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 // Test host codegen.
3 // RUN: %clang_cc1 -no-opaque-pointers -DHAS_INT128 -verify -fopenmp -fopenmp-version=50 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
4 // RUN: %clang_cc1 -no-opaque-pointers -DHAS_INT128 -fopenmp -fopenmp-version=50 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
5 // RUN: %clang_cc1 -no-opaque-pointers -DHAS_INT128 -fopenmp -fopenmp-version=50 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
6 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=50 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
7 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=50 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
8 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=50 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
9
10 // Test target codegen - host bc file has to be created first.
11 // RUN: %clang_cc1 -no-opaque-pointers -DHAS_INT128 -verify -fopenmp -fopenmp-version=50 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
12 // RUN: %clang_cc1 -no-opaque-pointers -DHAS_INT128 -verify -fopenmp -fopenmp-version=50 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=CHECK5
13 // RUN: %clang_cc1 -no-opaque-pointers -DHAS_INT128 -fopenmp -fopenmp-version=50 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
14 // RUN: %clang_cc1 -no-opaque-pointers -DHAS_INT128 -fopenmp -fopenmp-version=50 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK5
15 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=50 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
16 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=50 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix=CHECK7
17 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=50 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
18 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=50 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK7
19
20 // expected-no-diagnostics
21 #ifndef HEADER
22 #define HEADER
23
24
mapWithPrivate()25 void mapWithPrivate() {
26 int x, y;
27 #pragma omp target teams private(x) map(x,y) private(y)
28 ;
29 }
30
mapWithFirstprivate()31 void mapWithFirstprivate() {
32 int x, y;
33 #pragma omp target teams firstprivate(x) map(x,y) firstprivate(y)
34 ;
35 }
36
mapWithReduction()37 void mapWithReduction() {
38 int x, y;
39 #pragma omp target teams reduction(+:x) map(x,y) reduction(+:y)
40 ;
41 }
42
mapFrom()43 void mapFrom() {
44 int x;
45 #pragma omp target teams firstprivate(x) map(from:x)
46 ;
47 }
48
mapTo()49 void mapTo() {
50 int x;
51 #pragma omp target teams firstprivate(x) map(to:x)
52 ;
53 }
54
mapAlloc()55 void mapAlloc() {
56 int x;
57 #pragma omp target teams firstprivate(x) map(alloc:x)
58 ;
59 }
60
mapArray()61 void mapArray() {
62 int x[77], y[88], z[99];
63 #pragma omp target teams private(x) firstprivate(y) reduction(+:z) map(x,y,z)
64 ;
65 #pragma omp target teams private(x) firstprivate(y) reduction(+:z) map(to:x,y,z)
66 ;
67 }
68
69 # if HAS_INT128
mapInt128()70 void mapInt128() {
71 __int128 x, y, z;
72 #pragma omp target teams private(x) firstprivate(y) reduction(+:z) map(x,y,z)
73 ;
74 #pragma omp target teams private(x) firstprivate(y) reduction(+:z) map(from:x,y,z)
75 ;
76 }
77 # endif
78 #endif
79 // CHECK1-LABEL: define {{[^@]+}}@_Z14mapWithPrivatev
80 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
81 // CHECK1-NEXT: entry:
82 // CHECK1-NEXT: [[X:%.*]] = alloca i32, align 4
83 // CHECK1-NEXT: [[Y:%.*]] = alloca i32, align 4
84 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 8
85 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 8
86 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 8
87 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
88 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32**
89 // CHECK1-NEXT: store i32* [[X]], i32** [[TMP1]], align 8
90 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
91 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
92 // CHECK1-NEXT: store i32* [[X]], i32** [[TMP3]], align 8
93 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
94 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
95 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
96 // CHECK1-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32**
97 // CHECK1-NEXT: store i32* [[Y]], i32** [[TMP6]], align 8
98 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
99 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32**
100 // CHECK1-NEXT: store i32* [[Y]], i32** [[TMP8]], align 8
101 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
102 // CHECK1-NEXT: store i8* null, i8** [[TMP9]], align 8
103 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
104 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
105 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
106 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
107 // CHECK1-NEXT: store i32 1, i32* [[TMP12]], align 4
108 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
109 // CHECK1-NEXT: store i32 2, i32* [[TMP13]], align 4
110 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
111 // CHECK1-NEXT: store i8** [[TMP10]], i8*** [[TMP14]], align 8
112 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
113 // CHECK1-NEXT: store i8** [[TMP11]], i8*** [[TMP15]], align 8
114 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
115 // CHECK1-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP16]], align 8
116 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
117 // CHECK1-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP17]], align 8
118 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
119 // CHECK1-NEXT: store i8** null, i8*** [[TMP18]], align 8
120 // CHECK1-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
121 // CHECK1-NEXT: store i8** null, i8*** [[TMP19]], align 8
122 // CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
123 // CHECK1-NEXT: store i64 0, i64* [[TMP20]], align 8
124 // CHECK1-NEXT: [[TMP21:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z14mapWithPrivatev_l27.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
125 // CHECK1-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
126 // CHECK1-NEXT: br i1 [[TMP22]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
127 // CHECK1: omp_offload.failed:
128 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z14mapWithPrivatev_l27() #[[ATTR2:[0-9]+]]
129 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
130 // CHECK1: omp_offload.cont:
131 // CHECK1-NEXT: ret void
132 //
133 //
134 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z14mapWithPrivatev_l27
135 // CHECK1-SAME: () #[[ATTR1:[0-9]+]] {
136 // CHECK1-NEXT: entry:
137 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
138 // CHECK1-NEXT: ret void
139 //
140 //
141 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
142 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
143 // CHECK1-NEXT: entry:
144 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
145 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
146 // CHECK1-NEXT: [[X:%.*]] = alloca i32, align 4
147 // CHECK1-NEXT: [[Y:%.*]] = alloca i32, align 4
148 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
149 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
150 // CHECK1-NEXT: ret void
151 //
152 //
153 // CHECK1-LABEL: define {{[^@]+}}@_Z19mapWithFirstprivatev
154 // CHECK1-SAME: () #[[ATTR0]] {
155 // CHECK1-NEXT: entry:
156 // CHECK1-NEXT: [[X:%.*]] = alloca i32, align 4
157 // CHECK1-NEXT: [[Y:%.*]] = alloca i32, align 4
158 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 8
159 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 8
160 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 8
161 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
162 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32**
163 // CHECK1-NEXT: store i32* [[X]], i32** [[TMP1]], align 8
164 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
165 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
166 // CHECK1-NEXT: store i32* [[X]], i32** [[TMP3]], align 8
167 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
168 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
169 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
170 // CHECK1-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32**
171 // CHECK1-NEXT: store i32* [[Y]], i32** [[TMP6]], align 8
172 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
173 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32**
174 // CHECK1-NEXT: store i32* [[Y]], i32** [[TMP8]], align 8
175 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
176 // CHECK1-NEXT: store i8* null, i8** [[TMP9]], align 8
177 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
178 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
179 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
180 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
181 // CHECK1-NEXT: store i32 1, i32* [[TMP12]], align 4
182 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
183 // CHECK1-NEXT: store i32 2, i32* [[TMP13]], align 4
184 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
185 // CHECK1-NEXT: store i8** [[TMP10]], i8*** [[TMP14]], align 8
186 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
187 // CHECK1-NEXT: store i8** [[TMP11]], i8*** [[TMP15]], align 8
188 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
189 // CHECK1-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP16]], align 8
190 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
191 // CHECK1-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP17]], align 8
192 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
193 // CHECK1-NEXT: store i8** null, i8*** [[TMP18]], align 8
194 // CHECK1-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
195 // CHECK1-NEXT: store i8** null, i8*** [[TMP19]], align 8
196 // CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
197 // CHECK1-NEXT: store i64 0, i64* [[TMP20]], align 8
198 // CHECK1-NEXT: [[TMP21:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z19mapWithFirstprivatev_l33.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
199 // CHECK1-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
200 // CHECK1-NEXT: br i1 [[TMP22]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
201 // CHECK1: omp_offload.failed:
202 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z19mapWithFirstprivatev_l33(i32* [[X]], i32* [[Y]]) #[[ATTR2]]
203 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
204 // CHECK1: omp_offload.cont:
205 // CHECK1-NEXT: ret void
206 //
207 //
208 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z19mapWithFirstprivatev_l33
209 // CHECK1-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[Y:%.*]]) #[[ATTR1]] {
210 // CHECK1-NEXT: entry:
211 // CHECK1-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 8
212 // CHECK1-NEXT: [[Y_ADDR:%.*]] = alloca i32*, align 8
213 // CHECK1-NEXT: [[X_CASTED:%.*]] = alloca i64, align 8
214 // CHECK1-NEXT: [[Y_CASTED:%.*]] = alloca i64, align 8
215 // CHECK1-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 8
216 // CHECK1-NEXT: store i32* [[Y]], i32** [[Y_ADDR]], align 8
217 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 8
218 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[Y_ADDR]], align 8
219 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
220 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32*
221 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[CONV]], align 4
222 // CHECK1-NEXT: [[TMP3:%.*]] = load i64, i64* [[X_CASTED]], align 8
223 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP1]], align 4
224 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[Y_CASTED]] to i32*
225 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[CONV1]], align 4
226 // CHECK1-NEXT: [[TMP5:%.*]] = load i64, i64* [[Y_CASTED]], align 8
227 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP3]], i64 [[TMP5]])
228 // CHECK1-NEXT: ret void
229 //
230 //
231 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
232 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[X:%.*]], i64 noundef [[Y:%.*]]) #[[ATTR1]] {
233 // CHECK1-NEXT: entry:
234 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
235 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
236 // CHECK1-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8
237 // CHECK1-NEXT: [[Y_ADDR:%.*]] = alloca i64, align 8
238 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
239 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
240 // CHECK1-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8
241 // CHECK1-NEXT: store i64 [[Y]], i64* [[Y_ADDR]], align 8
242 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32*
243 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[Y_ADDR]] to i32*
244 // CHECK1-NEXT: ret void
245 //
246 //
247 // CHECK1-LABEL: define {{[^@]+}}@_Z16mapWithReductionv
248 // CHECK1-SAME: () #[[ATTR0]] {
249 // CHECK1-NEXT: entry:
250 // CHECK1-NEXT: [[X:%.*]] = alloca i32, align 4
251 // CHECK1-NEXT: [[Y:%.*]] = alloca i32, align 4
252 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 8
253 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 8
254 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 8
255 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
256 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32**
257 // CHECK1-NEXT: store i32* [[X]], i32** [[TMP1]], align 8
258 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
259 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
260 // CHECK1-NEXT: store i32* [[X]], i32** [[TMP3]], align 8
261 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
262 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
263 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
264 // CHECK1-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32**
265 // CHECK1-NEXT: store i32* [[Y]], i32** [[TMP6]], align 8
266 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
267 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32**
268 // CHECK1-NEXT: store i32* [[Y]], i32** [[TMP8]], align 8
269 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
270 // CHECK1-NEXT: store i8* null, i8** [[TMP9]], align 8
271 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
272 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
273 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
274 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
275 // CHECK1-NEXT: store i32 1, i32* [[TMP12]], align 4
276 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
277 // CHECK1-NEXT: store i32 2, i32* [[TMP13]], align 4
278 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
279 // CHECK1-NEXT: store i8** [[TMP10]], i8*** [[TMP14]], align 8
280 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
281 // CHECK1-NEXT: store i8** [[TMP11]], i8*** [[TMP15]], align 8
282 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
283 // CHECK1-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64** [[TMP16]], align 8
284 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
285 // CHECK1-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i64** [[TMP17]], align 8
286 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
287 // CHECK1-NEXT: store i8** null, i8*** [[TMP18]], align 8
288 // CHECK1-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
289 // CHECK1-NEXT: store i8** null, i8*** [[TMP19]], align 8
290 // CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
291 // CHECK1-NEXT: store i64 0, i64* [[TMP20]], align 8
292 // CHECK1-NEXT: [[TMP21:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16mapWithReductionv_l39.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
293 // CHECK1-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
294 // CHECK1-NEXT: br i1 [[TMP22]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
295 // CHECK1: omp_offload.failed:
296 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16mapWithReductionv_l39(i32* [[X]], i32* [[Y]]) #[[ATTR2]]
297 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
298 // CHECK1: omp_offload.cont:
299 // CHECK1-NEXT: ret void
300 //
301 //
302 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16mapWithReductionv_l39
303 // CHECK1-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[Y:%.*]]) #[[ATTR1]] {
304 // CHECK1-NEXT: entry:
305 // CHECK1-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 8
306 // CHECK1-NEXT: [[Y_ADDR:%.*]] = alloca i32*, align 8
307 // CHECK1-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 8
308 // CHECK1-NEXT: store i32* [[Y]], i32** [[Y_ADDR]], align 8
309 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 8
310 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[Y_ADDR]], align 8
311 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32* [[TMP0]], i32* [[TMP1]])
312 // CHECK1-NEXT: ret void
313 //
314 //
315 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
316 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[Y:%.*]]) #[[ATTR1]] {
317 // CHECK1-NEXT: entry:
318 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
319 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
320 // CHECK1-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 8
321 // CHECK1-NEXT: [[Y_ADDR:%.*]] = alloca i32*, align 8
322 // CHECK1-NEXT: [[X1:%.*]] = alloca i32, align 4
323 // CHECK1-NEXT: [[Y2:%.*]] = alloca i32, align 4
324 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 8
325 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
326 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
327 // CHECK1-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 8
328 // CHECK1-NEXT: store i32* [[Y]], i32** [[Y_ADDR]], align 8
329 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 8
330 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[Y_ADDR]], align 8
331 // CHECK1-NEXT: store i32 0, i32* [[X1]], align 4
332 // CHECK1-NEXT: store i32 0, i32* [[Y2]], align 4
333 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
334 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i32* [[X1]] to i8*
335 // CHECK1-NEXT: store i8* [[TMP3]], i8** [[TMP2]], align 8
336 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
337 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i32* [[Y2]] to i8*
338 // CHECK1-NEXT: store i8* [[TMP5]], i8** [[TMP4]], align 8
339 // CHECK1-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
340 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
341 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
342 // CHECK1-NEXT: [[TMP9:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP7]], i32 2, i64 16, i8* [[TMP8]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
343 // CHECK1-NEXT: switch i32 [[TMP9]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
344 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
345 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
346 // CHECK1-NEXT: ]
347 // CHECK1: .omp.reduction.case1:
348 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP0]], align 4
349 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[X1]], align 4
350 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
351 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP0]], align 4
352 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP1]], align 4
353 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[Y2]], align 4
354 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
355 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[TMP1]], align 4
356 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP7]], [8 x i32]* @.gomp_critical_user_.reduction.var)
357 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
358 // CHECK1: .omp.reduction.case2:
359 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[X1]], align 4
360 // CHECK1-NEXT: [[TMP15:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP14]] monotonic, align 4
361 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[Y2]], align 4
362 // CHECK1-NEXT: [[TMP17:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP16]] monotonic, align 4
363 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP7]], [8 x i32]* @.gomp_critical_user_.reduction.var)
364 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
365 // CHECK1: .omp.reduction.default:
366 // CHECK1-NEXT: ret void
367 //
368 //
369 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
370 // CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
371 // CHECK1-NEXT: entry:
372 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
373 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
374 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
375 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
376 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
377 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [2 x i8*]*
378 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
379 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
380 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
381 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
382 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
383 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 0
384 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
385 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
386 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 1
387 // CHECK1-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
388 // CHECK1-NEXT: [[TMP14:%.*]] = bitcast i8* [[TMP13]] to i32*
389 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 1
390 // CHECK1-NEXT: [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
391 // CHECK1-NEXT: [[TMP17:%.*]] = bitcast i8* [[TMP16]] to i32*
392 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP11]], align 4
393 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP8]], align 4
394 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
395 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
396 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP17]], align 4
397 // CHECK1-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP14]], align 4
398 // CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
399 // CHECK1-NEXT: store i32 [[ADD2]], i32* [[TMP17]], align 4
400 // CHECK1-NEXT: ret void
401 //
402 //
403 // CHECK1-LABEL: define {{[^@]+}}@_Z7mapFromv
404 // CHECK1-SAME: () #[[ATTR0]] {
405 // CHECK1-NEXT: entry:
406 // CHECK1-NEXT: [[X:%.*]] = alloca i32, align 4
407 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
408 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
409 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
410 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
411 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32**
412 // CHECK1-NEXT: store i32* [[X]], i32** [[TMP1]], align 8
413 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
414 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
415 // CHECK1-NEXT: store i32* [[X]], i32** [[TMP3]], align 8
416 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
417 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
418 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
419 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
420 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
