown-linux -verify -fopenmp-simd -x c++ -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
7 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
8 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -debug-info-kind=limited -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
9 // expected-no-diagnostics
10 #ifndef HEADER
11 #define HEADER
12
main(int argc,char ** argv)13 int main(int argc, char **argv) {
14 #pragma omp parallel
15 #pragma omp 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: 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_IV:%.*]] = alloca i64, align 8
58 // CHECK1-NEXT: [[TMP:%.*]] = alloca i64, align 8
59 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8
60 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8
61 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
62 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
63 // CHECK1-NEXT: [[ARGC1:%.*]] = alloca i32, align 4
64 // CHECK1-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
65 // CHECK1-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
66 // CHECK1-NEXT: [[_TMP5:%.*]] = alloca i8**, align 8
67 // CHECK1-NEXT: [[_TMP6:%.*]] = alloca i8*, align 8
68 // CHECK1-NEXT: [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
69 // CHECK1-NEXT: [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
70 // CHECK1-NEXT: [[I:%.*]] = alloca i64, align 8
71 // CHECK1-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
72 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
73 // CHECK1-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
74 // CHECK1-NEXT: [[_TMP28:%.*]] = alloca i8, align 1
75 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
76 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
77 // CHECK1-NEXT: store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
78 // CHECK1-NEXT: store i8*** [[ARGV]], i8**** [[ARGV_ADDR]], align 8
79 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
80 // CHECK1-NEXT: [[TMP1:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
81 // CHECK1-NEXT: store i64 0, i64* [[DOTOMP_LB]], align 8
82 // CHECK1-NEXT: store i64 9, i64* [[DOTOMP_UB]], align 8
83 // CHECK1-NEXT: store i64 1, i64* [[DOTOMP_STRIDE]], align 8
84 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
85 // CHECK1-NEXT: store i32 0, i32* [[ARGC1]], align 4
86 // CHECK1-NEXT: [[TMP2:%.*]] = load i8**, i8*** [[TMP1]], align 8
87 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i64 0
88 // CHECK1-NEXT: [[TMP3:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
89 // CHECK1-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP3]], i64 0
90 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP0]], align 4
91 // CHECK1-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
92 // CHECK1-NEXT: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP5]]
93 // CHECK1-NEXT: [[TMP6:%.*]] = load i8**, i8*** [[TMP1]], align 8
94 // CHECK1-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP6]], i64 9
95 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
96 // CHECK1-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP7]], i64 [[LB_ADD_LEN]]
97 // CHECK1-NEXT: [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
98 // CHECK1-NEXT: [[TMP9:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
99 // CHECK1-NEXT: [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
100 // CHECK1-NEXT: [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
101 // CHECK1-NEXT: [[TMP12:%.*]] = add nuw i64 [[TMP11]], 1
102 // CHECK1-NEXT: [[TMP13:%.*]] = mul nuw i64 [[TMP12]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
103 // CHECK1-NEXT: [[TMP14:%.*]] = call i8* @llvm.stacksave()
104 // CHECK1-NEXT: store i8* [[TMP14]], i8** [[SAVED_STACK]], align 8
105 // CHECK1-NEXT: [[VLA:%.*]] = alloca i8, i64 [[TMP12]], align 16
106 // CHECK1-NEXT: store i64 [[TMP12]], i64* [[__VLA_EXPR0]], align 8
107 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP12]]
108 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP15]]
109 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
110 // CHECK1: omp.arrayinit.body:
111 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
112 // CHECK1-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
113 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
114 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP15]]
115 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
116 // CHECK1: omp.arrayinit.done:
117 // CHECK1-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[TMP1]], align 8
118 // CHECK1-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
119 // CHECK1-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
120 // CHECK1-NEXT: [[TMP19:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
121 // CHECK1-NEXT: [[TMP20:%.*]] = sub i64 [[TMP18]], [[TMP19]]
122 // CHECK1-NEXT: [[TMP21:%.*]] = sdiv exact i64 [[TMP20]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
123 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP21]]
124 // CHECK1-NEXT: store i8** [[_TMP6]], i8*** [[_TMP5]], align 8
125 // CHECK1-NEXT: store i8* [[TMP22]], i8** [[_TMP6]], align 8
126 // 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
127 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
128 // CHECK1-NEXT: [[TMP24:%.*]] = bitcast i32* [[ARGC1]] to i8*
129 // CHECK1-NEXT: store i8* [[TMP24]], i8** [[TMP23]], align 8
130 // CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
131 // CHECK1-NEXT: [[TMP26:%.*]] = bitcast i32* [[TMP0]] to i8*
132 // CHECK1-NEXT: store i8* [[TMP26]], i8** [[TMP25]], align 8
133 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
134 // CHECK1-NEXT: store i64 4, i64* [[TMP27]], align 8
135 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
136 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP28]], align 8
137 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
138 // CHECK1-NEXT: store i8* null, i8** [[TMP29]], align 8
139 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
140 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP30]], align 8
141 // CHECK1-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
142 // CHECK1-NEXT: [[TMP32:%.*]] = bitcast i32* [[TMP31]] to i8*
143 // CHECK1-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP32]], i8 0, i64 4, i1 false)
144 // CHECK1-NEXT: [[DOTRD_INPUT_GEP_7:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t], [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 1
145 // CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 0
146 // CHECK1-NEXT: [[TMP34:%.*]] = load i8**, i8*** [[TMP1]], align 8
147 // CHECK1-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8*, i8** [[TMP34]], i64 0
148 // CHECK1-NEXT: [[TMP35:%.*]] = load i8*, i8** [[ARRAYIDX8]], align 8
149 // CHECK1-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds i8, i8* [[TMP35]], i64 0
150 // CHECK1-NEXT: [[TMP36:%.*]] = load i32, i32* [[TMP0]], align 4
151 // CHECK1-NEXT: [[TMP37:%.*]] = sext i32 [[TMP36]] to i64