421 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
422 // CHECK1-NEXT: store i32 1, i32* [[TMP7]], align 4
423 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
424 // CHECK1-NEXT: store i32 1, i32* [[TMP8]], align 4
425 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
426 // CHECK1-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8
427 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
428 // CHECK1-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8
429 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
430 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP11]], align 8
431 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
432 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP12]], align 8
433 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
434 // CHECK1-NEXT: store i8** null, i8*** [[TMP13]], align 8
435 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
436 // CHECK1-NEXT: store i8** null, i8*** [[TMP14]], align 8
437 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
438 // CHECK1-NEXT: store i64 0, i64* [[TMP15]], align 8
439 // CHECK1-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z7mapFromv_l45.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
440 // CHECK1-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
441 // CHECK1-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
442 // CHECK1: omp_offload.failed:
443 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z7mapFromv_l45(i32* [[X]]) #[[ATTR2]]
444 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
445 // CHECK1: omp_offload.cont:
446 // CHECK1-NEXT: ret void
447 //
448 //
449 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z7mapFromv_l45
450 // CHECK1-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
451 // CHECK1-NEXT: entry:
452 // CHECK1-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 8
453 // CHECK1-NEXT: [[X_CASTED:%.*]] = alloca i64, align 8
454 // CHECK1-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 8
455 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 8
456 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
457 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32*
458 // CHECK1-NEXT: store i32 [[TMP1]], i32* [[CONV]], align 4
459 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[X_CASTED]], align 8
460 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP2]])
461 // CHECK1-NEXT: ret void
462 //
463 //
464 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
465 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[X:%.*]]) #[[ATTR1]] {
466 // CHECK1-NEXT: entry:
467 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
468 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
469 // CHECK1-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8
470 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
471 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
472 // CHECK1-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8
473 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32*
474 // CHECK1-NEXT: ret void
475 //
476 //
477 // CHECK1-LABEL: define {{[^@]+}}@_Z5mapTov
478 // CHECK1-SAME: () #[[ATTR0]] {
479 // CHECK1-NEXT: entry:
480 // CHECK1-NEXT: [[X:%.*]] = alloca i32, align 4
481 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
482 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
483 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
484 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
485 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32**
486 // CHECK1-NEXT: store i32* [[X]], i32** [[TMP1]], align 8
487 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
488 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
489 // CHECK1-NEXT: store i32* [[X]], i32** [[TMP3]], align 8
490 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
491 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
492 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
493 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
494 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
495 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
496 // CHECK1-NEXT: store i32 1, i32* [[TMP7]], align 4
497 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
498 // CHECK1-NEXT: store i32 1, i32* [[TMP8]], align 4
499 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
500 // CHECK1-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8
501 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
502 // CHECK1-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8
503 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
504 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.11, i32 0, i32 0), i64** [[TMP11]], align 8
505 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
506 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.12, i32 0, i32 0), i64** [[TMP12]], align 8
507 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
508 // CHECK1-NEXT: store i8** null, i8*** [[TMP13]], align 8
509 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
510 // CHECK1-NEXT: store i8** null, i8*** [[TMP14]], align 8
511 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
512 // CHECK1-NEXT: store i64 0, i64* [[TMP15]], align 8
513 // CHECK1-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5mapTov_l51.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
514 // CHECK1-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
515 // CHECK1-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
516 // CHECK1: omp_offload.failed:
517 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5mapTov_l51(i32* [[X]]) #[[ATTR2]]
518 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
519 // CHECK1: omp_offload.cont:
520 // CHECK1-NEXT: ret void
521 //
522 //
523 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5mapTov_l51
524 // CHECK1-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
525 // CHECK1-NEXT: entry:
526 // CHECK1-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 8
527 // CHECK1-NEXT: [[X_CASTED:%.*]] = alloca i64, align 8
528 // CHECK1-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 8
529 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 8
530 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
531 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32*
532 // CHECK1-NEXT: store i32 [[TMP1]], i32* [[CONV]], align 4
533 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[X_CASTED]], align 8
534 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..10 to void (i32*, i32*, ...)*), i64 [[TMP2]])
535 // CHECK1-NEXT: ret void
536 //
537 //
538 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..10
539 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[X:%.*]]) #[[ATTR1]] {
540 // CHECK1-NEXT: entry:
541 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
542 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
543 // CHECK1-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8
544 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
545 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
546 // CHECK1-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8
547 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32*
548 // CHECK1-NEXT: ret void
549 //
550 //
551 // CHECK1-LABEL: define {{[^@]+}}@_Z8mapAllocv
552 // CHECK1-SAME: () #[[ATTR0]] {
553 // CHECK1-NEXT: entry:
554 // CHECK1-NEXT: [[X:%.*]] = alloca i32, align 4
555 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
556 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
557 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
558 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
559 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32**
560 // CHECK1-NEXT: store i32* [[X]], i32** [[TMP1]], align 8
561 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
562 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
563 // CHECK1-NEXT: store i32* [[X]], i32** [[TMP3]], align 8
564 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
565 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
566 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
567 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
568 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
569 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
570 // CHECK1-NEXT: store i32 1, i32* [[TMP7]], align 4
571 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
572 // CHECK1-NEXT: store i32 1, i32* [[TMP8]], align 4
573 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
574 // CHECK1-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8
575 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
576 // CHECK1-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8
577 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
578 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.14, i32 0, i32 0), i64** [[TMP11]], align 8
579 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
580 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.15, i32 0, i32 0), i64** [[TMP12]], align 8
581 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
582 // CHECK1-NEXT: store i8** null, i8*** [[TMP13]], align 8
583 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
584 // CHECK1-NEXT: store i8** null, i8*** [[TMP14]], align 8
585 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
586 // CHECK1-NEXT: store i64 0, i64* [[TMP15]], align 8
587 // CHECK1-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapAllocv_l57.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
588 // CHECK1-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
589 // CHECK1-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
590 // CHECK1: omp_offload.failed:
591 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapAllocv_l57(i32* [[X]]) #[[ATTR2]]
592 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
593 // CHECK1: omp_offload.cont:
594 // CHECK1-NEXT: ret void
595 //
596 //
597 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapAllocv_l57
598 // CHECK1-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
599 // CHECK1-NEXT: entry:
600 // CHECK1-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 8
601 // CHECK1-NEXT: [[X_CASTED:%.*]] = alloca i64, align 8
602 // CHECK1-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 8
603 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 8
604 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
605 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32*
606 // CHECK1-NEXT: store i32 [[TMP1]], i32* [[CONV]], align 4
607 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[X_CASTED]], align 8
608 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i64 [[TMP2]])
609 // CHECK1-NEXT: ret void
610 //
611 //
612 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..13
613 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[X:%.*]]) #[[ATTR1]] {
614 // CHECK1-NEXT: entry:
615 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
616 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
617 // CHECK1-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8
618 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
619 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
620 // CHECK1-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8
621 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32*
622 // CHECK1-NEXT: ret void
623 //
624 //
625 // CHECK1-LABEL: define {{[^@]+}}@_Z8mapArrayv
626 // CHECK1-SAME: () #[[ATTR0]] {
627 // CHECK1-NEXT: entry:
628 // CHECK1-NEXT: [[X:%.*]] = alloca [77 x i32], align 4
629 // CHECK1-NEXT: [[Y:%.*]] = alloca [88 x i32], align 4
630 // CHECK1-NEXT: [[Z:%.*]] = alloca [99 x i32], align 4
631 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
632 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
633 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
634 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [3 x i8*], align 8
635 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [3 x i8*], align 8
636 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [3 x i8*], align 8
637 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
638 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [88 x i32]**
639 // CHECK1-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[TMP1]], align 8
640 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
641 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [88 x i32]**
642 // CHECK1-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[TMP3]], align 8
643 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
644 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
645 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
646 // CHECK1-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to [99 x i32]**
647 // CHECK1-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[TMP6]], align 8
648 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
649 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [99 x i32]**
650 // CHECK1-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[TMP8]], align 8
651 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
652 // CHECK1-NEXT: store i8* null, i8** [[TMP9]], align 8
653 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
654 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8** [[TMP10]] to [77 x i32]**
655 // CHECK1-NEXT: store [77 x i32]* [[X]], [77 x i32]** [[TMP11]], align 8
656 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
657 // CHECK1-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to [77 x i32]**
658 // CHECK1-NEXT: store [77 x i32]* [[X]], [77 x i32]** [[TMP13]], align 8
659 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
660 // CHECK1-NEXT: store i8* null, i8** [[TMP14]], align 8
661 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
662 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
663 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
664 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
665 // CHECK1-NEXT: store i32 1, i32* [[TMP17]], align 4
666 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
667 // CHECK1-NEXT: store i32 3, i32* [[TMP18]], align 4
668 // CHECK1-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
669 // CHECK1-NEXT: store i8** [[TMP15]], i8*** [[TMP19]], align 8
670 // CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
671 // CHECK1-NEXT: store i8** [[TMP16]], i8*** [[TMP20]], align 8
672 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
673 // CHECK1-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.18, i32 0, i32 0), i64** [[TMP21]], align 8
674 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
675 // CHECK1-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.19, i32 0, i32 0), i64** [[TMP22]], align 8
676 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
677 // CHECK1-NEXT: store i8** null, i8*** [[TMP23]], align 8
678 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
679 // CHECK1-NEXT: store i8** null, i8*** [[TMP24]], align 8
680 // CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
681 // CHECK1-NEXT: store i64 0, i64* [[TMP25]], align 8
682 // CHECK1-NEXT: [[TMP26:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l63.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
683 // CHECK1-NEXT: [[TMP27:%.*]] = icmp ne i32 [[TMP26]], 0
684 // CHECK1-NEXT: br i1 [[TMP27]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
685 // CHECK1: omp_offload.failed:
686 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l63([88 x i32]* [[Y]], [99 x i32]* [[Z]]) #[[ATTR2]]
687 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
688 // CHECK1: omp_offload.cont:
689 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
690 // CHECK1-NEXT: [[TMP29:%.*]] = bitcast i8** [[TMP28]] to [88 x i32]**
691 // CHECK1-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[TMP29]], align 8
692 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
693 // CHECK1-NEXT: [[TMP31:%.*]] = bitcast i8** [[TMP30]] to [88 x i32]**
694 // CHECK1-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[TMP31]], align 8
695 // CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
696 // CHECK1-NEXT: store i8* null, i8** [[TMP32]], align 8
697 // CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 1
698 // CHECK1-NEXT: [[TMP34:%.*]] = bitcast i8** [[TMP33]] to [99 x i32]**
699 // CHECK1-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[TMP34]], align 8
700 // CHECK1-NEXT: [[TMP35:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 1
701 // CHECK1-NEXT: [[TMP36:%.*]] = bitcast i8** [[TMP35]] to [99 x i32]**
702 // CHECK1-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[TMP36]], align 8
703 // CHECK1-NEXT: [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 1
704 // CHECK1-NEXT: store i8* null, i8** [[TMP37]], align 8
705 // CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 2
706 // CHECK1-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [77 x i32]**
707 // CHECK1-NEXT: store [77 x i32]* [[X]], [77 x i32]** [[TMP39]], align 8
708 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 2
709 // CHECK1-NEXT: [[TMP41:%.*]] = bitcast i8** [[TMP40]] to [77 x i32]**
710 // CHECK1-NEXT: store [77 x i32]* [[X]], [77 x i32]** [[TMP41]], align 8
711 // CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 2
712 // CHECK1-NEXT: store i8* null, i8** [[TMP42]], align 8
713 // CHECK1-NEXT: [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
714 // CHECK1-NEXT: [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
715 // CHECK1-NEXT: [[KERNEL_ARGS4:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
716 // CHECK1-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 0
717 // CHECK1-NEXT: store i32 1, i32* [[TMP45]], align 4
718 // CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 1
719 // CHECK1-NEXT: store i32 3, i32* [[TMP46]], align 4
720 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 2
721 // CHECK1-NEXT: store i8** [[TMP43]], i8*** [[TMP47]], align 8
722 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 3
723 // CHECK1-NEXT: store i8** [[TMP44]], i8*** [[TMP48]], align 8
724 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 4
725 // CHECK1-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.22, i32 0, i32 0), i64** [[TMP49]], align 8
726 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 5
727 // CHECK1-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.23, i32 0, i32 0), i64** [[TMP50]], align 8
728 // CHECK1-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 6
729 // CHECK1-NEXT: store i8** null, i8*** [[TMP51]], align 8
730 // CHECK1-NEXT: [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 7
731 // CHECK1-NEXT: store i8** null, i8*** [[TMP52]], align 8
732 // CHECK1-NEXT: [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 8
733 // CHECK1-NEXT: store i64 0, i64* [[TMP53]], align 8
734 // CHECK1-NEXT: [[TMP54:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l65.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]])
735 // CHECK1-NEXT: [[TMP55:%.*]] = icmp ne i32 [[TMP54]], 0
736 // CHECK1-NEXT: br i1 [[TMP55]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
737 // CHECK1: omp_offload.failed5:
738 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l65([88 x i32]* [[Y]], [99 x i32]* [[Z]]) #[[ATTR2]]
739 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT6]]
740 // CHECK1: omp_offload.cont6:
741 // CHECK1-NEXT: ret void
742 //
743 //
744 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l63
745 // CHECK1-SAME: ([88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR1]] {
746 // CHECK1-NEXT: entry:
747 // CHECK1-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 8
748 // CHECK1-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 8
749 // CHECK1-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 8
750 // CHECK1-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 8
751 // CHECK1-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 8
752 // CHECK1-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 8
753 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [88 x i32]*, [99 x i32]*)* @.omp_outlined..16 to void (i32*, i32*, ...)*), [88 x i32]* [[TMP0]], [99 x i32]* [[TMP1]])
754 // CHECK1-NEXT: ret void
755 //
756 //
757 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..16
758 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR1]] {
759 // CHECK1-NEXT: entry:
760 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
761 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
762 // CHECK1-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 8
763 // CHECK1-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 8
764 // CHECK1-NEXT: [[Y1:%.*]] = alloca [88 x i32], align 4
765 // CHECK1-NEXT: [[X:%.*]] = alloca [77 x i32], align 4
766 // CHECK1-NEXT: [[Z2:%.*]] = alloca [99 x i32], align 4
767 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
768 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
769 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
770 // CHECK1-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 8
771 // CHECK1-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 8
772 // CHECK1-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 8
773 // CHECK1-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 8
774 // CHECK1-NEXT: [[TMP2:%.*]] = bitcast [88 x i32]* [[Y1]] to i8*
775 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast [88 x i32]* [[TMP0]] to i8*
776 // CHECK1-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP2]], i8* align 4 [[TMP3]], i64 352, i1 false)
777 // CHECK1-NEXT: [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [99 x i32], [99 x i32]* [[Z2]], i32 0, i32 0
778 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i64 99
779 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP4]]
780 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
781 // CHECK1: omp.arrayinit.body:
782 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
783 // CHECK1-NEXT: store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
784 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
785 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
786 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
787 // CHECK1: omp.arrayinit.done:
788 // CHECK1-NEXT: [[LHS_BEGIN:%.*]] = bitcast [99 x i32]* [[TMP1]] to i32*
789 // CHECK1-NEXT: [[RHS_BEGIN:%.*]] = bitcast [99 x i32]* [[Z2]] to i32*
790 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
791 // CHECK1-NEXT: [[TMP6:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
792 // CHECK1-NEXT: store i8* [[TMP6]], i8** [[TMP5]], align 8
793 // CHECK1-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
794 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
795 // CHECK1-NEXT: [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
796 // CHECK1-NEXT: [[TMP10:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 1, i64 8, i8* [[TMP9]], void (i8*, i8*)* @.omp.reduction.reduction_func.17, [8 x i32]* @.gomp_critical_user_.reduction.var)
797 // CHECK1-NEXT: switch i32 [[TMP10]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
798 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
799 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
800 // CHECK1-NEXT: ]
801 // CHECK1: .omp.reduction.case1:
802 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i64 99
803 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP11]]
804 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
805 // CHECK1: omp.arraycpy.body:
806 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
807 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST3:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT4:%.*]], [[OMP_ARRAYCPY_BODY]] ]
808 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
809 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
810 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
811 // CHECK1-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
812 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT4]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], i32 1
813 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
814 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE5:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT4]], [[TMP11]]
815 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
816 // CHECK1: omp.arraycpy.done6:
817 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
818 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
819 // CHECK1: .omp.reduction.case2:
820 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i64 99
821 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY7:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP14]]
822 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY7]], label [[OMP_ARRAYCPY_DONE14:%.*]], label [[OMP_ARRAYCPY_BODY8:%.*]]
823 // CHECK1: omp.arraycpy.body8:
824 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST9:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT12:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
825 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST10:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT11:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
826 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], align 4
827 // CHECK1-NEXT: [[TMP16:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 [[TMP15]] monotonic, align 4
828 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT11]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 1
829 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT12]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], i32 1
830 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE13:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT11]], [[TMP14]]
831 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE13]], label [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_BODY8]]
832 // CHECK1: omp.arraycpy.done14:
833 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
834 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
835 // CHECK1: .omp.reduction.default:
836 // CHECK1-NEXT: ret void
837 //
838 //
839 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.17
840 // CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
841 // CHECK1-NEXT: entry:
842 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
843 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
844 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
845 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
846 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
847 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
848 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
849 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
850 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
851 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
852 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
853 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
854 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
855 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
856 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i64 99
857 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
858 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
859 // CHECK1: omp.arraycpy.body:
860 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
861 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
862 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
863 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
864 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
865 // CHECK1-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
866 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
867 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
868 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
869 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
870 // CHECK1: omp.arraycpy.done2:
871 // CHECK1-NEXT: ret void
872 //
873 //
874 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l65
875 // CHECK1-SAME: ([88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR1]] {
876 // CHECK1-NEXT: entry:
877 // CHECK1-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 8
878 // CHECK1-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 8
879 // CHECK1-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 8
880 // CHECK1-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 8
881 // CHECK1-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 8
882 // CHECK1-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 8
883 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [88 x i32]*, [99 x i32]*)* @.omp_outlined..20 to void (i32*, i32*, ...)*), [88 x i32]* [[TMP0]], [99 x i32]* [[TMP1]])
884 // CHECK1-NEXT: ret void
885 //
886 //
887 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..20
888 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR1]] {
889 // CHECK1-NEXT: entry:
890 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
891 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
892 // CHECK1-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 8
893 // CHECK1-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 8
894 // CHECK1-NEXT: [[Y1:%.*]] = alloca [88 x i32], align 4
895 // CHECK1-NEXT: [[X:%.*]] = alloca [77 x i32], align 4
896 // CHECK1-NEXT: [[Z2:%.*]] = alloca [99 x i32], align 4
897 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
898 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
899 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
900 // CHECK1-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 8
901 // CHECK1-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 8
902 // CHECK1-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 8
903 // CHECK1-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 8
904 // CHECK1-NEXT: [[TMP2:%.*]] = bitcast [88 x i32]* [[Y1]] to i8*
905 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast [88 x i32]* [[TMP0]] to i8*
906 // CHECK1-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP2]], i8* align 4 [[TMP3]], i64 352, i1 false)
907 // CHECK1-NEXT: [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [99 x i32], [99 x i32]* [[Z2]], i32 0, i32 0
908 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i64 99
909 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP4]]
910 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
911 // CHECK1: omp.arrayinit.body:
912 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
913 // CHECK1-NEXT: store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
914 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
915 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
916 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
917 // CHECK1: omp.arrayinit.done:
918 // CHECK1-NEXT: [[LHS_BEGIN:%.*]] = bitcast [99 x i32]* [[TMP1]] to i32*
919 // CHECK1-NEXT: [[RHS_BEGIN:%.*]] = bitcast [99 x i32]* [[Z2]] to i32*
920 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
921 // CHECK1-NEXT: [[TMP6:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
922 // CHECK1-NEXT: store i8* [[TMP6]], i8** [[TMP5]], align 8
923 // CHECK1-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
924 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
925 // CHECK1-NEXT: [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
926 // CHECK1-NEXT: [[TMP10:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 1, i64 8, i8* [[TMP9]], void (i8*, i8*)* @.omp.reduction.reduction_func.21, [8 x i32]* @.gomp_critical_user_.reduction.var)
927 // CHECK1-NEXT: switch i32 [[TMP10]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
928 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
929 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
930 // CHECK1-NEXT: ]
931 // CHECK1: .omp.reduction.case1:
932 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i64 99
933 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP11]]
934 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
935 // CHECK1: omp.arraycpy.body:
936 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
937 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST3:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT4:%.*]], [[OMP_ARRAYCPY_BODY]] ]
938 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
939 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
940 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
941 // CHECK1-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
942 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT4]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], i32 1
943 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
944 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE5:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT4]], [[TMP11]]
945 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
946 // CHECK1: omp.arraycpy.done6:
947 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
948 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
949 // CHECK1: .omp.reduction.case2:
950 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i64 99
951 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY7:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP14]]
952 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY7]], label [[OMP_ARRAYCPY_DONE14:%.*]], label [[OMP_ARRAYCPY_BODY8:%.*]]
953 // CHECK1: omp.arraycpy.body8:
954 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST9:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT12:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
955 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST10:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT11:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
956 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], align 4
957 // CHECK1-NEXT: [[TMP16:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 [[TMP15]] monotonic, align 4
958 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT11]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 1
959 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT12]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], i32 1
960 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE13:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT11]], [[TMP14]]
961 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE13]], label [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_BODY8]]
962 // CHECK1: omp.arraycpy.done14:
963 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
964 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
965 // CHECK1: .omp.reduction.default:
966 // CHECK1-NEXT: ret void
967 //
968 //
969 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.21
970 // CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
971 // CHECK1-NEXT: entry:
972 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
973 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
974 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
975 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
976 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
977 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
978 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
979 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
980 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
981 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
982 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
983 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
984 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
985 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
986 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i64 99
987 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
988 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
989 // CHECK1: omp.arraycpy.body:
990 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
991 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
992 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
993 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
994 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
995 // CHECK1-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
996 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
997 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
998 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
999 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
1000 // CHECK1: omp.arraycpy.done2:
1001 // CHECK1-NEXT: ret void
1002 //
1003 //
1004 // CHECK1-LABEL: define {{[^@]+}}@_Z9mapInt128v
1005 // CHECK1-SAME: () #[[ATTR0]] {
1006 // CHECK1-NEXT: entry:
1007 // CHECK1-NEXT: [[X:%.*]] = alloca i128, align 16
1008 // CHECK1-NEXT: [[Y:%.*]] = alloca i128, align 16
1009 // CHECK1-NEXT: [[Z:%.*]] = alloca i128, align 16
1010 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1011 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1012 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1013 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [3 x i8*], align 8
1014 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [3 x i8*], align 8
1015 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [3 x i8*], align 8
1016 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1017 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i128**
1018 // CHECK1-NEXT: store i128* [[Y]], i128** [[TMP1]], align 8
1019 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1020 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i128**
1021 // CHECK1-NEXT: store i128* [[Y]], i128** [[TMP3]], align 8
1022 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1023 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
1024 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1025 // CHECK1-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i128**
1026 // CHECK1-NEXT: store i128* [[Z]], i128** [[TMP6]], align 8
1027 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1028 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i128**
1029 // CHECK1-NEXT: store i128* [[Z]], i128** [[TMP8]], align 8
1030 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1031 // CHECK1-NEXT: store i8* null, i8** [[TMP9]], align 8
1032 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1033 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i128**
1034 // CHECK1-NEXT: store i128* [[X]], i128** [[TMP11]], align 8
1035 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1036 // CHECK1-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i128**
1037 // CHECK1-NEXT: store i128* [[X]], i128** [[TMP13]], align 8
1038 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1039 // CHECK1-NEXT: store i8* null, i8** [[TMP14]], align 8
1040 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1041 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1042 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1043 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1044 // CHECK1-NEXT: store i32 1, i32* [[TMP17]], align 4
1045 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1046 // CHECK1-NEXT: store i32 3, i32* [[TMP18]], align 4
1047 // CHECK1-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1048 // CHECK1-NEXT: store i8** [[TMP15]], i8*** [[TMP19]], align 8
1049 // CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1050 // CHECK1-NEXT: store i8** [[TMP16]], i8*** [[TMP20]], align 8
1051 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1052 // CHECK1-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.26, i32 0, i32 0), i64** [[TMP21]], align 8
1053 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1054 // CHECK1-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.27, i32 0, i32 0), i64** [[TMP22]], align 8
1055 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1056 // CHECK1-NEXT: store i8** null, i8*** [[TMP23]], align 8
1057 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1058 // CHECK1-NEXT: store i8** null, i8*** [[TMP24]], align 8
1059 // CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1060 // CHECK1-NEXT: store i64 0, i64* [[TMP25]], align 8
1061 // CHECK1-NEXT: [[TMP26:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9mapInt128v_l72.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1062 // CHECK1-NEXT: [[TMP27:%.*]] = icmp ne i32 [[TMP26]], 0
1063 // CHECK1-NEXT: br i1 [[TMP27]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1064 // CHECK1: omp_offload.failed:
1065 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9mapInt128v_l72(i128* [[Y]], i128* [[Z]]) #[[ATTR2]]
1066 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
1067 // CHECK1: omp_offload.cont:
1068 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
1069 // CHECK1-NEXT: [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i128**
1070 // CHECK1-NEXT: store i128* [[Y]], i128** [[TMP29]], align 8
1071 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
1072 // CHECK1-NEXT: [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i128**
1073 // CHECK1-NEXT: store i128* [[Y]], i128** [[TMP31]], align 8
1074 // CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
1075 // CHECK1-NEXT: store i8* null, i8** [[TMP32]], align 8
1076 // CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 1
1077 // CHECK1-NEXT: [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i128**
1078 // CHECK1-NEXT: store i128* [[Z]], i128** [[TMP34]], align 8
1079 // CHECK1-NEXT: [[TMP35:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 1
1080 // CHECK1-NEXT: [[TMP36:%.*]] = bitcast i8** [[TMP35]] to i128**
1081 // CHECK1-NEXT: store i128* [[Z]], i128** [[TMP36]], align 8
1082 // CHECK1-NEXT: [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 1
1083 // CHECK1-NEXT: store i8* null, i8** [[TMP37]], align 8
1084 // CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 2
1085 // CHECK1-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i128**
1086 // CHECK1-NEXT: store i128* [[X]], i128** [[TMP39]], align 8
1087 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 2
1088 // CHECK1-NEXT: [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i128**
1089 // CHECK1-NEXT: store i128* [[X]], i128** [[TMP41]], align 8
1090 // CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 2
1091 // CHECK1-NEXT: store i8* null, i8** [[TMP42]], align 8
1092 // CHECK1-NEXT: [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
1093 // CHECK1-NEXT: [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
1094 // CHECK1-NEXT: [[KERNEL_ARGS4:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1095 // CHECK1-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 0
1096 // CHECK1-NEXT: store i32 1, i32* [[TMP45]], align 4
1097 // CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 1
1098 // CHECK1-NEXT: store i32 3, i32* [[TMP46]], align 4
1099 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 2
1100 // CHECK1-NEXT: store i8** [[TMP43]], i8*** [[TMP47]], align 8
1101 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 3
1102 // CHECK1-NEXT: store i8** [[TMP44]], i8*** [[TMP48]], align 8
1103 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 4
1104 // CHECK1-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.30, i32 0, i32 0), i64** [[TMP49]], align 8
1105 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 5
1106 // CHECK1-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.31, i32 0, i32 0), i64** [[TMP50]], align 8
1107 // CHECK1-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 6
1108 // CHECK1-NEXT: store i8** null, i8*** [[TMP51]], align 8
1109 // CHECK1-NEXT: [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 7
1110 // CHECK1-NEXT: store i8** null, i8*** [[TMP52]], align 8
1111 // CHECK1-NEXT: [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 8
1112 // CHECK1-NEXT: store i64 0, i64* [[TMP53]], align 8
1113 // CHECK1-NEXT: [[TMP54:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9mapInt128v_l74.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]])
1114 // CHECK1-NEXT: [[TMP55:%.*]] = icmp ne i32 [[TMP54]], 0
1115 // CHECK1-NEXT: br i1 [[TMP55]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
1116 // CHECK1: omp_offload.failed5:
1117 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9mapInt128v_l74(i128* [[Y]], i128* [[Z]]) #[[ATTR2]]
1118 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT6]]
1119 // CHECK1: omp_offload.cont6:
1120 // CHECK1-NEXT: ret void
1121 //
1122 //
1123 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9mapInt128v_l72
1124 // CHECK1-SAME: (i128* noundef nonnull align 16 dereferenceable(16) [[Y:%.*]], i128* noundef nonnull align 16 dereferenceable(16) [[Z:%.*]]) #[[ATTR1]] {
1125 // CHECK1-NEXT: entry:
1126 // CHECK1-NEXT: [[Y_ADDR:%.*]] = alloca i128*, align 8
1127 // CHECK1-NEXT: [[Z_ADDR:%.*]] = alloca i128*, align 8
1128 // CHECK1-NEXT: store i128* [[Y]], i128** [[Y_ADDR]], align 8
1129 // CHECK1-NEXT: store i128* [[Z]], i128** [[Z_ADDR]], align 8
1130 // CHECK1-NEXT: [[TMP0:%.*]] = load i128*, i128** [[Y_ADDR]], align 8
1131 // CHECK1-NEXT: [[TMP1:%.*]] = load i128*, i128** [[Z_ADDR]], align 8
1132 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i128*, i128*)* @.omp_outlined..24 to void (i32*, i32*, ...)*), i128* [[TMP0]], i128* [[TMP1]])
1133 // CHECK1-NEXT: ret void
1134 //
1135 //
1136 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..24
1137 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i128* noundef nonnull align 16 dereferenceable(16) [[Y:%.*]], i128* noundef nonnull align 16 dereferenceable(16) [[Z:%.*]]) #[[ATTR1]] {
1138 // CHECK1-NEXT: entry:
1139 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1140 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1141 // CHECK1-NEXT: [[Y_ADDR:%.*]] = alloca i128*, align 8
1142 // CHECK1-NEXT: [[Z_ADDR:%.*]] = alloca i128*, align 8
1143 // CHECK1-NEXT: [[Y1:%.*]] = alloca i128, align 16
1144 // CHECK1-NEXT: [[X:%.*]] = alloca i128, align 16
1145 // CHECK1-NEXT: [[Z2:%.*]] = alloca i128, align 16
1146 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
1147 // CHECK1-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i128, align 16
1148 // CHECK1-NEXT: [[ATOMIC_TEMP3:%.*]] = alloca i128, align 16
1149 // CHECK1-NEXT: [[TMP:%.*]] = alloca i128, align 16
1150 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1151 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1152 // CHECK1-NEXT: store i128* [[Y]], i128** [[Y_ADDR]], align 8
1153 // CHECK1-NEXT: store i128* [[Z]], i128** [[Z_ADDR]], align 8
1154 // CHECK1-NEXT: [[TMP0:%.*]] = load i128*, i128** [[Y_ADDR]], align 8
1155 // CHECK1-NEXT: [[TMP1:%.*]] = load i128*, i128** [[Z_ADDR]], align 8