152 // CHECK1-NEXT: [[LB_ADD_LEN10:%.*]] = add nsw i64 -1, [[TMP37]]
153 // CHECK1-NEXT: [[TMP38:%.*]] = load i8**, i8*** [[TMP1]], align 8
154 // CHECK1-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8*, i8** [[TMP38]], i64 9
155 // CHECK1-NEXT: [[TMP39:%.*]] = load i8*, i8** [[ARRAYIDX11]], align 8
156 // CHECK1-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds i8, i8* [[TMP39]], i64 [[LB_ADD_LEN10]]
157 // CHECK1-NEXT: store i8* [[VLA]], i8** [[TMP33]], align 8
158 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 1
159 // CHECK1-NEXT: store i8* [[ARRAYIDX9]], i8** [[TMP40]], align 8
160 // CHECK1-NEXT: [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX12]] to i64
161 // CHECK1-NEXT: [[TMP42:%.*]] = ptrtoint i8* [[ARRAYIDX9]] to i64
162 // CHECK1-NEXT: [[TMP43:%.*]] = sub i64 [[TMP41]], [[TMP42]]
163 // CHECK1-NEXT: [[TMP44:%.*]] = sdiv exact i64 [[TMP43]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
164 // CHECK1-NEXT: [[TMP45:%.*]] = add nuw i64 [[TMP44]], 1
165 // CHECK1-NEXT: [[TMP46:%.*]] = mul nuw i64 [[TMP45]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
166 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 2
167 // CHECK1-NEXT: store i64 [[TMP46]], i64* [[TMP47]], align 8
168 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 3
169 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP48]], align 8
170 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 4
171 // CHECK1-NEXT: store i8* null, i8** [[TMP49]], align 8
172 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 5
173 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP50]], align 8
174 // CHECK1-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 6
175 // CHECK1-NEXT: store i32 1, i32* [[TMP51]], align 8
176 // CHECK1-NEXT: [[TMP52:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
177 // CHECK1-NEXT: [[TMP53:%.*]] = load i32, i32* [[TMP52]], align 4
178 // CHECK1-NEXT: [[TMP54:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
179 // CHECK1-NEXT: [[TMP55:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP53]], i32 1, i32 2, i8* [[TMP54]])
180 // CHECK1-NEXT: store i8* [[TMP55]], i8** [[DOTTASK_RED_]], align 8
181 // CHECK1-NEXT: [[TMP56:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
182 // CHECK1-NEXT: [[TMP57:%.*]] = load i32, i32* [[TMP56]], align 4
183 // CHECK1-NEXT: call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP57]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
184 // CHECK1-NEXT: [[TMP58:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
185 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i64 [[TMP58]], 9
186 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
187 // CHECK1: cond.true:
188 // CHECK1-NEXT: br label [[COND_END:%.*]]
189 // CHECK1: cond.false:
190 // CHECK1-NEXT: [[TMP59:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
191 // CHECK1-NEXT: br label [[COND_END]]
192 // CHECK1: cond.end:
193 // CHECK1-NEXT: [[COND:%.*]] = phi i64 [ 9, [[COND_TRUE]] ], [ [[TMP59]], [[COND_FALSE]] ]
194 // CHECK1-NEXT: store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
195 // CHECK1-NEXT: [[TMP60:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
196 // CHECK1-NEXT: store i64 [[TMP60]], i64* [[DOTOMP_IV]], align 8
197 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
198 // CHECK1: omp.inner.for.cond:
199 // CHECK1-NEXT: [[TMP61:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
200 // CHECK1-NEXT: [[TMP62:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
201 // CHECK1-NEXT: [[CMP13:%.*]] = icmp sle i64 [[TMP61]], [[TMP62]]
202 // CHECK1-NEXT: br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
203 // CHECK1: omp.inner.for.cond.cleanup:
204 // CHECK1-NEXT: br label [[OMP_INNER_FOR_END:%.*]]
205 // CHECK1: omp.inner.for.body:
206 // CHECK1-NEXT: [[TMP63:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
207 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i64 [[TMP63]], 1
208 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i64 0, [[MUL]]
209 // CHECK1-NEXT: store i64 [[ADD]], i64* [[I]], align 8
210 // CHECK1-NEXT: [[TMP64:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
211 // CHECK1-NEXT: store i8** [[DOTTASK_RED_]], i8*** [[TMP64]], align 8
212 // CHECK1-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
213 // CHECK1-NEXT: store i32* [[ARGC1]], i32** [[TMP65]], align 8
214 // CHECK1-NEXT: [[TMP66:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
215 // CHECK1-NEXT: [[TMP67:%.*]] = load i8**, i8*** [[_TMP5]], align 8
216 // CHECK1-NEXT: store i8** [[TMP67]], i8*** [[TMP66]], align 8
217 // CHECK1-NEXT: [[TMP68:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
218 // CHECK1-NEXT: [[TMP69:%.*]] = load i32, i32* [[TMP68]], align 4
219 // CHECK1-NEXT: [[TMP70:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP69]], i32 1, i64 48, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
220 // CHECK1-NEXT: [[TMP71:%.*]] = bitcast i8* [[TMP70]] to %struct.kmp_task_t_with_privates*
221 // CHECK1-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP71]], i32 0, i32 0
222 // CHECK1-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP72]], i32 0, i32 0
223 // CHECK1-NEXT: [[TMP74:%.*]] = load i8*, i8** [[TMP73]], align 8
224 // CHECK1-NEXT: [[TMP75:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
225 // CHECK1-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP74]], i8* align 8 [[TMP75]], i64 24, i1 false)
226 // CHECK1-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP71]], i32 0, i32 1
227 // CHECK1-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP76]], i32 0, i32 0
228 // CHECK1-NEXT: [[TMP78:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
229 // CHECK1-NEXT: store i8* [[TMP78]], i8** [[TMP77]], align 8
230 // CHECK1-NEXT: [[TMP79:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
231 // CHECK1-NEXT: [[TMP80:%.*]] = load i32, i32* [[TMP79]], align 4
232 // CHECK1-NEXT: [[TMP81:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP80]], i8* [[TMP70]])
233 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
234 // CHECK1: omp.body.continue:
235 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
236 // CHECK1: omp.inner.for.inc:
237 // CHECK1-NEXT: [[TMP82:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
238 // CHECK1-NEXT: [[ADD14:%.*]] = add nsw i64 [[TMP82]], 1
239 // CHECK1-NEXT: store i64 [[ADD14]], i64* [[DOTOMP_IV]], align 8
240 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
241 // CHECK1: omp.inner.for.end:
242 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
243 // CHECK1: omp.loop.exit:
244 // CHECK1-NEXT: [[TMP83:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
245 // CHECK1-NEXT: [[TMP84:%.*]] = load i32, i32* [[TMP83]], align 4
246 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP84]])
247 // CHECK1-NEXT: [[TMP85:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