1156 // CHECK1-NEXT: [[TMP2:%.*]] = load i128, i128* [[TMP0]], align 16
1157 // CHECK1-NEXT: store i128 [[TMP2]], i128* [[Y1]], align 16
1158 // CHECK1-NEXT: store i128 0, i128* [[Z2]], align 16
1159 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1160 // CHECK1-NEXT: [[TMP4:%.*]] = bitcast i128* [[Z2]] to i8*
1161 // CHECK1-NEXT: store i8* [[TMP4]], i8** [[TMP3]], align 8
1162 // CHECK1-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1163 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
1164 // CHECK1-NEXT: [[TMP7:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
1165 // CHECK1-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], i32 1, i64 8, i8* [[TMP7]], void (i8*, i8*)* @.omp.reduction.reduction_func.25, [8 x i32]* @.gomp_critical_user_.reduction.var)
1166 // CHECK1-NEXT: switch i32 [[TMP8]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1167 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1168 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1169 // CHECK1-NEXT: ]
1170 // CHECK1: .omp.reduction.case1:
1171 // CHECK1-NEXT: [[TMP9:%.*]] = load i128, i128* [[TMP1]], align 16
1172 // CHECK1-NEXT: [[TMP10:%.*]] = load i128, i128* [[Z2]], align 16
1173 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i128 [[TMP9]], [[TMP10]]
1174 // CHECK1-NEXT: store i128 [[ADD]], i128* [[TMP1]], align 16
1175 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], [8 x i32]* @.gomp_critical_user_.reduction.var)
1176 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
1177 // CHECK1: .omp.reduction.case2:
1178 // CHECK1-NEXT: [[TMP11:%.*]] = load i128, i128* [[Z2]], align 16
1179 // CHECK1-NEXT: [[TMP12:%.*]] = bitcast i128* [[TMP1]] to i8*
1180 // CHECK1-NEXT: [[TMP13:%.*]] = bitcast i128* [[ATOMIC_TEMP]] to i8*
1181 // CHECK1-NEXT: call void @__atomic_load(i64 noundef 16, i8* noundef [[TMP12]], i8* noundef [[TMP13]], i32 noundef signext 0)
1182 // CHECK1-NEXT: br label [[ATOMIC_CONT:%.*]]
1183 // CHECK1: atomic_cont:
1184 // CHECK1-NEXT: [[TMP14:%.*]] = load i128, i128* [[ATOMIC_TEMP]], align 16
1185 // CHECK1-NEXT: store i128 [[TMP14]], i128* [[TMP]], align 16
1186 // CHECK1-NEXT: [[TMP15:%.*]] = load i128, i128* [[TMP]], align 16
1187 // CHECK1-NEXT: [[TMP16:%.*]] = load i128, i128* [[Z2]], align 16
1188 // CHECK1-NEXT: [[ADD4:%.*]] = add nsw i128 [[TMP15]], [[TMP16]]
1189 // CHECK1-NEXT: store i128 [[ADD4]], i128* [[ATOMIC_TEMP3]], align 16
1190 // CHECK1-NEXT: [[TMP17:%.*]] = bitcast i128* [[TMP1]] to i8*
1191 // CHECK1-NEXT: [[TMP18:%.*]] = bitcast i128* [[ATOMIC_TEMP]] to i8*
1192 // CHECK1-NEXT: [[TMP19:%.*]] = bitcast i128* [[ATOMIC_TEMP3]] to i8*
1193 // CHECK1-NEXT: [[CALL:%.*]] = call noundef zeroext i1 @__atomic_compare_exchange(i64 noundef 16, i8* noundef [[TMP17]], i8* noundef [[TMP18]], i8* noundef [[TMP19]], i32 noundef signext 0, i32 noundef signext 0)
1194 // CHECK1-NEXT: br i1 [[CALL]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
1195 // CHECK1: atomic_exit:
1196 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], [8 x i32]* @.gomp_critical_user_.reduction.var)
1197 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
1198 // CHECK1: .omp.reduction.default:
1199 // CHECK1-NEXT: ret void
1200 //
1201 //
1202 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.25
1203 // CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
1204 // CHECK1-NEXT: entry:
1205 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1206 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1207 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1208 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1209 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1210 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
1211 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1212 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
1213 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
1214 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1215 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i128*
1216 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
1217 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1218 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i128*
1219 // CHECK1-NEXT: [[TMP12:%.*]] = load i128, i128* [[TMP11]], align 16
1220 // CHECK1-NEXT: [[TMP13:%.*]] = load i128, i128* [[TMP8]], align 16
1221 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i128 [[TMP12]], [[TMP13]]
1222 // CHECK1-NEXT: store i128 [[ADD]], i128* [[TMP11]], align 16
1223 // CHECK1-NEXT: ret void
1224 //
1225 //
1226 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9mapInt128v_l74
1227 // CHECK1-SAME: (i128* noundef nonnull align 16 dereferenceable(16) [[Y:%.*]], i128* noundef nonnull align 16 dereferenceable(16) [[Z:%.*]]) #[[ATTR1]] {
1228 // CHECK1-NEXT: entry:
1229 // CHECK1-NEXT: [[Y_ADDR:%.*]] = alloca i128*, align 8
1230 // CHECK1-NEXT: [[Z_ADDR:%.*]] = alloca i128*, align 8
1231 // CHECK1-NEXT: store i128* [[Y]], i128** [[Y_ADDR]], align 8
1232 // CHECK1-NEXT: store i128* [[Z]], i128** [[Z_ADDR]], align 8
1233 // CHECK1-NEXT: [[TMP0:%.*]] = load i128*, i128** [[Y_ADDR]], align 8
1234 // CHECK1-NEXT: [[TMP1:%.*]] = load i128*, i128** [[Z_ADDR]], align 8
1235 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i128*, i128*)* @.omp_outlined..28 to void (i32*, i32*, ...)*), i128* [[TMP0]], i128* [[TMP1]])
1236 // CHECK1-NEXT: ret void
1237 //
1238 //
1239 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..28
1240 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i128* noundef nonnull align 16 dereferenceable(16) [[Y:%.*]], i128* noundef nonnull align 16 dereferenceable(16) [[Z:%.*]]) #[[ATTR1]] {
1241 // CHECK1-NEXT: entry:
1242 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1243 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1244 // CHECK1-NEXT: [[Y_ADDR:%.*]] = alloca i128*, align 8
1245 // CHECK1-NEXT: [[Z_ADDR:%.*]] = alloca i128*, align 8
1246 // CHECK1-NEXT: [[Y1:%.*]] = alloca i128, align 16
1247 // CHECK1-NEXT: [[X:%.*]] = alloca i128, align 16
1248 // CHECK1-NEXT: [[Z2:%.*]] = alloca i128, align 16
1249 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
1250 // CHECK1-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i128, align 16
1251 // CHECK1-NEXT: [[ATOMIC_TEMP3:%.*]] = alloca i128, align 16
1252 // CHECK1-NEXT: [[TMP:%.*]] = alloca i128, align 16
1253 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1254 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1255 // CHECK1-NEXT: store i128* [[Y]], i128** [[Y_ADDR]], align 8
1256 // CHECK1-NEXT: store i128* [[Z]], i128** [[Z_ADDR]], align 8
1257 // CHECK1-NEXT: [[TMP0:%.*]] = load i128*, i128** [[Y_ADDR]], align 8
1258 // CHECK1-NEXT: [[TMP1:%.*]] = load i128*, i128** [[Z_ADDR]], align 8
1259 // CHECK1-NEXT: [[TMP2:%.*]] = load i128, i128* [[TMP0]], align 16
1260 // CHECK1-NEXT: store i128 [[TMP2]], i128* [[Y1]], align 16
1261 // CHECK1-NEXT: store i128 0, i128* [[Z2]], align 16
1262 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1263 // CHECK1-NEXT: [[TMP4:%.*]] = bitcast i128* [[Z2]] to i8*
1264 // CHECK1-NEXT: store i8* [[TMP4]], i8** [[TMP3]], align 8
1265 // CHECK1-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1266 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
1267 // CHECK1-NEXT: [[TMP7:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
1268 // CHECK1-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], i32 1, i64 8, i8* [[TMP7]], void (i8*, i8*)* @.omp.reduction.reduction_func.29, [8 x i32]* @.gomp_critical_user_.reduction.var)
1269 // CHECK1-NEXT: switch i32 [[TMP8]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1270 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1271 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1272 // CHECK1-NEXT: ]
1273 // CHECK1: .omp.reduction.case1:
1274 // CHECK1-NEXT: [[TMP9:%.*]] = load i128, i128* [[TMP1]], align 16
1275 // CHECK1-NEXT: [[TMP10:%.*]] = load i128, i128* [[Z2]], align 16
1276 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i128 [[TMP9]], [[TMP10]]
1277 // CHECK1-NEXT: store i128 [[ADD]], i128* [[TMP1]], align 16
1278 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], [8 x i32]* @.gomp_critical_user_.reduction.var)
1279 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
1280 // CHECK1: .omp.reduction.case2:
1281 // CHECK1-NEXT: [[TMP11:%.*]] = load i128, i128* [[Z2]], align 16
1282 // CHECK1-NEXT: [[TMP12:%.*]] = bitcast i128* [[TMP1]] to i8*
1283 // CHECK1-NEXT: [[TMP13:%.*]] = bitcast i128* [[ATOMIC_TEMP]] to i8*
1284 // CHECK1-NEXT: call void @__atomic_load(i64 noundef 16, i8* noundef [[TMP12]], i8* noundef [[TMP13]], i32 noundef signext 0)
1285 // CHECK1-NEXT: br label [[ATOMIC_CONT:%.*]]
1286 // CHECK1: atomic_cont:
1287 // CHECK1-NEXT: [[TMP14:%.*]] = load i128, i128* [[ATOMIC_TEMP]], align 16
1288 // CHECK1-NEXT: store i128 [[TMP14]], i128* [[TMP]], align 16
1289 // CHECK1-NEXT: [[TMP15:%.*]] = load i128, i128* [[TMP]], align 16
1290 // CHECK1-NEXT: [[TMP16:%.*]] = load i128, i128* [[Z2]], align 16
1291 // CHECK1-NEXT: [[ADD4:%.*]] = add nsw i128 [[TMP15]], [[TMP16]]
1292 // CHECK1-NEXT: store i128 [[ADD4]], i128* [[ATOMIC_TEMP3]], align 16
1293 // CHECK1-NEXT: [[TMP17:%.*]] = bitcast i128* [[TMP1]] to i8*
1294 // CHECK1-NEXT: [[TMP18:%.*]] = bitcast i128* [[ATOMIC_TEMP]] to i8*
1295 // CHECK1-NEXT: [[TMP19:%.*]] = bitcast i128* [[ATOMIC_TEMP3]] to i8*
1296 // CHECK1-NEXT: [[CALL:%.*]] = call noundef zeroext i1 @__atomic_compare_exchange(i64 noundef 16, i8* noundef [[TMP17]], i8* noundef [[TMP18]], i8* noundef [[TMP19]], i32 noundef signext 0, i32 noundef signext 0)
1297 // CHECK1-NEXT: br i1 [[CALL]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
1298 // CHECK1: atomic_exit:
1299 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], [8 x i32]* @.gomp_critical_user_.reduction.var)
1300 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
1301 // CHECK1: .omp.reduction.default:
1302 // CHECK1-NEXT: ret void
1303 //
1304 //
1305 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.29
1306 // CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
1307 // CHECK1-NEXT: entry:
1308 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1309 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1310 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1311 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1312 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1313 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
1314 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1315 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
1316 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
1317 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1318 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i128*
1319 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
1320 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1321 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i128*
1322 // CHECK1-NEXT: [[TMP12:%.*]] = load i128, i128* [[TMP11]], align 16
1323 // CHECK1-NEXT: [[TMP13:%.*]] = load i128, i128* [[TMP8]], align 16
1324 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i128 [[TMP12]], [[TMP13]]
1325 // CHECK1-NEXT: store i128 [[ADD]], i128* [[TMP11]], align 16
1326 // CHECK1-NEXT: ret void
1327 //
1328 //
1329 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1330 // CHECK1-SAME: () #[[ATTR7:[0-9]+]] {
1331 // CHECK1-NEXT: entry:
1332 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
1333 // CHECK1-NEXT: ret void
1334 //
1335 //
1336 // CHECK3-LABEL: define {{[^@]+}}@_Z14mapWithPrivatev
1337 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
1338 // CHECK3-NEXT: entry:
1339 // CHECK3-NEXT: [[X:%.*]] = alloca i32, align 4
1340 // CHECK3-NEXT: [[Y:%.*]] = alloca i32, align 4
1341 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 4
1342 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 4
1343 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 4
1344 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1345 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32**
1346 // CHECK3-NEXT: store i32* [[X]], i32** [[TMP1]], align 4
1347 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1348 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
1349 // CHECK3-NEXT: store i32* [[X]], i32** [[TMP3]], align 4
1350 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1351 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
1352 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1353 // CHECK3-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32**
1354 // CHECK3-NEXT: store i32* [[Y]], i32** [[TMP6]], align 4
1355 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1356 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32**
1357 // CHECK3-NEXT: store i32* [[Y]], i32** [[TMP8]], align 4
1358 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1359 // CHECK3-NEXT: store i8* null, i8** [[TMP9]], align 4
1360 // CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1361 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1362 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1363 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1364 // CHECK3-NEXT: store i32 1, i32* [[TMP12]], align 4
1365 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1366 // CHECK3-NEXT: store i32 2, i32* [[TMP13]], align 4
1367 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1368 // CHECK3-NEXT: store i8** [[TMP10]], i8*** [[TMP14]], align 4
1369 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1370 // CHECK3-NEXT: store i8** [[TMP11]], i8*** [[TMP15]], align 4
1371 // CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1372 // CHECK3-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP16]], align 4
1373 // CHECK3-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1374 // CHECK3-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP17]], align 4
1375 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1376 // CHECK3-NEXT: store i8** null, i8*** [[TMP18]], align 4
1377 // CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1378 // CHECK3-NEXT: store i8** null, i8*** [[TMP19]], align 4
1379 // CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1380 // CHECK3-NEXT: store i64 0, i64* [[TMP20]], align 8
1381 // CHECK3-NEXT: [[TMP21:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z14mapWithPrivatev_l27.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1382 // CHECK3-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
1383 // CHECK3-NEXT: br i1 [[TMP22]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1384 // CHECK3: omp_offload.failed:
1385 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z14mapWithPrivatev_l27() #[[ATTR2:[0-9]+]]
1386 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
1387 // CHECK3: omp_offload.cont:
1388 // CHECK3-NEXT: ret void
1389 //
1390 //
1391 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z14mapWithPrivatev_l27
1392 // CHECK3-SAME: () #[[ATTR1:[0-9]+]] {
1393 // CHECK3-NEXT: entry:
1394 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
1395 // CHECK3-NEXT: ret void
1396 //
1397 //
1398 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
1399 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1400 // CHECK3-NEXT: entry:
1401 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1402 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1403 // CHECK3-NEXT: [[X:%.*]] = alloca i32, align 4
1404 // CHECK3-NEXT: [[Y:%.*]] = alloca i32, align 4
1405 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1406 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1407 // CHECK3-NEXT: ret void
1408 //
1409 //
1410 // CHECK3-LABEL: define {{[^@]+}}@_Z19mapWithFirstprivatev
1411 // CHECK3-SAME: () #[[ATTR0]] {
1412 // CHECK3-NEXT: entry:
1413 // CHECK3-NEXT: [[X:%.*]] = alloca i32, align 4
1414 // CHECK3-NEXT: [[Y:%.*]] = alloca i32, align 4
1415 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 4
1416 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 4
1417 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 4
1418 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1419 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32**
1420 // CHECK3-NEXT: store i32* [[X]], i32** [[TMP1]], align 4
1421 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1422 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
1423 // CHECK3-NEXT: store i32* [[X]], i32** [[TMP3]], align 4
1424 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1425 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
1426 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1427 // CHECK3-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32**
1428 // CHECK3-NEXT: store i32* [[Y]], i32** [[TMP6]], align 4
1429 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1430 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32**
1431 // CHECK3-NEXT: store i32* [[Y]], i32** [[TMP8]], align 4
1432 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1433 // CHECK3-NEXT: store i8* null, i8** [[TMP9]], align 4
1434 // CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1435 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1436 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1437 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1438 // CHECK3-NEXT: store i32 1, i32* [[TMP12]], align 4
1439 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1440 // CHECK3-NEXT: store i32 2, i32* [[TMP13]], align 4
1441 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1442 // CHECK3-NEXT: store i8** [[TMP10]], i8*** [[TMP14]], align 4
1443 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1444 // CHECK3-NEXT: store i8** [[TMP11]], i8*** [[TMP15]], align 4
1445 // CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1446 // CHECK3-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP16]], align 4
1447 // CHECK3-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1448 // CHECK3-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP17]], align 4
1449 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1450 // CHECK3-NEXT: store i8** null, i8*** [[TMP18]], align 4
1451 // CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1452 // CHECK3-NEXT: store i8** null, i8*** [[TMP19]], align 4
1453 // CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1454 // CHECK3-NEXT: store i64 0, i64* [[TMP20]], align 8
1455 // CHECK3-NEXT: [[TMP21:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z19mapWithFirstprivatev_l33.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1456 // CHECK3-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
1457 // CHECK3-NEXT: br i1 [[TMP22]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1458 // CHECK3: omp_offload.failed:
1459 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z19mapWithFirstprivatev_l33(i32* [[X]], i32* [[Y]]) #[[ATTR2]]
1460 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
1461 // CHECK3: omp_offload.cont:
1462 // CHECK3-NEXT: ret void
1463 //
1464 //
1465 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z19mapWithFirstprivatev_l33
1466 // CHECK3-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[Y:%.*]]) #[[ATTR1]] {
1467 // CHECK3-NEXT: entry:
1468 // CHECK3-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 4
1469 // CHECK3-NEXT: [[Y_ADDR:%.*]] = alloca i32*, align 4
1470 // CHECK3-NEXT: [[X_CASTED:%.*]] = alloca i32, align 4
1471 // CHECK3-NEXT: [[Y_CASTED:%.*]] = alloca i32, align 4
1472 // CHECK3-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 4
1473 // CHECK3-NEXT: store i32* [[Y]], i32** [[Y_ADDR]], align 4
1474 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 4
1475 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[Y_ADDR]], align 4
1476 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
1477 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[X_CASTED]], align 4
1478 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[X_CASTED]], align 4
1479 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP1]], align 4
1480 // CHECK3-NEXT: store i32 [[TMP4]], i32* [[Y_CASTED]], align 4
1481 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[Y_CASTED]], align 4
1482 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP3]], i32 [[TMP5]])
1483 // CHECK3-NEXT: ret void
1484 //
1485 //
1486 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
1487 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[X:%.*]], i32 noundef [[Y:%.*]]) #[[ATTR1]] {
1488 // CHECK3-NEXT: entry:
1489 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1490 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1491 // CHECK3-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4
1492 // CHECK3-NEXT: [[Y_ADDR:%.*]] = alloca i32, align 4
1493 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1494 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1495 // CHECK3-NEXT: store i32 [[X]], i32* [[X_ADDR]], align 4
1496 // CHECK3-NEXT: store i32 [[Y]], i32* [[Y_ADDR]], align 4
1497 // CHECK3-NEXT: ret void
1498 //
1499 //
1500 // CHECK3-LABEL: define {{[^@]+}}@_Z16mapWithReductionv
1501 // CHECK3-SAME: () #[[ATTR0]] {
1502 // CHECK3-NEXT: entry:
1503 // CHECK3-NEXT: [[X:%.*]] = alloca i32, align 4
1504 // CHECK3-NEXT: [[Y:%.*]] = alloca i32, align 4
1505 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 4
1506 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 4
1507 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 4
1508 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1509 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32**
1510 // CHECK3-NEXT: store i32* [[X]], i32** [[TMP1]], align 4
1511 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1512 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
1513 // CHECK3-NEXT: store i32* [[X]], i32** [[TMP3]], align 4
1514 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1515 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
1516 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1517 // CHECK3-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32**
1518 // CHECK3-NEXT: store i32* [[Y]], i32** [[TMP6]], align 4
1519 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1520 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32**
1521 // CHECK3-NEXT: store i32* [[Y]], i32** [[TMP8]], align 4
1522 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1523 // CHECK3-NEXT: store i8* null, i8** [[TMP9]], align 4
1524 // CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1525 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1526 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1527 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1528 // CHECK3-NEXT: store i32 1, i32* [[TMP12]], align 4
1529 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1530 // CHECK3-NEXT: store i32 2, i32* [[TMP13]], align 4
1531 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1532 // CHECK3-NEXT: store i8** [[TMP10]], i8*** [[TMP14]], align 4
1533 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1534 // CHECK3-NEXT: store i8** [[TMP11]], i8*** [[TMP15]], align 4
1535 // CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1536 // CHECK3-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64** [[TMP16]], align 4
1537 // CHECK3-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1538 // CHECK3-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i64** [[TMP17]], align 4
1539 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1540 // CHECK3-NEXT: store i8** null, i8*** [[TMP18]], align 4
1541 // CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1542 // CHECK3-NEXT: store i8** null, i8*** [[TMP19]], align 4
1543 // CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1544 // CHECK3-NEXT: store i64 0, i64* [[TMP20]], align 8
1545 // CHECK3-NEXT: [[TMP21:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16mapWithReductionv_l39.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1546 // CHECK3-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
1547 // CHECK3-NEXT: br i1 [[TMP22]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1548 // CHECK3: omp_offload.failed:
1549 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16mapWithReductionv_l39(i32* [[X]], i32* [[Y]]) #[[ATTR2]]
1550 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
1551 // CHECK3: omp_offload.cont:
1552 // CHECK3-NEXT: ret void
1553 //
1554 //
1555 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16mapWithReductionv_l39
1556 // CHECK3-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[Y:%.*]]) #[[ATTR1]] {
1557 // CHECK3-NEXT: entry:
1558 // CHECK3-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 4
1559 // CHECK3-NEXT: [[Y_ADDR:%.*]] = alloca i32*, align 4
1560 // CHECK3-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 4
1561 // CHECK3-NEXT: store i32* [[Y]], i32** [[Y_ADDR]], align 4
1562 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 4
1563 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[Y_ADDR]], align 4
1564 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32* [[TMP0]], i32* [[TMP1]])
1565 // CHECK3-NEXT: ret void
1566 //
1567 //
1568 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
1569 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[Y:%.*]]) #[[ATTR1]] {
1570 // CHECK3-NEXT: entry:
1571 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1572 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1573 // CHECK3-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 4
1574 // CHECK3-NEXT: [[Y_ADDR:%.*]] = alloca i32*, align 4
1575 // CHECK3-NEXT: [[X1:%.*]] = alloca i32, align 4
1576 // CHECK3-NEXT: [[Y2:%.*]] = alloca i32, align 4
1577 // CHECK3-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 4
1578 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1579 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1580 // CHECK3-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 4
1581 // CHECK3-NEXT: store i32* [[Y]], i32** [[Y_ADDR]], align 4
1582 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 4
1583 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[Y_ADDR]], align 4
1584 // CHECK3-NEXT: store i32 0, i32* [[X1]], align 4
1585 // CHECK3-NEXT: store i32 0, i32* [[Y2]], align 4
1586 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
1587 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i32* [[X1]] to i8*
1588 // CHECK3-NEXT: store i8* [[TMP3]], i8** [[TMP2]], align 4
1589 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
1590 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i32* [[Y2]] to i8*
1591 // CHECK3-NEXT: store i8* [[TMP5]], i8** [[TMP4]], align 4
1592 // CHECK3-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1593 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
1594 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
1595 // CHECK3-NEXT: [[TMP9:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP7]], i32 2, i32 8, i8* [[TMP8]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
1596 // CHECK3-NEXT: switch i32 [[TMP9]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1597 // CHECK3-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1598 // CHECK3-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1599 // CHECK3-NEXT: ]
1600 // CHECK3: .omp.reduction.case1:
1601 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP0]], align 4
1602 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[X1]], align 4
1603 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
1604 // CHECK3-NEXT: store i32 [[ADD]], i32* [[TMP0]], align 4
1605 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP1]], align 4
1606 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[Y2]], align 4
1607 // CHECK3-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1608 // CHECK3-NEXT: store i32 [[ADD3]], i32* [[TMP1]], align 4
1609 // CHECK3-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP7]], [8 x i32]* @.gomp_critical_user_.reduction.var)
1610 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
1611 // CHECK3: .omp.reduction.case2:
1612 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[X1]], align 4
1613 // CHECK3-NEXT: [[TMP15:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP14]] monotonic, align 4
1614 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[Y2]], align 4
1615 // CHECK3-NEXT: [[TMP17:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP16]] monotonic, align 4
1616 // CHECK3-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP7]], [8 x i32]* @.gomp_critical_user_.reduction.var)
1617 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
1618 // CHECK3: .omp.reduction.default:
1619 // CHECK3-NEXT: ret void
1620 //
1621 //
1622 // CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
1623 // CHECK3-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
1624 // CHECK3-NEXT: entry:
1625 // CHECK3-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4
1626 // CHECK3-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4
1627 // CHECK3-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4
1628 // CHECK3-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4
1629 // CHECK3-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4
1630 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [2 x i8*]*
1631 // CHECK3-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4
1632 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
1633 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
1634 // CHECK3-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
1635 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1636 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i32 0, i32 0
1637 // CHECK3-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
1638 // CHECK3-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1639 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
1640 // CHECK3-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 4
1641 // CHECK3-NEXT: [[TMP14:%.*]] = bitcast i8* [[TMP13]] to i32*
1642 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i32 0, i32 1
1643 // CHECK3-NEXT: [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 4
1644 // CHECK3-NEXT: [[TMP17:%.*]] = bitcast i8* [[TMP16]] to i32*
1645 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP11]], align 4
1646 // CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP8]], align 4
1647 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
1648 // CHECK3-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
1649 // CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP17]], align 4
1650 // CHECK3-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP14]], align 4
1651 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
1652 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[TMP17]], align 4
1653 // CHECK3-NEXT: ret void
1654 //
1655 //
1656 // CHECK3-LABEL: define {{[^@]+}}@_Z7mapFromv
1657 // CHECK3-SAME: () #[[ATTR0]] {
1658 // CHECK3-NEXT: entry:
1659 // CHECK3-NEXT: [[X:%.*]] = alloca i32, align 4
1660 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
1661 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
1662 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
1663 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1664 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32**
1665 // CHECK3-NEXT: store i32* [[X]], i32** [[TMP1]], align 4
1666 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1667 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
1668 // CHECK3-NEXT: store i32* [[X]], i32** [[TMP3]], align 4
1669 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1670 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
1671 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1672 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1673 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1674 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1675 // CHECK3-NEXT: store i32 1, i32* [[TMP7]], align 4
1676 // CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1677 // CHECK3-NEXT: store i32 1, i32* [[TMP8]], align 4
1678 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1679 // CHECK3-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4
1680 // CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1681 // CHECK3-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4
1682 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1683 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP11]], align 4
1684 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1685 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP12]], align 4
1686 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1687 // CHECK3-NEXT: store i8** null, i8*** [[TMP13]], align 4
1688 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1689 // CHECK3-NEXT: store i8** null, i8*** [[TMP14]], align 4
1690 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1691 // CHECK3-NEXT: store i64 0, i64* [[TMP15]], align 8
1692 // CHECK3-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z7mapFromv_l45.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1693 // CHECK3-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
1694 // CHECK3-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1695 // CHECK3: omp_offload.failed:
1696 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z7mapFromv_l45(i32* [[X]]) #[[ATTR2]]
1697 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
1698 // CHECK3: omp_offload.cont:
1699 // CHECK3-NEXT: ret void
1700 //
1701 //
1702 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z7mapFromv_l45
1703 // CHECK3-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
1704 // CHECK3-NEXT: entry:
1705 // CHECK3-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 4
1706 // CHECK3-NEXT: [[X_CASTED:%.*]] = alloca i32, align 4
1707 // CHECK3-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 4
1708 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 4
1709 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1710 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[X_CASTED]], align 4
1711 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[X_CASTED]], align 4
1712 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP2]])
1713 // CHECK3-NEXT: ret void
1714 //
1715 //
1716 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
1717 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[X:%.*]]) #[[ATTR1]] {
1718 // CHECK3-NEXT: entry:
1719 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1720 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1721 // CHECK3-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4
1722 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1723 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1724 // CHECK3-NEXT: store i32 [[X]], i32* [[X_ADDR]], align 4
1725 // CHECK3-NEXT: ret void
1726 //
1727 //
1728 // CHECK3-LABEL: define {{[^@]+}}@_Z5mapTov
1729 // CHECK3-SAME: () #[[ATTR0]] {
1730 // CHECK3-NEXT: entry:
1731 // CHECK3-NEXT: [[X:%.*]] = alloca i32, align 4
1732 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
1733 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
1734 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
1735 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1736 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32**
1737 // CHECK3-NEXT: store i32* [[X]], i32** [[TMP1]], align 4
1738 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1739 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
1740 // CHECK3-NEXT: store i32* [[X]], i32** [[TMP3]], align 4
1741 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1742 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
1743 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1744 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1745 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1746 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1747 // CHECK3-NEXT: store i32 1, i32* [[TMP7]], align 4
1748 // CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1749 // CHECK3-NEXT: store i32 1, i32* [[TMP8]], align 4
1750 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1751 // CHECK3-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4
1752 // CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1753 // CHECK3-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4
1754 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1755 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.11, i32 0, i32 0), i64** [[TMP11]], align 4
1756 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1757 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.12, i32 0, i32 0), i64** [[TMP12]], align 4
1758 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1759 // CHECK3-NEXT: store i8** null, i8*** [[TMP13]], align 4
1760 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1761 // CHECK3-NEXT: store i8** null, i8*** [[TMP14]], align 4
1762 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1763 // CHECK3-NEXT: store i64 0, i64* [[TMP15]], align 8
1764 // CHECK3-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5mapTov_l51.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1765 // CHECK3-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
1766 // CHECK3-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1767 // CHECK3: omp_offload.failed:
1768 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5mapTov_l51(i32* [[X]]) #[[ATTR2]]
1769 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
1770 // CHECK3: omp_offload.cont:
1771 // CHECK3-NEXT: ret void
1772 //
1773 //
1774 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5mapTov_l51
1775 // CHECK3-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
1776 // CHECK3-NEXT: entry:
1777 // CHECK3-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 4
1778 // CHECK3-NEXT: [[X_CASTED:%.*]] = alloca i32, align 4
1779 // CHECK3-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 4
1780 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 4
1781 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1782 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[X_CASTED]], align 4
1783 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[X_CASTED]], align 4
1784 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..10 to void (i32*, i32*, ...)*), i32 [[TMP2]])
1785 // CHECK3-NEXT: ret void
1786 //
1787 //
1788 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..10
1789 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[X:%.*]]) #[[ATTR1]] {
1790 // CHECK3-NEXT: entry:
1791 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1792 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1793 // CHECK3-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4
1794 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1795 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1796 // CHECK3-NEXT: store i32 [[X]], i32* [[X_ADDR]], align 4
1797 // CHECK3-NEXT: ret void
1798 //
1799 //
1800 // CHECK3-LABEL: define {{[^@]+}}@_Z8mapAllocv
1801 // CHECK3-SAME: () #[[ATTR0]] {
1802 // CHECK3-NEXT: entry:
1803 // CHECK3-NEXT: [[X:%.*]] = alloca i32, align 4
1804 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
1805 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
1806 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
1807 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1808 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32**
1809 // CHECK3-NEXT: store i32* [[X]], i32** [[TMP1]], align 4
1810 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1811 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32**
1812 // CHECK3-NEXT: store i32* [[X]], i32** [[TMP3]], align 4
1813 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1814 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
1815 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1816 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1817 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1818 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1819 // CHECK3-NEXT: store i32 1, i32* [[TMP7]], align 4
1820 // CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1821 // CHECK3-NEXT: store i32 1, i32* [[TMP8]], align 4
1822 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1823 // CHECK3-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4
1824 // CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1825 // CHECK3-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4
1826 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1827 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.14, i32 0, i32 0), i64** [[TMP11]], align 4
1828 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1829 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.15, i32 0, i32 0), i64** [[TMP12]], align 4
1830 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1831 // CHECK3-NEXT: store i8** null, i8*** [[TMP13]], align 4
1832 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1833 // CHECK3-NEXT: store i8** null, i8*** [[TMP14]], align 4
1834 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1835 // CHECK3-NEXT: store i64 0, i64* [[TMP15]], align 8
1836 // CHECK3-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapAllocv_l57.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1837 // CHECK3-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
1838 // CHECK3-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1839 // CHECK3: omp_offload.failed:
1840 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapAllocv_l57(i32* [[X]]) #[[ATTR2]]
1841 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
1842 // CHECK3: omp_offload.cont:
1843 // CHECK3-NEXT: ret void
1844 //
1845 //
1846 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapAllocv_l57
1847 // CHECK3-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
1848 // CHECK3-NEXT: entry:
1849 // CHECK3-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 4
1850 // CHECK3-NEXT: [[X_CASTED:%.*]] = alloca i32, align 4
1851 // CHECK3-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 4
1852 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 4
1853 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1854 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[X_CASTED]], align 4
1855 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[X_CASTED]], align 4
1856 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i32 [[TMP2]])
1857 // CHECK3-NEXT: ret void
1858 //
1859 //
1860 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..13
1861 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[X:%.*]]) #[[ATTR1]] {
1862 // CHECK3-NEXT: entry:
1863 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1864 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1865 // CHECK3-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4
1866 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1867 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1868 // CHECK3-NEXT: store i32 [[X]], i32* [[X_ADDR]], align 4
1869 // CHECK3-NEXT: ret void
1870 //
1871 //
1872 // CHECK3-LABEL: define {{[^@]+}}@_Z8mapArrayv
1873 // CHECK3-SAME: () #[[ATTR0]] {
1874 // CHECK3-NEXT: entry:
1875 // CHECK3-NEXT: [[X:%.*]] = alloca [77 x i32], align 4
1876 // CHECK3-NEXT: [[Y:%.*]] = alloca [88 x i32], align 4
1877 // CHECK3-NEXT: [[Z:%.*]] = alloca [99 x i32], align 4
1878 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
1879 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
1880 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
1881 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [3 x i8*], align 4
1882 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [3 x i8*], align 4
1883 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [3 x i8*], align 4
1884 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1885 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [88 x i32]**
1886 // CHECK3-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[TMP1]], align 4
1887 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1888 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [88 x i32]**
1889 // CHECK3-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[TMP3]], align 4
1890 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1891 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
1892 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1893 // CHECK3-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to [99 x i32]**
1894 // CHECK3-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[TMP6]], align 4
1895 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1896 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [99 x i32]**
1897 // CHECK3-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[TMP8]], align 4
1898 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1899 // CHECK3-NEXT: store i8* null, i8** [[TMP9]], align 4
1900 // CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1901 // CHECK3-NEXT: [[TMP11:%.*]] = bitcast i8** [[TMP10]] to [77 x i32]**
1902 // CHECK3-NEXT: store [77 x i32]* [[X]], [77 x i32]** [[TMP11]], align 4
1903 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1904 // CHECK3-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to [77 x i32]**
1905 // CHECK3-NEXT: store [77 x i32]* [[X]], [77 x i32]** [[TMP13]], align 4
1906 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1907 // CHECK3-NEXT: store i8* null, i8** [[TMP14]], align 4
1908 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1909 // CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1910 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1911 // CHECK3-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1912 // CHECK3-NEXT: store i32 1, i32* [[TMP17]], align 4
1913 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1914 // CHECK3-NEXT: store i32 3, i32* [[TMP18]], align 4
1915 // CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1916 // CHECK3-NEXT: store i8** [[TMP15]], i8*** [[TMP19]], align 4
1917 // CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1918 // CHECK3-NEXT: store i8** [[TMP16]], i8*** [[TMP20]], align 4
1919 // CHECK3-NEXT: [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1920 // CHECK3-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.18, i32 0, i32 0), i64** [[TMP21]], align 4
1921 // CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1922 // CHECK3-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.19, i32 0, i32 0), i64** [[TMP22]], align 4
1923 // CHECK3-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1924 // CHECK3-NEXT: store i8** null, i8*** [[TMP23]], align 4