248 // CHECK1-NEXT: [[TMP86:%.*]] = load i32, i32* [[TMP85]], align 4
249 // CHECK1-NEXT: call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP86]], i32 1)
250 // CHECK1-NEXT: [[TMP87:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
251 // CHECK1-NEXT: [[TMP88:%.*]] = bitcast i32* [[ARGC1]] to i8*
252 // CHECK1-NEXT: store i8* [[TMP88]], i8** [[TMP87]], align 8
253 // CHECK1-NEXT: [[TMP89:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
254 // CHECK1-NEXT: store i8* [[VLA]], i8** [[TMP89]], align 8
255 // CHECK1-NEXT: [[TMP90:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
256 // CHECK1-NEXT: [[TMP91:%.*]] = inttoptr i64 [[TMP12]] to i8*
257 // CHECK1-NEXT: store i8* [[TMP91]], i8** [[TMP90]], align 8
258 // CHECK1-NEXT: [[TMP92:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
259 // CHECK1-NEXT: [[TMP93:%.*]] = load i32, i32* [[TMP92]], align 4
260 // CHECK1-NEXT: [[TMP94:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
261 // CHECK1-NEXT: [[TMP95:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP93]], i32 2, i64 24, i8* [[TMP94]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
262 // CHECK1-NEXT: switch i32 [[TMP95]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
263 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
264 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
265 // CHECK1-NEXT: ]
266 // CHECK1: .omp.reduction.case1:
267 // CHECK1-NEXT: [[TMP96:%.*]] = load i32, i32* [[TMP0]], align 4
268 // CHECK1-NEXT: [[TMP97:%.*]] = load i32, i32* [[ARGC1]], align 4
269 // CHECK1-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP96]], [[TMP97]]
270 // CHECK1-NEXT: store i32 [[ADD15]], i32* [[TMP0]], align 4
271 // CHECK1-NEXT: [[TMP98:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP12]]
272 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP98]]
273 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE22:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
274 // CHECK1: omp.arraycpy.body:
275 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
276 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST16:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT20:%.*]], [[OMP_ARRAYCPY_BODY]] ]
277 // CHECK1-NEXT: [[TMP99:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
278 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP99]] to i32
279 // CHECK1-NEXT: [[TMP100:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
280 // CHECK1-NEXT: [[CONV17:%.*]] = sext i8 [[TMP100]] to i32
281 // CHECK1-NEXT: [[ADD18:%.*]] = add nsw i32 [[CONV]], [[CONV17]]
282 // CHECK1-NEXT: [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
283 // CHECK1-NEXT: store i8 [[CONV19]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
284 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT20]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], i32 1
285 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
286 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE21:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT20]], [[TMP98]]
287 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_DONE22]], label [[OMP_ARRAYCPY_BODY]]
288 // CHECK1: omp.arraycpy.done22:
289 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB3]], i32 [[TMP93]], [8 x i32]* @.gomp_critical_user_.reduction.var)
290 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
291 // CHECK1: .omp.reduction.case2:
292 // CHECK1-NEXT: [[TMP101:%.*]] = load i32, i32* [[ARGC1]], align 4
293 // CHECK1-NEXT: [[TMP102:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP101]] monotonic, align 4
294 // CHECK1-NEXT: [[TMP103:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP12]]
295 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY23:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP103]]
296 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY23]], label [[OMP_ARRAYCPY_DONE36:%.*]], label [[OMP_ARRAYCPY_BODY24:%.*]]
297 // CHECK1: omp.arraycpy.body24:
298 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST25:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT34:%.*]], [[ATOMIC_EXIT:%.*]] ]
299 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST26:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT33:%.*]], [[ATOMIC_EXIT]] ]
300 // CHECK1-NEXT: [[TMP104:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
301 // CHECK1-NEXT: [[CONV27:%.*]] = sext i8 [[TMP104]] to i32
302 // CHECK1-NEXT: [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]] monotonic, align 1
303 // CHECK1-NEXT: br label [[ATOMIC_CONT:%.*]]
304 // CHECK1: atomic_cont:
305 // CHECK1-NEXT: [[TMP105:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY24]] ], [ [[TMP110:%.*]], [[ATOMIC_CONT]] ]
306 // CHECK1-NEXT: store i8 [[TMP105]], i8* [[_TMP28]], align 1
307 // CHECK1-NEXT: [[TMP106:%.*]] = load i8, i8* [[_TMP28]], align 1
308 // CHECK1-NEXT: [[CONV29:%.*]] = sext i8 [[TMP106]] to i32
309 // CHECK1-NEXT: [[TMP107:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
310 // CHECK1-NEXT: [[CONV30:%.*]] = sext i8 [[TMP107]] to i32
311 // CHECK1-NEXT: [[ADD31:%.*]] = add nsw i32 [[CONV29]], [[CONV30]]
312 // CHECK1-NEXT: [[CONV32:%.*]] = trunc i32 [[ADD31]] to i8
313 // CHECK1-NEXT: store i8 [[CONV32]], i8* [[ATOMIC_TEMP]], align 1
314 // CHECK1-NEXT: [[TMP108:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
315 // CHECK1-NEXT: [[TMP109:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i8 [[TMP105]], i8 [[TMP108]] monotonic monotonic, align 1
316 // CHECK1-NEXT: [[TMP110]] = extractvalue { i8, i1 } [[TMP109]], 0
317 // CHECK1-NEXT: [[TMP111:%.*]] = extractvalue { i8, i1 } [[TMP109]], 1
318 // CHECK1-NEXT: br i1 [[TMP111]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
319 // CHECK1: atomic_exit:
320 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT33]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i32 1
321 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT34]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], i32 1
322 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE35:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT33]], [[TMP103]]
323 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE35]], label [[OMP_ARRAYCPY_DONE36]], label [[OMP_ARRAYCPY_BODY24]]
324 // CHECK1: omp.arraycpy.done36:
325 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB3]], i32 [[TMP93]], [8 x i32]* @.gomp_critical_user_.reduction.var)
326 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
327 // CHECK1: .omp.reduction.default:
328 // CHECK1-NEXT: [[TMP112:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
329 // CHECK1-NEXT: call void @llvm.stackrestore(i8* [[TMP112]])
330 // CHECK1-NEXT: [[TMP113:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
331 // CHECK1-NEXT: [[TMP114:%.*]] = load i32, i32* [[TMP113]], align 4
332 // CHECK1-NEXT: call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP114]])
333 // CHECK1-NEXT: ret void
334 //
335 //
336 // CHECK1-LABEL: define {{[^@]+}}@.red_init.