1925 // CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1926 // CHECK3-NEXT: store i8** null, i8*** [[TMP24]], align 4
1927 // CHECK3-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1928 // CHECK3-NEXT: store i64 0, i64* [[TMP25]], align 8
1929 // CHECK3-NEXT: [[TMP26:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l63.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1930 // CHECK3-NEXT: [[TMP27:%.*]] = icmp ne i32 [[TMP26]], 0
1931 // CHECK3-NEXT: br i1 [[TMP27]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1932 // CHECK3: omp_offload.failed:
1933 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l63([88 x i32]* [[Y]], [99 x i32]* [[Z]]) #[[ATTR2]]
1934 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
1935 // CHECK3: omp_offload.cont:
1936 // CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
1937 // CHECK3-NEXT: [[TMP29:%.*]] = bitcast i8** [[TMP28]] to [88 x i32]**
1938 // CHECK3-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[TMP29]], align 4
1939 // CHECK3-NEXT: [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
1940 // CHECK3-NEXT: [[TMP31:%.*]] = bitcast i8** [[TMP30]] to [88 x i32]**
1941 // CHECK3-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[TMP31]], align 4
1942 // CHECK3-NEXT: [[TMP32:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
1943 // CHECK3-NEXT: store i8* null, i8** [[TMP32]], align 4
1944 // CHECK3-NEXT: [[TMP33:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 1
1945 // CHECK3-NEXT: [[TMP34:%.*]] = bitcast i8** [[TMP33]] to [99 x i32]**
1946 // CHECK3-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[TMP34]], align 4
1947 // CHECK3-NEXT: [[TMP35:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 1
1948 // CHECK3-NEXT: [[TMP36:%.*]] = bitcast i8** [[TMP35]] to [99 x i32]**
1949 // CHECK3-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[TMP36]], align 4
1950 // CHECK3-NEXT: [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 1
1951 // CHECK3-NEXT: store i8* null, i8** [[TMP37]], align 4
1952 // CHECK3-NEXT: [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 2
1953 // CHECK3-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [77 x i32]**
1954 // CHECK3-NEXT: store [77 x i32]* [[X]], [77 x i32]** [[TMP39]], align 4
1955 // CHECK3-NEXT: [[TMP40:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 2
1956 // CHECK3-NEXT: [[TMP41:%.*]] = bitcast i8** [[TMP40]] to [77 x i32]**
1957 // CHECK3-NEXT: store [77 x i32]* [[X]], [77 x i32]** [[TMP41]], align 4
1958 // CHECK3-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 2
1959 // CHECK3-NEXT: store i8* null, i8** [[TMP42]], align 4
1960 // CHECK3-NEXT: [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
1961 // CHECK3-NEXT: [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
1962 // CHECK3-NEXT: [[KERNEL_ARGS4:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1963 // CHECK3-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 0
1964 // CHECK3-NEXT: store i32 1, i32* [[TMP45]], align 4
1965 // CHECK3-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 1
1966 // CHECK3-NEXT: store i32 3, i32* [[TMP46]], align 4
1967 // CHECK3-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 2
1968 // CHECK3-NEXT: store i8** [[TMP43]], i8*** [[TMP47]], align 4
1969 // CHECK3-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 3
1970 // CHECK3-NEXT: store i8** [[TMP44]], i8*** [[TMP48]], align 4
1971 // CHECK3-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 4
1972 // CHECK3-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.22, i32 0, i32 0), i64** [[TMP49]], align 4
1973 // CHECK3-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 5
1974 // CHECK3-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.23, i32 0, i32 0), i64** [[TMP50]], align 4
1975 // CHECK3-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 6
1976 // CHECK3-NEXT: store i8** null, i8*** [[TMP51]], align 4
1977 // CHECK3-NEXT: [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 7
1978 // CHECK3-NEXT: store i8** null, i8*** [[TMP52]], align 4
1979 // CHECK3-NEXT: [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]], i32 0, i32 8
1980 // CHECK3-NEXT: store i64 0, i64* [[TMP53]], align 8
1981 // CHECK3-NEXT: [[TMP54:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l65.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS4]])
1982 // CHECK3-NEXT: [[TMP55:%.*]] = icmp ne i32 [[TMP54]], 0
1983 // CHECK3-NEXT: br i1 [[TMP55]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
1984 // CHECK3: omp_offload.failed5:
1985 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l65([88 x i32]* [[Y]], [99 x i32]* [[Z]]) #[[ATTR2]]
1986 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT6]]
1987 // CHECK3: omp_offload.cont6:
1988 // CHECK3-NEXT: ret void
1989 //
1990 //
1991 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l63
1992 // CHECK3-SAME: ([88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR1]] {
1993 // CHECK3-NEXT: entry:
1994 // CHECK3-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 4
1995 // CHECK3-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 4
1996 // CHECK3-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 4
1997 // CHECK3-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 4
1998 // CHECK3-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 4
1999 // CHECK3-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 4
2000 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [88 x i32]*, [99 x i32]*)* @.omp_outlined..16 to void (i32*, i32*, ...)*), [88 x i32]* [[TMP0]], [99 x i32]* [[TMP1]])
2001 // CHECK3-NEXT: ret void
2002 //
2003 //
2004 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..16
2005 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR1]] {
2006 // CHECK3-NEXT: entry:
2007 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2008 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2009 // CHECK3-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 4
2010 // CHECK3-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 4
2011 // CHECK3-NEXT: [[Y1:%.*]] = alloca [88 x i32], align 4
2012 // CHECK3-NEXT: [[X:%.*]] = alloca [77 x i32], align 4
2013 // CHECK3-NEXT: [[Z2:%.*]] = alloca [99 x i32], align 4
2014 // CHECK3-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
2015 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2016 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2017 // CHECK3-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 4
2018 // CHECK3-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 4
2019 // CHECK3-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 4
2020 // CHECK3-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 4
2021 // CHECK3-NEXT: [[TMP2:%.*]] = bitcast [88 x i32]* [[Y1]] to i8*
2022 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast [88 x i32]* [[TMP0]] to i8*
2023 // CHECK3-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP2]], i8* align 4 [[TMP3]], i32 352, i1 false)
2024 // CHECK3-NEXT: [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [99 x i32], [99 x i32]* [[Z2]], i32 0, i32 0
2025 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i32 99
2026 // CHECK3-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP4]]
2027 // CHECK3-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
2028 // CHECK3: omp.arrayinit.body:
2029 // CHECK3-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
2030 // CHECK3-NEXT: store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
2031 // CHECK3-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
2032 // CHECK3-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
2033 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
2034 // CHECK3: omp.arrayinit.done:
2035 // CHECK3-NEXT: [[LHS_BEGIN:%.*]] = bitcast [99 x i32]* [[TMP1]] to i32*
2036 // CHECK3-NEXT: [[RHS_BEGIN:%.*]] = bitcast [99 x i32]* [[Z2]] to i32*
2037 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
2038 // CHECK3-NEXT: [[TMP6:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
2039 // CHECK3-NEXT: store i8* [[TMP6]], i8** [[TMP5]], align 4
2040 // CHECK3-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2041 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
2042 // CHECK3-NEXT: [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
2043 // CHECK3-NEXT: [[TMP10:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 1, i32 4, i8* [[TMP9]], void (i8*, i8*)* @.omp.reduction.reduction_func.17, [8 x i32]* @.gomp_critical_user_.reduction.var)
2044 // CHECK3-NEXT: switch i32 [[TMP10]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2045 // CHECK3-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2046 // CHECK3-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2047 // CHECK3-NEXT: ]
2048 // CHECK3: .omp.reduction.case1:
2049 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i32 99
2050 // CHECK3-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP11]]
2051 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
2052 // CHECK3: omp.arraycpy.body:
2053 // CHECK3-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2054 // CHECK3-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST3:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT4:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2055 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
2056 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
2057 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2058 // CHECK3-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
2059 // CHECK3-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT4]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], i32 1
2060 // CHECK3-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
2061 // CHECK3-NEXT: [[OMP_ARRAYCPY_DONE5:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT4]], [[TMP11]]
2062 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
2063 // CHECK3: omp.arraycpy.done6:
2064 // CHECK3-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2065 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2066 // CHECK3: .omp.reduction.case2:
2067 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i32 99
2068 // CHECK3-NEXT: [[OMP_ARRAYCPY_ISEMPTY7:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP14]]
2069 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY7]], label [[OMP_ARRAYCPY_DONE14:%.*]], label [[OMP_ARRAYCPY_BODY8:%.*]]
2070 // CHECK3: omp.arraycpy.body8:
2071 // CHECK3-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST9:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT12:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
2072 // CHECK3-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST10:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT11:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
2073 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], align 4
2074 // CHECK3-NEXT: [[TMP16:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 [[TMP15]] monotonic, align 4
2075 // CHECK3-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT11]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 1
2076 // CHECK3-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT12]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], i32 1
2077 // CHECK3-NEXT: [[OMP_ARRAYCPY_DONE13:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT11]], [[TMP14]]
2078 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_DONE13]], label [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_BODY8]]
2079 // CHECK3: omp.arraycpy.done14:
2080 // CHECK3-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2081 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2082 // CHECK3: .omp.reduction.default:
2083 // CHECK3-NEXT: ret void
2084 //
2085 //
2086 // CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.17
2087 // CHECK3-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
2088 // CHECK3-NEXT: entry:
2089 // CHECK3-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4
2090 // CHECK3-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4
2091 // CHECK3-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4
2092 // CHECK3-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4
2093 // CHECK3-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4
2094 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
2095 // CHECK3-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4
2096 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
2097 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
2098 // CHECK3-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
2099 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
2100 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0
2101 // CHECK3-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
2102 // CHECK3-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
2103 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i32 99
2104 // CHECK3-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
2105 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
2106 // CHECK3: omp.arraycpy.body:
2107 // CHECK3-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2108 // CHECK3-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2109 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
2110 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
2111 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
2112 // CHECK3-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
2113 // CHECK3-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
2114 // CHECK3-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
2115 // CHECK3-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
2116 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
2117 // CHECK3: omp.arraycpy.done2:
2118 // CHECK3-NEXT: ret void
2119 //
2120 //
2121 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l65
2122 // CHECK3-SAME: ([88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR1]] {
2123 // CHECK3-NEXT: entry:
2124 // CHECK3-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 4
2125 // CHECK3-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 4
2126 // CHECK3-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 4
2127 // CHECK3-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 4
2128 // CHECK3-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 4
2129 // CHECK3-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 4
2130 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [88 x i32]*, [99 x i32]*)* @.omp_outlined..20 to void (i32*, i32*, ...)*), [88 x i32]* [[TMP0]], [99 x i32]* [[TMP1]])
2131 // CHECK3-NEXT: ret void
2132 //
2133 //
2134 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..20
2135 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR1]] {
2136 // CHECK3-NEXT: entry:
2137 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2138 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2139 // CHECK3-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 4
2140 // CHECK3-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 4
2141 // CHECK3-NEXT: [[Y1:%.*]] = alloca [88 x i32], align 4
2142 // CHECK3-NEXT: [[X:%.*]] = alloca [77 x i32], align 4
2143 // CHECK3-NEXT: [[Z2:%.*]] = alloca [99 x i32], align 4
2144 // CHECK3-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
2145 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2146 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2147 // CHECK3-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 4
2148 // CHECK3-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 4
2149 // CHECK3-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 4
2150 // CHECK3-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 4
2151 // CHECK3-NEXT: [[TMP2:%.*]] = bitcast [88 x i32]* [[Y1]] to i8*
2152 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast [88 x i32]* [[TMP0]] to i8*
2153 // CHECK3-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP2]], i8* align 4 [[TMP3]], i32 352, i1 false)
2154 // CHECK3-NEXT: [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [99 x i32], [99 x i32]* [[Z2]], i32 0, i32 0
2155 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i32 99
2156 // CHECK3-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP4]]
2157 // CHECK3-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
2158 // CHECK3: omp.arrayinit.body:
2159 // CHECK3-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
2160 // CHECK3-NEXT: store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
2161 // CHECK3-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
2162 // CHECK3-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
2163 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
2164 // CHECK3: omp.arrayinit.done:
2165 // CHECK3-NEXT: [[LHS_BEGIN:%.*]] = bitcast [99 x i32]* [[TMP1]] to i32*
2166 // CHECK3-NEXT: [[RHS_BEGIN:%.*]] = bitcast [99 x i32]* [[Z2]] to i32*
2167 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
2168 // CHECK3-NEXT: [[TMP6:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
2169 // CHECK3-NEXT: store i8* [[TMP6]], i8** [[TMP5]], align 4
2170 // CHECK3-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2171 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
2172 // CHECK3-NEXT: [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
2173 // CHECK3-NEXT: [[TMP10:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 1, i32 4, i8* [[TMP9]], void (i8*, i8*)* @.omp.reduction.reduction_func.21, [8 x i32]* @.gomp_critical_user_.reduction.var)
2174 // CHECK3-NEXT: switch i32 [[TMP10]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2175 // CHECK3-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2176 // CHECK3-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2177 // CHECK3-NEXT: ]
2178 // CHECK3: .omp.reduction.case1:
2179 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i32 99
2180 // CHECK3-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP11]]
2181 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
2182 // CHECK3: omp.arraycpy.body:
2183 // CHECK3-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2184 // CHECK3-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST3:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT4:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2185 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
2186 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
2187 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2188 // CHECK3-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
2189 // CHECK3-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT4]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], i32 1
2190 // CHECK3-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
2191 // CHECK3-NEXT: [[OMP_ARRAYCPY_DONE5:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT4]], [[TMP11]]
2192 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
2193 // CHECK3: omp.arraycpy.done6:
2194 // CHECK3-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2195 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2196 // CHECK3: .omp.reduction.case2:
2197 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i32 99
2198 // CHECK3-NEXT: [[OMP_ARRAYCPY_ISEMPTY7:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP14]]
2199 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY7]], label [[OMP_ARRAYCPY_DONE14:%.*]], label [[OMP_ARRAYCPY_BODY8:%.*]]
2200 // CHECK3: omp.arraycpy.body8:
2201 // CHECK3-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST9:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT12:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
2202 // CHECK3-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST10:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT11:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
2203 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], align 4
2204 // CHECK3-NEXT: [[TMP16:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 [[TMP15]] monotonic, align 4
2205 // CHECK3-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT11]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 1
2206 // CHECK3-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT12]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], i32 1
2207 // CHECK3-NEXT: [[OMP_ARRAYCPY_DONE13:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT11]], [[TMP14]]
2208 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_DONE13]], label [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_BODY8]]
2209 // CHECK3: omp.arraycpy.done14:
2210 // CHECK3-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2211 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2212 // CHECK3: .omp.reduction.default:
2213 // CHECK3-NEXT: ret void
2214 //
2215 //
2216 // CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.21
2217 // CHECK3-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
2218 // CHECK3-NEXT: entry:
2219 // CHECK3-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4
2220 // CHECK3-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4
2221 // CHECK3-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4
2222 // CHECK3-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4
2223 // CHECK3-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4
2224 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
2225 // CHECK3-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4
2226 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
2227 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
2228 // CHECK3-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
2229 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
2230 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0
2231 // CHECK3-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
2232 // CHECK3-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
2233 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i32 99
2234 // CHECK3-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
2235 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
2236 // CHECK3: omp.arraycpy.body:
2237 // CHECK3-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2238 // CHECK3-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2239 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
2240 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
2241 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
2242 // CHECK3-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
2243 // CHECK3-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
2244 // CHECK3-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
2245 // CHECK3-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
2246 // CHECK3-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
2247 // CHECK3: omp.arraycpy.done2:
2248 // CHECK3-NEXT: ret void
2249 //
2250 //
2251 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2252 // CHECK3-SAME: () #[[ATTR6:[0-9]+]] {
2253 // CHECK3-NEXT: entry:
2254 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
2255 // CHECK3-NEXT: ret void
2256 //
2257 //
2258 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z14mapWithPrivatev_l27
2259 // CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
2260 // CHECK5-NEXT: entry:
2261 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
2262 // CHECK5-NEXT: ret void
2263 //
2264 //
2265 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined.
2266 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
2267 // CHECK5-NEXT: entry:
2268 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2269 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2270 // CHECK5-NEXT: [[X:%.*]] = alloca i32, align 4
2271 // CHECK5-NEXT: [[Y:%.*]] = alloca i32, align 4
2272 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2273 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2274 // CHECK5-NEXT: ret void
2275 //
2276 //
2277 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z19mapWithFirstprivatev_l33
2278 // CHECK5-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[Y:%.*]]) #[[ATTR0]] {
2279 // CHECK5-NEXT: entry:
2280 // CHECK5-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 8
2281 // CHECK5-NEXT: [[Y_ADDR:%.*]] = alloca i32*, align 8
2282 // CHECK5-NEXT: [[X_CASTED:%.*]] = alloca i64, align 8
2283 // CHECK5-NEXT: [[Y_CASTED:%.*]] = alloca i64, align 8
2284 // CHECK5-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 8
2285 // CHECK5-NEXT: store i32* [[Y]], i32** [[Y_ADDR]], align 8
2286 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 8
2287 // CHECK5-NEXT: [[TMP1:%.*]] = load i32*, i32** [[Y_ADDR]], align 8
2288 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
2289 // CHECK5-NEXT: [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32*
2290 // CHECK5-NEXT: store i32 [[TMP2]], i32* [[CONV]], align 4
2291 // CHECK5-NEXT: [[TMP3:%.*]] = load i64, i64* [[X_CASTED]], align 8
2292 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP1]], align 4
2293 // CHECK5-NEXT: [[CONV1:%.*]] = bitcast i64* [[Y_CASTED]] to i32*
2294 // CHECK5-NEXT: store i32 [[TMP4]], i32* [[CONV1]], align 4
2295 // CHECK5-NEXT: [[TMP5:%.*]] = load i64, i64* [[Y_CASTED]], align 8
2296 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP3]], i64 [[TMP5]])
2297 // CHECK5-NEXT: ret void
2298 //
2299 //
2300 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..1
2301 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[X:%.*]], i64 noundef [[Y:%.*]]) #[[ATTR0]] {
2302 // CHECK5-NEXT: entry:
2303 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2304 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2305 // CHECK5-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8
2306 // CHECK5-NEXT: [[Y_ADDR:%.*]] = alloca i64, align 8
2307 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2308 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2309 // CHECK5-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8
2310 // CHECK5-NEXT: store i64 [[Y]], i64* [[Y_ADDR]], align 8
2311 // CHECK5-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32*
2312 // CHECK5-NEXT: [[CONV1:%.*]] = bitcast i64* [[Y_ADDR]] to i32*
2313 // CHECK5-NEXT: ret void
2314 //
2315 //
2316 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16mapWithReductionv_l39
2317 // CHECK5-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[Y:%.*]]) #[[ATTR0]] {
2318 // CHECK5-NEXT: entry:
2319 // CHECK5-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 8
2320 // CHECK5-NEXT: [[Y_ADDR:%.*]] = alloca i32*, align 8
2321 // CHECK5-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 8
2322 // CHECK5-NEXT: store i32* [[Y]], i32** [[Y_ADDR]], align 8
2323 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 8
2324 // CHECK5-NEXT: [[TMP1:%.*]] = load i32*, i32** [[Y_ADDR]], align 8
2325 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[TMP0]], i32* [[TMP1]])
2326 // CHECK5-NEXT: ret void
2327 //
2328 //
2329 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..2
2330 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[Y:%.*]]) #[[ATTR0]] {
2331 // CHECK5-NEXT: entry:
2332 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2333 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2334 // CHECK5-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 8
2335 // CHECK5-NEXT: [[Y_ADDR:%.*]] = alloca i32*, align 8
2336 // CHECK5-NEXT: [[X1:%.*]] = alloca i32, align 4
2337 // CHECK5-NEXT: [[Y2:%.*]] = alloca i32, align 4
2338 // CHECK5-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 8
2339 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2340 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2341 // CHECK5-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 8
2342 // CHECK5-NEXT: store i32* [[Y]], i32** [[Y_ADDR]], align 8
2343 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 8
2344 // CHECK5-NEXT: [[TMP1:%.*]] = load i32*, i32** [[Y_ADDR]], align 8
2345 // CHECK5-NEXT: store i32 0, i32* [[X1]], align 4
2346 // CHECK5-NEXT: store i32 0, i32* [[Y2]], align 4
2347 // CHECK5-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
2348 // CHECK5-NEXT: [[TMP3:%.*]] = bitcast i32* [[X1]] to i8*
2349 // CHECK5-NEXT: store i8* [[TMP3]], i8** [[TMP2]], align 8
2350 // CHECK5-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
2351 // CHECK5-NEXT: [[TMP5:%.*]] = bitcast i32* [[Y2]] to i8*
2352 // CHECK5-NEXT: store i8* [[TMP5]], i8** [[TMP4]], align 8
2353 // CHECK5-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2354 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
2355 // CHECK5-NEXT: [[TMP8:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
2356 // CHECK5-NEXT: [[TMP9:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP7]], i32 2, i64 16, i8* [[TMP8]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
2357 // CHECK5-NEXT: switch i32 [[TMP9]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2358 // CHECK5-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2359 // CHECK5-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2360 // CHECK5-NEXT: ]
2361 // CHECK5: .omp.reduction.case1:
2362 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP0]], align 4
2363 // CHECK5-NEXT: [[TMP11:%.*]] = load i32, i32* [[X1]], align 4
2364 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
2365 // CHECK5-NEXT: store i32 [[ADD]], i32* [[TMP0]], align 4
2366 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP1]], align 4
2367 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[Y2]], align 4
2368 // CHECK5-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2369 // CHECK5-NEXT: store i32 [[ADD3]], i32* [[TMP1]], align 4
2370 // CHECK5-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP7]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2371 // CHECK5-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2372 // CHECK5: .omp.reduction.case2:
2373 // CHECK5-NEXT: [[TMP14:%.*]] = load i32, i32* [[X1]], align 4
2374 // CHECK5-NEXT: [[TMP15:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP14]] monotonic, align 4
2375 // CHECK5-NEXT: [[TMP16:%.*]] = load i32, i32* [[Y2]], align 4
2376 // CHECK5-NEXT: [[TMP17:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP16]] monotonic, align 4
2377 // CHECK5-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP7]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2378 // CHECK5-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2379 // CHECK5: .omp.reduction.default:
2380 // CHECK5-NEXT: ret void
2381 //
2382 //
2383 // CHECK5-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
2384 // CHECK5-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] {
2385 // CHECK5-NEXT: entry:
2386 // CHECK5-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
2387 // CHECK5-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
2388 // CHECK5-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
2389 // CHECK5-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
2390 // CHECK5-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
2391 // CHECK5-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [2 x i8*]*
2392 // CHECK5-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
2393 // CHECK5-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
2394 // CHECK5-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 0
2395 // CHECK5-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2396 // CHECK5-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
2397 // CHECK5-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 0
2398 // CHECK5-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
2399 // CHECK5-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
2400 // CHECK5-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i64 0, i64 1
2401 // CHECK5-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
2402 // CHECK5-NEXT: [[TMP14:%.*]] = bitcast i8* [[TMP13]] to i32*
2403 // CHECK5-NEXT: [[TMP15:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i64 0, i64 1
2404 // CHECK5-NEXT: [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
2405 // CHECK5-NEXT: [[TMP17:%.*]] = bitcast i8* [[TMP16]] to i32*
2406 // CHECK5-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP11]], align 4
2407 // CHECK5-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP8]], align 4
2408 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
2409 // CHECK5-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
2410 // CHECK5-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP17]], align 4
2411 // CHECK5-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP14]], align 4
2412 // CHECK5-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
2413 // CHECK5-NEXT: store i32 [[ADD2]], i32* [[TMP17]], align 4
2414 // CHECK5-NEXT: ret void
2415 //
2416 //
2417 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z7mapFromv_l45
2418 // CHECK5-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
2419 // CHECK5-NEXT: entry:
2420 // CHECK5-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 8
2421 // CHECK5-NEXT: [[X_CASTED:%.*]] = alloca i64, align 8
2422 // CHECK5-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 8
2423 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 8
2424 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2425 // CHECK5-NEXT: [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32*
2426 // CHECK5-NEXT: store i32 [[TMP1]], i32* [[CONV]], align 4
2427 // CHECK5-NEXT: [[TMP2:%.*]] = load i64, i64* [[X_CASTED]], align 8
2428 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP2]])
2429 // CHECK5-NEXT: ret void
2430 //
2431 //
2432 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..3
2433 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[X:%.*]]) #[[ATTR0]] {
2434 // CHECK5-NEXT: entry:
2435 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2436 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2437 // CHECK5-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8
2438 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2439 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2440 // CHECK5-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8
2441 // CHECK5-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32*
2442 // CHECK5-NEXT: ret void
2443 //
2444 //
2445 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5mapTov_l51
2446 // CHECK5-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
2447 // CHECK5-NEXT: entry:
2448 // CHECK5-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 8
2449 // CHECK5-NEXT: [[X_CASTED:%.*]] = alloca i64, align 8
2450 // CHECK5-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 8
2451 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 8
2452 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2453 // CHECK5-NEXT: [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32*
2454 // CHECK5-NEXT: store i32 [[TMP1]], i32* [[CONV]], align 4
2455 // CHECK5-NEXT: [[TMP2:%.*]] = load i64, i64* [[X_CASTED]], align 8
2456 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP2]])
2457 // CHECK5-NEXT: ret void
2458 //
2459 //
2460 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..4
2461 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[X:%.*]]) #[[ATTR0]] {
2462 // CHECK5-NEXT: entry:
2463 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2464 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2465 // CHECK5-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8
2466 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2467 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2468 // CHECK5-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8
2469 // CHECK5-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32*
2470 // CHECK5-NEXT: ret void
2471 //
2472 //
2473 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapAllocv_l57
2474 // CHECK5-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
2475 // CHECK5-NEXT: entry:
2476 // CHECK5-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 8
2477 // CHECK5-NEXT: [[X_CASTED:%.*]] = alloca i64, align 8
2478 // CHECK5-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 8
2479 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 8
2480 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2481 // CHECK5-NEXT: [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32*
2482 // CHECK5-NEXT: store i32 [[TMP1]], i32* [[CONV]], align 4
2483 // CHECK5-NEXT: [[TMP2:%.*]] = load i64, i64* [[X_CASTED]], align 8
2484 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP2]])
2485 // CHECK5-NEXT: ret void
2486 //
2487 //
2488 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..5
2489 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[X:%.*]]) #[[ATTR0]] {
2490 // CHECK5-NEXT: entry:
2491 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2492 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2493 // CHECK5-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8
2494 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2495 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2496 // CHECK5-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8
2497 // CHECK5-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32*
2498 // CHECK5-NEXT: ret void
2499 //
2500 //
2501 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l63
2502 // CHECK5-SAME: ([88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR0]] {
2503 // CHECK5-NEXT: entry:
2504 // CHECK5-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 8
2505 // CHECK5-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 8
2506 // CHECK5-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 8
2507 // CHECK5-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 8
2508 // CHECK5-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 8
2509 // CHECK5-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 8
2510 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [88 x i32]*, [99 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), [88 x i32]* [[TMP0]], [99 x i32]* [[TMP1]])
2511 // CHECK5-NEXT: ret void
2512 //
2513 //
2514 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..6
2515 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR0]] {
2516 // CHECK5-NEXT: entry:
2517 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2518 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2519 // CHECK5-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 8
2520 // CHECK5-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 8
2521 // CHECK5-NEXT: [[Y1:%.*]] = alloca [88 x i32], align 4
2522 // CHECK5-NEXT: [[X:%.*]] = alloca [77 x i32], align 4
2523 // CHECK5-NEXT: [[Z2:%.*]] = alloca [99 x i32], align 4
2524 // CHECK5-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
2525 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2526 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2527 // CHECK5-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 8
2528 // CHECK5-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 8
2529 // CHECK5-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 8
2530 // CHECK5-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 8
2531 // CHECK5-NEXT: [[TMP2:%.*]] = bitcast [88 x i32]* [[Y1]] to i8*
2532 // CHECK5-NEXT: [[TMP3:%.*]] = bitcast [88 x i32]* [[TMP0]] to i8*
2533 // CHECK5-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP2]], i8* align 4 [[TMP3]], i64 352, i1 false)
2534 // CHECK5-NEXT: [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [99 x i32], [99 x i32]* [[Z2]], i32 0, i32 0
2535 // CHECK5-NEXT: [[TMP4:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i64 99
2536 // CHECK5-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP4]]
2537 // CHECK5-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
2538 // CHECK5: omp.arrayinit.body:
2539 // CHECK5-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
2540 // CHECK5-NEXT: store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
2541 // CHECK5-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
2542 // CHECK5-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
2543 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
2544 // CHECK5: omp.arrayinit.done:
2545 // CHECK5-NEXT: [[LHS_BEGIN:%.*]] = bitcast [99 x i32]* [[TMP1]] to i32*
2546 // CHECK5-NEXT: [[RHS_BEGIN:%.*]] = bitcast [99 x i32]* [[Z2]] to i32*
2547 // CHECK5-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
2548 // CHECK5-NEXT: [[TMP6:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
2549 // CHECK5-NEXT: store i8* [[TMP6]], i8** [[TMP5]], align 8
2550 // CHECK5-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2551 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
2552 // CHECK5-NEXT: [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
2553 // CHECK5-NEXT: [[TMP10:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 1, i64 8, i8* [[TMP9]], void (i8*, i8*)* @.omp.reduction.reduction_func.7, [8 x i32]* @.gomp_critical_user_.reduction.var)
2554 // CHECK5-NEXT: switch i32 [[TMP10]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2555 // CHECK5-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2556 // CHECK5-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2557 // CHECK5-NEXT: ]
2558 // CHECK5: .omp.reduction.case1:
2559 // CHECK5-NEXT: [[TMP11:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i64 99
2560 // CHECK5-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP11]]
2561 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
2562 // CHECK5: omp.arraycpy.body:
2563 // CHECK5-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2564 // CHECK5-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST3:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT4:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2565 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
2566 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
2567 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2568 // CHECK5-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