337 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
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: store i32 0, i32* [[TMP3]], align 8
346 // CHECK1-NEXT: ret void
347 //
348 //
349 // CHECK1-LABEL: define {{[^@]+}}@.red_comb.
350 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
351 // CHECK1-NEXT: entry:
352 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
353 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
354 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
355 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
356 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
357 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
358 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
359 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
360 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
361 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
362 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
363 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP3]], align 8
364 // CHECK1-NEXT: ret void
365 //
366 //
367 // CHECK1-LABEL: define {{[^@]+}}@.red_init..1
368 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
369 // CHECK1-NEXT: entry:
370 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
371 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
372 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
373 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
374 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
375 // CHECK1-NEXT: [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
376 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
377 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
378 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
379 // CHECK1: omp.arrayinit.body:
380 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
381 // CHECK1-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
382 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
383 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
384 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
385 // CHECK1: omp.arrayinit.done:
386 // CHECK1-NEXT: ret void
387 //
388 //
389 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..2
390 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
391 // CHECK1-NEXT: entry:
392 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
393 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
394 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
395 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
396 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
397 // CHECK1-NEXT: [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
398 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
399 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
400 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
401 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
402 // CHECK1: omp.arraycpy.body:
403 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
404 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
405 // CHECK1-NEXT: [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
406 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP6]] to i32
407 // CHECK1-NEXT: [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
408 // CHECK1-NEXT: [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
409 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
410 // CHECK1-NEXT: [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
411 // CHECK1-NEXT: store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
412 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
413 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
414 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
415 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
416 // CHECK1: omp.arraycpy.done4:
417 // CHECK1-NEXT: ret void
418 //
419 //
420 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
421 // CHECK1-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
422 // CHECK1-NEXT: entry:
423 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
424 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8***, align 8
425 // CHECK1-NEXT: store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
426 // CHECK1-NEXT: store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
427 // CHECK1-NEXT: [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
428 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
429 // CHECK1-NEXT: [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
430 // CHECK1-NEXT: store i8** [[TMP3]], i8*** [[TMP4]], align 8
431 // CHECK1-NEXT: ret void
432 //
433 //
434 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
435 // CHECK1-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
436 // CHECK1-NEXT: entry:
437 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
438 // CHECK1-NEXT: [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
439 // CHECK1-NEXT: [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
440 // CHECK1-NEXT: [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
441 // CHECK1-NEXT: [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
442 // CHECK1-NEXT: [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
443 // CHECK1-NEXT: [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
444 // CHECK1-NEXT: [[TMP_I:%.*]] = alloca i8**, align 8
445 // CHECK1-NEXT: [[TMP4_I:%.*]] = alloca i8*, align 8
446 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i32, align 4
447 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
448 // CHECK1-NEXT: store i32 [[TMP0]], i32* [[DOTADDR]], align 4
449 // CHECK1-NEXT: store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
450 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
451 // CHECK1-NEXT: [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
452 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
453 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
454 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
455 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
456 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
457 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
458 // CHECK1-NEXT: [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
459 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
460 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
461 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
462 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
463 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
464 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
465 // CHECK1-NEXT: store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
466 // CHECK1-NEXT: store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
467 // 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
468 // CHECK1-NEXT: store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
469 // CHECK1-NEXT: store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
470 // CHECK1-NEXT: [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
471 // CHECK1-NEXT: [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
472 // CHECK1-NEXT: [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
473 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
474 // CHECK1-NEXT: call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
475 // CHECK1-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
476 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
477 // CHECK1-NEXT: [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
478 // CHECK1-NEXT: [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
479 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
480 // CHECK1-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
481 // CHECK1-NEXT: [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
482 // CHECK1-NEXT: [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
483 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
484 // CHECK1-NEXT: [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
485 // CHECK1-NEXT: [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
486 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
487 // CHECK1-NEXT: [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
488 // CHECK1-NEXT: [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
489 // CHECK1-NEXT: [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
490 // CHECK1-NEXT: [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
491 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
492 // CHECK1-NEXT: [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
493 // CHECK1-NEXT: [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
494 // CHECK1-NEXT: [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
495 // CHECK1-NEXT: [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
496 // CHECK1-NEXT: [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
497 // CHECK1-NEXT: [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
498 // CHECK1-NEXT: [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
499 // CHECK1-NEXT: [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
500 // CHECK1-NEXT: [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
501 // CHECK1-NEXT: [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
502 // CHECK1-NEXT: store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
503 // CHECK1-NEXT: [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
504 // CHECK1-NEXT: [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
505 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
506 // CHECK1-NEXT: [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
507 // CHECK1-NEXT: [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
508 // CHECK1-NEXT: [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
509 // CHECK1-NEXT: [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
510 // CHECK1-NEXT: [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
511 // CHECK1-NEXT: [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
512 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
513 // CHECK1-NEXT: store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
514 // CHECK1-NEXT: store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
515 // CHECK1-NEXT: ret i32 0
516 //
517 //
518 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
519 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
520 // CHECK1-NEXT: entry:
521 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
522 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
523 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
524 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
525 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
526 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
527 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
528 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
529 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
530 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
531 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
532 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
533 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
534 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
535 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
536 // CHECK1-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
537 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
538 // CHECK1-NEXT: [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
539 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
540 // CHECK1-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
541 // CHECK1-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
542 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
543 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
544 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
545 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
546 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
547 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
548 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
549 // CHECK1: omp.arraycpy.body:
550 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
551 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
552 // CHECK1-NEXT: [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
553 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP22]] to i32
554 // CHECK1-NEXT: [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
555 // CHECK1-NEXT: [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
556 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
557 // CHECK1-NEXT: [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
558 // CHECK1-NEXT: store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
559 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
560 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
561 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
562 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
563 // CHECK1: omp.arraycpy.done5:
564 // CHECK1-NEXT: ret void
565 //
566 //
567 // CHECK2-LABEL: define {{[^@]+}}@main
568 // CHECK2-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
569 // CHECK2-NEXT: entry:
570 // CHECK2-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
571 // CHECK2-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
572 // CHECK2-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
573 // CHECK2-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
574 // 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]])
575 // CHECK2-NEXT: ret i32 0
576 //
577 //
578 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
579 // 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]+]] {
580 // CHECK2-NEXT: entry:
581 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
582 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
583 // CHECK2-NEXT: [[ARGC_ADDR:%.*]] = alloca i32*, align 8
584 // CHECK2-NEXT: [[ARGV_ADDR:%.*]] = alloca i8***, align 8
585 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8
586 // CHECK2-NEXT: [[TMP:%.*]] = alloca i64, align 8
587 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8
588 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8
589 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
590 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
591 // CHECK2-NEXT: [[ARGC1:%.*]] = alloca i32, align 4
592 // CHECK2-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
593 // CHECK2-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
594 // CHECK2-NEXT: [[_TMP5:%.*]] = alloca i8**, align 8
595 // CHECK2-NEXT: [[_TMP6:%.*]] = alloca i8*, align 8
596 // CHECK2-NEXT: [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
597 // CHECK2-NEXT: [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
598 // CHECK2-NEXT: [[I:%.*]] = alloca i64, align 8
599 // CHECK2-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
600 // CHECK2-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
601 // CHECK2-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
602 // CHECK2-NEXT: [[_TMP28:%.*]] = alloca i8, align 1
603 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
604 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
605 // CHECK2-NEXT: store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
606 // CHECK2-NEXT: store i8*** [[ARGV]], i8**** [[ARGV_ADDR]], align 8
607 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
608 // CHECK2-NEXT: [[TMP1:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
609 // CHECK2-NEXT: store i64 0, i64* [[DOTOMP_LB]], align 8
610 // CHECK2-NEXT: store i64 9, i64* [[DOTOMP_UB]], align 8
611 // CHECK2-NEXT: store i64 1, i64* [[DOTOMP_STRIDE]], align 8
612 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
613 // CHECK2-NEXT: store i32 0, i32* [[ARGC1]], align 4
614 // CHECK2-NEXT: [[TMP2:%.*]] = load i8**, i8*** [[TMP1]], align 8
615 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i64 0
616 // CHECK2-NEXT: [[TMP3:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
617 // CHECK2-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP3]], i64 0
618 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP0]], align 4
619 // CHECK2-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
620 // CHECK2-NEXT: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP5]]
621 // CHECK2-NEXT: [[TMP6:%.*]] = load i8**, i8*** [[TMP1]], align 8
622 // CHECK2-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP6]], i64 9
623 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
624 // CHECK2-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP7]], i64 [[LB_ADD_LEN]]
625 // CHECK2-NEXT: [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
626 // CHECK2-NEXT: [[TMP9:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
627 // CHECK2-NEXT: [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
628 // CHECK2-NEXT: [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
629 // CHECK2-NEXT: [[TMP12:%.*]] = add nuw i64 [[TMP11]], 1
630 // CHECK2-NEXT: [[TMP13:%.*]] = mul nuw i64 [[TMP12]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
631 // CHECK2-NEXT: [[TMP14:%.*]] = call i8* @llvm.stacksave()
632 // CHECK2-NEXT: store i8* [[TMP14]], i8** [[SAVED_STACK]], align 8
633 // CHECK2-NEXT: [[VLA:%.*]] = alloca i8, i64 [[TMP12]], align 16
634 // CHECK2-NEXT: store i64 [[TMP12]], i64* [[__VLA_EXPR0]], align 8
635 // CHECK2-NEXT: [[TMP15:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP12]]
636 // CHECK2-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP15]]
637 // CHECK2-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
638 // CHECK2: omp.arrayinit.body:
639 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
640 // CHECK2-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
641 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
642 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP15]]
643 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
644 // CHECK2: omp.arrayinit.done:
645 // CHECK2-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[TMP1]], align 8
646 // CHECK2-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
647 // CHECK2-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
648 // CHECK2-NEXT: [[TMP19:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
649 // CHECK2-NEXT: [[TMP20:%.*]] = sub i64 [[TMP18]], [[TMP19]]
650 // CHECK2-NEXT: [[TMP21:%.*]] = sdiv exact i64 [[TMP20]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
651 // CHECK2-NEXT: [[TMP22:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP21]]
652 // CHECK2-NEXT: store i8** [[_TMP6]], i8*** [[_TMP5]], align 8
653 // CHECK2-NEXT: store i8* [[TMP22]], i8** [[_TMP6]], align 8
654 // 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
655 // CHECK2-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
656 // CHECK2-NEXT: [[TMP24:%.*]] = bitcast i32* [[ARGC1]] to i8*
657 // CHECK2-NEXT: store i8* [[TMP24]], i8** [[TMP23]], align 8
658 // CHECK2-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
659 // CHECK2-NEXT: [[TMP26:%.*]] = bitcast i32* [[TMP0]] to i8*
660 // CHECK2-NEXT: store i8* [[TMP26]], i8** [[TMP25]], align 8
661 // CHECK2-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
662 // CHECK2-NEXT: store i64 4, i64* [[TMP27]], align 8
663 // CHECK2-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
664 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP28]], align 8
665 // CHECK2-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
666 // CHECK2-NEXT: store i8* null, i8** [[TMP29]], align 8
667 // CHECK2-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
668 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP30]], align 8
669 // CHECK2-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
670 // CHECK2-NEXT: [[TMP32:%.*]] = bitcast i32* [[TMP31]] to i8*
671 // CHECK2-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP32]], i8 0, i64 4, i1 false)
672 // CHECK2-NEXT: [[DOTRD_INPUT_GEP_7:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t], [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 1
673 // CHECK2-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 0
674 // CHECK2-NEXT: [[TMP34:%.*]] = load i8**, i8*** [[TMP1]], align 8
675 // CHECK2-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8*, i8** [[TMP34]], i64 0
676 // CHECK2-NEXT: [[TMP35:%.*]] = load i8*, i8** [[ARRAYIDX8]], align 8
677 // CHECK2-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds i8, i8* [[TMP35]], i64 0
678 // CHECK2-NEXT: [[TMP36:%.*]] = load i32, i32* [[TMP0]], align 4
679 // CHECK2-NEXT: [[TMP37:%.*]] = sext i32 [[TMP36]] to i64
680 // CHECK2-NEXT: [[LB_ADD_LEN10:%.*]] = add nsw i64 -1, [[TMP37]]
681 // CHECK2-NEXT: [[TMP38:%.*]] = load i8**, i8*** [[TMP1]], align 8
682 // CHECK2-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8*, i8** [[TMP38]], i64 9
683 // CHECK2-NEXT: [[TMP39:%.*]] = load i8*, i8** [[ARRAYIDX11]], align 8
684 // CHECK2-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds i8, i8* [[TMP39]], i64 [[LB_ADD_LEN10]]
685 // CHECK2-NEXT: store i8* [[VLA]], i8** [[TMP33]], align 8
686 // CHECK2-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 1
687 // CHECK2-NEXT: store i8* [[ARRAYIDX9]], i8** [[TMP40]], align 8
688 // CHECK2-NEXT: [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX12]] to i64
689 // CHECK2-NEXT: [[TMP42:%.*]] = ptrtoint i8* [[ARRAYIDX9]] to i64
690 // CHECK2-NEXT: [[TMP43:%.*]] = sub i64 [[TMP41]], [[TMP42]]
691 // CHECK2-NEXT: [[TMP44:%.*]] = sdiv exact i64 [[TMP43]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
692 // CHECK2-NEXT: [[TMP45:%.*]] = add nuw i64 [[TMP44]], 1
693 // CHECK2-NEXT: [[TMP46:%.*]] = mul nuw i64 [[TMP45]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
694 // CHECK2-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 2
695 // CHECK2-NEXT: store i64 [[TMP46]], i64* [[TMP47]], align 8
696 // CHECK2-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 3
697 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP48]], align 8
698 // CHECK2-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 4
699 // CHECK2-NEXT: store i8* null, i8** [[TMP49]], align 8
700 // CHECK2-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 5
701 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP50]], align 8
702 // CHECK2-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 6
703 // CHECK2-NEXT: store i32 1, i32* [[TMP51]], align 8
704 // CHECK2-NEXT: [[TMP52:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
705 // CHECK2-NEXT: [[TMP53:%.*]] = load i32, i32* [[TMP52]], align 4
706 // CHECK2-NEXT: [[TMP54:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
707 // CHECK2-NEXT: [[TMP55:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP53]], i32 1, i32 2, i8* [[TMP54]])
708 // CHECK2-NEXT: store i8* [[TMP55]], i8** [[DOTTASK_RED_]], align 8
709 // CHECK2-NEXT: [[TMP56:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
710 // CHECK2-NEXT: [[TMP57:%.*]] = load i32, i32* [[TMP56]], align 4
711 // CHECK2-NEXT: call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP57]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
712 // CHECK2-NEXT: [[TMP58:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
713 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i64 [[TMP58]], 9
714 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
715 // CHECK2: cond.true:
716 // CHECK2-NEXT: br label [[COND_END:%.*]]
717 // CHECK2: cond.false:
718 // CHECK2-NEXT: [[TMP59:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
719 // CHECK2-NEXT: br label [[COND_END]]
720 // CHECK2: cond.end:
721 // CHECK2-NEXT: [[COND:%.*]] = phi i64 [ 9, [[COND_TRUE]] ], [ [[TMP59]], [[COND_FALSE]] ]
722 // CHECK2-NEXT: store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
723 // CHECK2-NEXT: [[TMP60:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
724 // CHECK2-NEXT: store i64 [[TMP60]], i64* [[DOTOMP_IV]], align 8
725 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
726 // CHECK2: omp.inner.for.cond:
727 // CHECK2-NEXT: [[TMP61:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
728 // CHECK2-NEXT: [[TMP62:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
729 // CHECK2-NEXT: [[CMP13:%.*]] = icmp sle i64 [[TMP61]], [[TMP62]]
730 // CHECK2-NEXT: br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
731 // CHECK2: omp.inner.for.cond.cleanup:
732 // CHECK2-NEXT: br label [[OMP_INNER_FOR_END:%.*]]
733 // CHECK2: omp.inner.for.body:
734 // CHECK2-NEXT: [[TMP63:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
735 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i64 [[TMP63]], 1
736 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i64 0, [[MUL]]
737 // CHECK2-NEXT: store i64 [[ADD]], i64* [[I]], align 8
738 // CHECK2-NEXT: [[TMP64:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
739 // CHECK2-NEXT: store i8** [[DOTTASK_RED_]], i8*** [[TMP64]], align 8
740 // CHECK2-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
741 // CHECK2-NEXT: store i32* [[ARGC1]], i32** [[TMP65]], align 8
742 // CHECK2-NEXT: [[TMP66:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
743 // CHECK2-NEXT: [[TMP67:%.*]] = load i8**, i8*** [[_TMP5]], align 8
744 // CHECK2-NEXT: store i8** [[TMP67]], i8*** [[TMP66]], align 8
745 // CHECK2-NEXT: [[TMP68:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
746 // CHECK2-NEXT: [[TMP69:%.*]] = load i32, i32* [[TMP68]], align 4
747 // CHECK2-NEXT: [[TMP70:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP69]], i32 1, i64 48, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
748 // CHECK2-NEXT: [[TMP71:%.*]] = bitcast i8* [[TMP70]] to %struct.kmp_task_t_with_privates*
749 // CHECK2-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP71]], i32 0, i32 0
750 // CHECK2-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP72]], i32 0, i32 0
751 // CHECK2-NEXT: [[TMP74:%.*]] = load i8*, i8** [[TMP73]], align 8
752 // CHECK2-NEXT: [[TMP75:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
753 // CHECK2-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP74]], i8* align 8 [[TMP75]], i64 24, i1 false)
754 // CHECK2-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP71]], i32 0, i32 1
755 // CHECK2-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP76]], i32 0, i32 0
756 // CHECK2-NEXT: [[TMP78:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
757 // CHECK2-NEXT: store i8* [[TMP78]], i8** [[TMP77]], align 8
758 // CHECK2-NEXT: [[TMP79:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
759 // CHECK2-NEXT: [[TMP80:%.*]] = load i32, i32* [[TMP79]], align 4
760 // CHECK2-NEXT: [[TMP81:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP80]], i8* [[TMP70]])
761 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
762 // CHECK2: omp.body.continue:
763 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
764 // CHECK2: omp.inner.for.inc:
765 // CHECK2-NEXT: [[TMP82:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
766 // CHECK2-NEXT: [[ADD14:%.*]] = add nsw i64 [[TMP82]], 1
767 // CHECK2-NEXT: store i64 [[ADD14]], i64* [[DOTOMP_IV]], align 8
768 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
769 // CHECK2: omp.inner.for.end:
770 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
771 // CHECK2: omp.loop.exit:
772 // CHECK2-NEXT: [[TMP83:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
773 // CHECK2-NEXT: [[TMP84:%.*]] = load i32, i32* [[TMP83]], align 4
774 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP84]])
775 // CHECK2-NEXT: [[TMP85:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
776 // CHECK2-NEXT: [[TMP86:%.*]] = load i32, i32* [[TMP85]], align 4
777 // CHECK2-NEXT: call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP86]], i32 1)
778 // CHECK2-NEXT: [[TMP87:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
779 // CHECK2-NEXT: [[TMP88:%.*]] = bitcast i32* [[ARGC1]] to i8*
780 // CHECK2-NEXT: store i8* [[TMP88]], i8** [[TMP87]], align 8
781 // CHECK2-NEXT: [[TMP89:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
782 // CHECK2-NEXT: store i8* [[VLA]], i8** [[TMP89]], align 8
783 // CHECK2-NEXT: [[TMP90:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
784 // CHECK2-NEXT: [[TMP91:%.*]] = inttoptr i64 [[TMP12]] to i8*
785 // CHECK2-NEXT: store i8* [[TMP91]], i8** [[TMP90]], align 8
786 // CHECK2-NEXT: [[TMP92:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
787 // CHECK2-NEXT: [[TMP93:%.*]] = load i32, i32* [[TMP92]], align 4
788 // CHECK2-NEXT: [[TMP94:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
789 // CHECK2-NEXT: [[TMP95:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP93]], i32 2, i64 24, i8* [[TMP94]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
790 // CHECK2-NEXT: switch i32 [[TMP95]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
791 // CHECK2-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
792 // CHECK2-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
793 // CHECK2-NEXT: ]
794 // CHECK2: .omp.reduction.case1:
795 // CHECK2-NEXT: [[TMP96:%.*]] = load i32, i32* [[TMP0]], align 4
796 // CHECK2-NEXT: [[TMP97:%.*]] = load i32, i32* [[ARGC1]], align 4
797 // CHECK2-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP96]], [[TMP97]]
798 // CHECK2-NEXT: store i32 [[ADD15]], i32* [[TMP0]], align 4
799 // CHECK2-NEXT: [[TMP98:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP12]]
800 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP98]]
801 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE22:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
802 // CHECK2: omp.arraycpy.body:
803 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
804 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST16:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT20:%.*]], [[OMP_ARRAYCPY_BODY]] ]
805 // CHECK2-NEXT: [[TMP99:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
806 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP99]] to i32
807 // CHECK2-NEXT: [[TMP100:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
808 // CHECK2-NEXT: [[CONV17:%.*]] = sext i8 [[TMP100]] to i32
809 // CHECK2-NEXT: [[ADD18:%.*]] = add nsw i32 [[CONV]], [[CONV17]]
810 // CHECK2-NEXT: [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
811 // CHECK2-NEXT: store i8 [[CONV19]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
812 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT20]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], i32 1
813 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
814 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE21:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT20]], [[TMP98]]
815 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_DONE22]], label [[OMP_ARRAYCPY_BODY]]
816 // CHECK2: omp.arraycpy.done22:
817 // CHECK2-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB3]], i32 [[TMP93]], [8 x i32]* @.gomp_critical_user_.reduction.var)
818 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
819 // CHECK2: .omp.reduction.case2:
820 // CHECK2-NEXT: [[TMP101:%.*]] = load i32, i32* [[ARGC1]], align 4
821 // CHECK2-NEXT: [[TMP102:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP101]] monotonic, align 4
822 // CHECK2-NEXT: [[TMP103:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP12]]
823 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY23:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP103]]
824 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY23]], label [[OMP_ARRAYCPY_DONE36:%.*]], label [[OMP_ARRAYCPY_BODY24:%.*]]
825 // CHECK2: omp.arraycpy.body24:
826 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST25:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT34:%.*]], [[ATOMIC_EXIT:%.*]] ]
827 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST26:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT33:%.*]], [[ATOMIC_EXIT]] ]
828 // CHECK2-NEXT: [[TMP104:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
829 // CHECK2-NEXT: [[CONV27:%.*]] = sext i8 [[TMP104]] to i32
830 // CHECK2-NEXT: [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]] monotonic, align 1
831 // CHECK2-NEXT: br label [[ATOMIC_CONT:%.*]]
832 // CHECK2: atomic_cont:
833 // CHECK2-NEXT: [[TMP105:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY24]] ], [ [[TMP110:%.*]], [[ATOMIC_CONT]] ]
834 // CHECK2-NEXT: store i8 [[TMP105]], i8* [[_TMP28]], align 1
835 // CHECK2-NEXT: [[TMP106:%.*]] = load i8, i8* [[_TMP28]], align 1
836 // CHECK2-NEXT: [[CONV29:%.*]] = sext i8 [[TMP106]] to i32
837 // CHECK2-NEXT: [[TMP107:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
838 // CHECK2-NEXT: [[CONV30:%.*]] = sext i8 [[TMP107]] to i32
839 // CHECK2-NEXT: [[ADD31:%.*]] = add nsw i32 [[CONV29]], [[CONV30]]
840 // CHECK2-NEXT: [[CONV32:%.*]] = trunc i32 [[ADD31]] to i8
841 // CHECK2-NEXT: store i8 [[CONV32]], i8* [[ATOMIC_TEMP]], align 1
842 // CHECK2-NEXT: [[TMP108:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
843 // CHECK2-NEXT: [[TMP109:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i8 [[TMP105]], i8 [[TMP108]] monotonic monotonic, align 1
844 // CHECK2-NEXT: [[TMP110]] = extractvalue { i8, i1 } [[TMP109]], 0
845 // CHECK2-NEXT: [[TMP111:%.*]] = extractvalue { i8, i1 } [[TMP109]], 1
846 // CHECK2-NEXT: br i1 [[TMP111]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
847 // CHECK2: atomic_exit:
848 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT33]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i32 1
849 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT34]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], i32 1
850 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE35:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT33]], [[TMP103]]
851 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE35]], label [[OMP_ARRAYCPY_DONE36]], label [[OMP_ARRAYCPY_BODY24]]
852 // CHECK2: omp.arraycpy.done36:
853 // CHECK2-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB3]], i32 [[TMP93]], [8 x i32]* @.gomp_critical_user_.reduction.var)
854 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
855 // CHECK2: .omp.reduction.default:
856 // CHECK2-NEXT: [[TMP112:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
857 // CHECK2-NEXT: call void @llvm.stackrestore(i8* [[TMP112]])
858 // CHECK2-NEXT: [[TMP113:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
859 // CHECK2-NEXT: [[TMP114:%.*]] = load i32, i32* [[TMP113]], align 4
860 // CHECK2-NEXT: call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP114]])
861 // CHECK2-NEXT: ret void
862 //
863 //
864 // CHECK2-LABEL: define {{[^@]+}}@.red_init.
865 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
866 // CHECK2-NEXT: entry:
867 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
868 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
869 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
870 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
871 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
872 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
873 // CHECK2-NEXT: store i32 0, i32* [[TMP3]], align 8
874 // CHECK2-NEXT: ret void
875 //
876 //
877 // CHECK2-LABEL: define {{[^@]+}}@.red_comb.
878 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
879 // CHECK2-NEXT: entry:
880 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
881 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
882 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
883 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
884 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
885 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
886 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
887 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
888 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
889 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
890 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
891 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP3]], align 8
892 // CHECK2-NEXT: ret void
893 //
894 //
895 // CHECK2-LABEL: define {{[^@]+}}@.red_init..1
896 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
897 // CHECK2-NEXT: entry:
898 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
899 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
900 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
901 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
902 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
903 // CHECK2-NEXT: [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
904 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
905 // CHECK2-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
906 // CHECK2-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
907 // CHECK2: omp.arrayinit.body:
908 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
909 // CHECK2-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
910 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
911 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
912 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
913 // CHECK2: omp.arrayinit.done:
914 // CHECK2-NEXT: ret void
915 //
916 //
917 // CHECK2-LABEL: define {{[^@]+}}@.red_comb..2
918 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
919 // CHECK2-NEXT: entry:
920 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
921 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
922 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
923 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
924 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
925 // CHECK2-NEXT: [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
926 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
927 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
928 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
929 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
930 // CHECK2: omp.arraycpy.body:
931 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
932 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
933 // CHECK2-NEXT: [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
934 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP6]] to i32
935 // CHECK2-NEXT: [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
936 // CHECK2-NEXT: [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
937 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
938 // CHECK2-NEXT: [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
939 // CHECK2-NEXT: store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
940 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
941 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
942 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
943 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
944 // CHECK2: omp.arraycpy.done4:
945 // CHECK2-NEXT: ret void
946 //
947 //
948 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_privates_map.
949 // CHECK2-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
950 // CHECK2-NEXT: entry:
951 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
952 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8***, align 8
953 // CHECK2-NEXT: store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
954 // CHECK2-NEXT: store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
955 // CHECK2-NEXT: [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
956 // CHECK2-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
957 // CHECK2-NEXT: [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
958 // CHECK2-NEXT: store i8** [[TMP3]], i8*** [[TMP4]], align 8
959 // CHECK2-NEXT: ret void
960 //
961 //
962 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
963 // CHECK2-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
964 // CHECK2-NEXT: entry:
965 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
966 // CHECK2-NEXT: [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
967 // CHECK2-NEXT: [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
968 // CHECK2-NEXT: [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
969 // CHECK2-NEXT: [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
970 // CHECK2-NEXT: [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
971 // CHECK2-NEXT: [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
972 // CHECK2-NEXT: [[TMP_I:%.*]] = alloca i8**, align 8
973 // CHECK2-NEXT: [[TMP4_I:%.*]] = alloca i8*, align 8
974 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i32, align 4
975 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
976 // CHECK2-NEXT: store i32 [[TMP0]], i32* [[DOTADDR]], align 4
977 // CHECK2-NEXT: store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
978 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
979 // CHECK2-NEXT: [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
980 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
981 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
982 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
983 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
984 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
985 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
986 // CHECK2-NEXT: [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
987 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
988 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
989 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
990 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
991 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
992 // CHECK2-NEXT: store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
993 // CHECK2-NEXT: store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
994 // CHECK2-NEXT: store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
995 // 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
996 // CHECK2-NEXT: store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
997 // CHECK2-NEXT: store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
998 // CHECK2-NEXT: [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
999 // CHECK2-NEXT: [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
1000 // CHECK2-NEXT: [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
1001 // CHECK2-NEXT: [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
1002 // CHECK2-NEXT: call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
1003 // CHECK2-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
1004 // CHECK2-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
1005 // CHECK2-NEXT: [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
1006 // CHECK2-NEXT: [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
1007 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
1008 // CHECK2-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
1009 // CHECK2-NEXT: [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
1010 // CHECK2-NEXT: [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
1011 // CHECK2-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
1012 // CHECK2-NEXT: [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
1013 // CHECK2-NEXT: [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
1014 // CHECK2-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
1015 // CHECK2-NEXT: [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
1016 // CHECK2-NEXT: [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
1017 // CHECK2-NEXT: [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
1018 // CHECK2-NEXT: [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
1019 // CHECK2-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
1020 // CHECK2-NEXT: [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
1021 // CHECK2-NEXT: [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
1022 // CHECK2-NEXT: [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
1023 // CHECK2-NEXT: [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
1024 // CHECK2-NEXT: [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
1025 // CHECK2-NEXT: [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
1026 // CHECK2-NEXT: [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
1027 // CHECK2-NEXT: [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1028 // CHECK2-NEXT: [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
1029 // CHECK2-NEXT: [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1030 // CHECK2-NEXT: store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
1031 // CHECK2-NEXT: [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
1032 // CHECK2-NEXT: [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
1033 // CHECK2-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
1034 // CHECK2-NEXT: [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
1035 // CHECK2-NEXT: [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
1036 // CHECK2-NEXT: [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
1037 // CHECK2-NEXT: [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
1038 // CHECK2-NEXT: [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
1039 // CHECK2-NEXT: [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1040 // CHECK2-NEXT: [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
1041 // CHECK2-NEXT: store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
1042 // CHECK2-NEXT: store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
1043 // CHECK2-NEXT: ret i32 0
1044 //
1045 //
1046 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
1047 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
1048 // CHECK2-NEXT: entry:
1049 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1050 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1051 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1052 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1053 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1054 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
1055 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1056 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
1057 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
1058 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1059 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1060 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
1061 // CHECK2-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1062 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1063 // CHECK2-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
1064 // CHECK2-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
1065 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
1066 // CHECK2-NEXT: [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
1067 // CHECK2-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
1068 // CHECK2-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
1069 // CHECK2-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
1070 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
1071 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
1072 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
1073 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
1074 // CHECK2-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
1075 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
1076 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1077 // CHECK2: omp.arraycpy.body:
1078 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1079 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1080 // CHECK2-NEXT: [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1081 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP22]] to i32
1082 // CHECK2-NEXT: [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
1083 // CHECK2-NEXT: [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
1084 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
1085 // CHECK2-NEXT: [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
1086 // CHECK2-NEXT: store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1087 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1088 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1089 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
1090 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
1091 // CHECK2: omp.arraycpy.done5:
1092 // CHECK2-NEXT: ret void
1093 //
1094