2569 // CHECK5-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT4]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], i32 1
2570 // CHECK5-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
2571 // CHECK5-NEXT: [[OMP_ARRAYCPY_DONE5:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT4]], [[TMP11]]
2572 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
2573 // CHECK5: omp.arraycpy.done6:
2574 // CHECK5-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2575 // CHECK5-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2576 // CHECK5: .omp.reduction.case2:
2577 // CHECK5-NEXT: [[TMP14:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i64 99
2578 // CHECK5-NEXT: [[OMP_ARRAYCPY_ISEMPTY7:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP14]]
2579 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY7]], label [[OMP_ARRAYCPY_DONE14:%.*]], label [[OMP_ARRAYCPY_BODY8:%.*]]
2580 // CHECK5: omp.arraycpy.body8:
2581 // CHECK5-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST9:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT12:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
2582 // CHECK5-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST10:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT11:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
2583 // CHECK5-NEXT: [[TMP15:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], align 4
2584 // CHECK5-NEXT: [[TMP16:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 [[TMP15]] monotonic, align 4
2585 // CHECK5-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT11]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 1
2586 // CHECK5-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT12]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], i32 1
2587 // CHECK5-NEXT: [[OMP_ARRAYCPY_DONE13:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT11]], [[TMP14]]
2588 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_DONE13]], label [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_BODY8]]
2589 // CHECK5: omp.arraycpy.done14:
2590 // CHECK5-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2591 // CHECK5-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2592 // CHECK5: .omp.reduction.default:
2593 // CHECK5-NEXT: ret void
2594 //
2595 //
2596 // CHECK5-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.7
2597 // CHECK5-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR2]] {
2598 // CHECK5-NEXT: entry:
2599 // CHECK5-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
2600 // CHECK5-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
2601 // CHECK5-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
2602 // CHECK5-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
2603 // CHECK5-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
2604 // CHECK5-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
2605 // CHECK5-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
2606 // CHECK5-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
2607 // CHECK5-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
2608 // CHECK5-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2609 // CHECK5-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
2610 // CHECK5-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
2611 // CHECK5-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
2612 // CHECK5-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
2613 // CHECK5-NEXT: [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i64 99
2614 // CHECK5-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
2615 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
2616 // CHECK5: omp.arraycpy.body:
2617 // CHECK5-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2618 // CHECK5-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2619 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
2620 // CHECK5-NEXT: [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
2621 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
2622 // CHECK5-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
2623 // CHECK5-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
2624 // CHECK5-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
2625 // CHECK5-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
2626 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
2627 // CHECK5: omp.arraycpy.done2:
2628 // CHECK5-NEXT: ret void
2629 //
2630 //
2631 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l65
2632 // CHECK5-SAME: ([88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR0]] {
2633 // CHECK5-NEXT: entry:
2634 // CHECK5-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 8
2635 // CHECK5-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 8
2636 // CHECK5-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 8
2637 // CHECK5-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 8
2638 // CHECK5-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 8
2639 // CHECK5-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 8
2640 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [88 x i32]*, [99 x i32]*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), [88 x i32]* [[TMP0]], [99 x i32]* [[TMP1]])
2641 // CHECK5-NEXT: ret void
2642 //
2643 //
2644 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..8
2645 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR0]] {
2646 // CHECK5-NEXT: entry:
2647 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2648 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2649 // CHECK5-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 8
2650 // CHECK5-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 8
2651 // CHECK5-NEXT: [[Y1:%.*]] = alloca [88 x i32], align 4
2652 // CHECK5-NEXT: [[X:%.*]] = alloca [77 x i32], align 4
2653 // CHECK5-NEXT: [[Z2:%.*]] = alloca [99 x i32], align 4
2654 // CHECK5-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
2655 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2656 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2657 // CHECK5-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 8
2658 // CHECK5-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 8
2659 // CHECK5-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 8
2660 // CHECK5-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 8
2661 // CHECK5-NEXT: [[TMP2:%.*]] = bitcast [88 x i32]* [[Y1]] to i8*
2662 // CHECK5-NEXT: [[TMP3:%.*]] = bitcast [88 x i32]* [[TMP0]] to i8*
2663 // CHECK5-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP2]], i8* align 4 [[TMP3]], i64 352, i1 false)
2664 // CHECK5-NEXT: [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [99 x i32], [99 x i32]* [[Z2]], i32 0, i32 0
2665 // CHECK5-NEXT: [[TMP4:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i64 99
2666 // CHECK5-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP4]]
2667 // CHECK5-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
2668 // CHECK5: omp.arrayinit.body:
2669 // CHECK5-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
2670 // CHECK5-NEXT: store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
2671 // CHECK5-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
2672 // CHECK5-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
2673 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
2674 // CHECK5: omp.arrayinit.done:
2675 // CHECK5-NEXT: [[LHS_BEGIN:%.*]] = bitcast [99 x i32]* [[TMP1]] to i32*
2676 // CHECK5-NEXT: [[RHS_BEGIN:%.*]] = bitcast [99 x i32]* [[Z2]] to i32*
2677 // CHECK5-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
2678 // CHECK5-NEXT: [[TMP6:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
2679 // CHECK5-NEXT: store i8* [[TMP6]], i8** [[TMP5]], align 8
2680 // CHECK5-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2681 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
2682 // CHECK5-NEXT: [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
2683 // CHECK5-NEXT: [[TMP10:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 1, i64 8, i8* [[TMP9]], void (i8*, i8*)* @.omp.reduction.reduction_func.9, [8 x i32]* @.gomp_critical_user_.reduction.var)
2684 // CHECK5-NEXT: switch i32 [[TMP10]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2685 // CHECK5-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2686 // CHECK5-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2687 // CHECK5-NEXT: ]
2688 // CHECK5: .omp.reduction.case1:
2689 // CHECK5-NEXT: [[TMP11:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i64 99
2690 // CHECK5-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP11]]
2691 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
2692 // CHECK5: omp.arraycpy.body:
2693 // CHECK5-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2694 // CHECK5-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST3:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT4:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2695 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
2696 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
2697 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2698 // CHECK5-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
2699 // CHECK5-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT4]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], i32 1
2700 // CHECK5-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
2701 // CHECK5-NEXT: [[OMP_ARRAYCPY_DONE5:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT4]], [[TMP11]]
2702 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
2703 // CHECK5: omp.arraycpy.done6:
2704 // CHECK5-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2705 // CHECK5-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2706 // CHECK5: .omp.reduction.case2:
2707 // CHECK5-NEXT: [[TMP14:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i64 99
2708 // CHECK5-NEXT: [[OMP_ARRAYCPY_ISEMPTY7:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP14]]
2709 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY7]], label [[OMP_ARRAYCPY_DONE14:%.*]], label [[OMP_ARRAYCPY_BODY8:%.*]]
2710 // CHECK5: omp.arraycpy.body8:
2711 // CHECK5-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST9:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT12:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
2712 // CHECK5-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST10:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT11:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
2713 // CHECK5-NEXT: [[TMP15:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], align 4
2714 // CHECK5-NEXT: [[TMP16:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 [[TMP15]] monotonic, align 4
2715 // CHECK5-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT11]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 1
2716 // CHECK5-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT12]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], i32 1
2717 // CHECK5-NEXT: [[OMP_ARRAYCPY_DONE13:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT11]], [[TMP14]]
2718 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_DONE13]], label [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_BODY8]]
2719 // CHECK5: omp.arraycpy.done14:
2720 // CHECK5-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2721 // CHECK5-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2722 // CHECK5: .omp.reduction.default:
2723 // CHECK5-NEXT: ret void
2724 //
2725 //
2726 // CHECK5-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.9
2727 // CHECK5-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR2]] {
2728 // CHECK5-NEXT: entry:
2729 // CHECK5-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
2730 // CHECK5-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
2731 // CHECK5-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
2732 // CHECK5-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
2733 // CHECK5-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
2734 // CHECK5-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
2735 // CHECK5-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
2736 // CHECK5-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
2737 // CHECK5-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
2738 // CHECK5-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2739 // CHECK5-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
2740 // CHECK5-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
2741 // CHECK5-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
2742 // CHECK5-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
2743 // CHECK5-NEXT: [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i64 99
2744 // CHECK5-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
2745 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
2746 // CHECK5: omp.arraycpy.body:
2747 // CHECK5-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2748 // CHECK5-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2749 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
2750 // CHECK5-NEXT: [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
2751 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
2752 // CHECK5-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
2753 // CHECK5-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
2754 // CHECK5-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
2755 // CHECK5-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
2756 // CHECK5-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
2757 // CHECK5: omp.arraycpy.done2:
2758 // CHECK5-NEXT: ret void
2759 //
2760 //
2761 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9mapInt128v_l72
2762 // CHECK5-SAME: (i128* noundef nonnull align 16 dereferenceable(16) [[Y:%.*]], i128* noundef nonnull align 16 dereferenceable(16) [[Z:%.*]]) #[[ATTR0]] {
2763 // CHECK5-NEXT: entry:
2764 // CHECK5-NEXT: [[Y_ADDR:%.*]] = alloca i128*, align 8
2765 // CHECK5-NEXT: [[Z_ADDR:%.*]] = alloca i128*, align 8
2766 // CHECK5-NEXT: store i128* [[Y]], i128** [[Y_ADDR]], align 8
2767 // CHECK5-NEXT: store i128* [[Z]], i128** [[Z_ADDR]], align 8
2768 // CHECK5-NEXT: [[TMP0:%.*]] = load i128*, i128** [[Y_ADDR]], align 8
2769 // CHECK5-NEXT: [[TMP1:%.*]] = load i128*, i128** [[Z_ADDR]], align 8
2770 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i128*, i128*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), i128* [[TMP0]], i128* [[TMP1]])
2771 // CHECK5-NEXT: ret void
2772 //
2773 //
2774 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..10
2775 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i128* noundef nonnull align 16 dereferenceable(16) [[Y:%.*]], i128* noundef nonnull align 16 dereferenceable(16) [[Z:%.*]]) #[[ATTR0]] {
2776 // CHECK5-NEXT: entry:
2777 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2778 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2779 // CHECK5-NEXT: [[Y_ADDR:%.*]] = alloca i128*, align 8
2780 // CHECK5-NEXT: [[Z_ADDR:%.*]] = alloca i128*, align 8
2781 // CHECK5-NEXT: [[Y1:%.*]] = alloca i128, align 16
2782 // CHECK5-NEXT: [[X:%.*]] = alloca i128, align 16
2783 // CHECK5-NEXT: [[Z2:%.*]] = alloca i128, align 16
2784 // CHECK5-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
2785 // CHECK5-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i128, align 16
2786 // CHECK5-NEXT: [[ATOMIC_TEMP3:%.*]] = alloca i128, align 16
2787 // CHECK5-NEXT: [[TMP:%.*]] = alloca i128, align 16
2788 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2789 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2790 // CHECK5-NEXT: store i128* [[Y]], i128** [[Y_ADDR]], align 8
2791 // CHECK5-NEXT: store i128* [[Z]], i128** [[Z_ADDR]], align 8
2792 // CHECK5-NEXT: [[TMP0:%.*]] = load i128*, i128** [[Y_ADDR]], align 8
2793 // CHECK5-NEXT: [[TMP1:%.*]] = load i128*, i128** [[Z_ADDR]], align 8
2794 // CHECK5-NEXT: [[TMP2:%.*]] = load i128, i128* [[TMP0]], align 16
2795 // CHECK5-NEXT: store i128 [[TMP2]], i128* [[Y1]], align 16
2796 // CHECK5-NEXT: store i128 0, i128* [[Z2]], align 16
2797 // CHECK5-NEXT: [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
2798 // CHECK5-NEXT: [[TMP4:%.*]] = bitcast i128* [[Z2]] to i8*
2799 // CHECK5-NEXT: store i8* [[TMP4]], i8** [[TMP3]], align 8
2800 // CHECK5-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2801 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
2802 // CHECK5-NEXT: [[TMP7:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
2803 // CHECK5-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], i32 1, i64 8, i8* [[TMP7]], void (i8*, i8*)* @.omp.reduction.reduction_func.11, [8 x i32]* @.gomp_critical_user_.reduction.var)
2804 // CHECK5-NEXT: switch i32 [[TMP8]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2805 // CHECK5-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2806 // CHECK5-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2807 // CHECK5-NEXT: ]
2808 // CHECK5: .omp.reduction.case1:
2809 // CHECK5-NEXT: [[TMP9:%.*]] = load i128, i128* [[TMP1]], align 16
2810 // CHECK5-NEXT: [[TMP10:%.*]] = load i128, i128* [[Z2]], align 16
2811 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i128 [[TMP9]], [[TMP10]]
2812 // CHECK5-NEXT: store i128 [[ADD]], i128* [[TMP1]], align 16
2813 // CHECK5-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2814 // CHECK5-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2815 // CHECK5: .omp.reduction.case2:
2816 // CHECK5-NEXT: [[TMP11:%.*]] = load i128, i128* [[Z2]], align 16
2817 // CHECK5-NEXT: [[TMP12:%.*]] = bitcast i128* [[TMP1]] to i8*
2818 // CHECK5-NEXT: [[TMP13:%.*]] = bitcast i128* [[ATOMIC_TEMP]] to i8*
2819 // CHECK5-NEXT: call void @__atomic_load(i64 noundef 16, i8* noundef [[TMP12]], i8* noundef [[TMP13]], i32 noundef signext 0) #[[ATTR6:[0-9]+]]
2820 // CHECK5-NEXT: br label [[ATOMIC_CONT:%.*]]
2821 // CHECK5: atomic_cont:
2822 // CHECK5-NEXT: [[TMP14:%.*]] = load i128, i128* [[ATOMIC_TEMP]], align 16
2823 // CHECK5-NEXT: store i128 [[TMP14]], i128* [[TMP]], align 16
2824 // CHECK5-NEXT: [[TMP15:%.*]] = load i128, i128* [[TMP]], align 16
2825 // CHECK5-NEXT: [[TMP16:%.*]] = load i128, i128* [[Z2]], align 16
2826 // CHECK5-NEXT: [[ADD4:%.*]] = add nsw i128 [[TMP15]], [[TMP16]]
2827 // CHECK5-NEXT: store i128 [[ADD4]], i128* [[ATOMIC_TEMP3]], align 16
2828 // CHECK5-NEXT: [[TMP17:%.*]] = bitcast i128* [[TMP1]] to i8*
2829 // CHECK5-NEXT: [[TMP18:%.*]] = bitcast i128* [[ATOMIC_TEMP]] to i8*
2830 // CHECK5-NEXT: [[TMP19:%.*]] = bitcast i128* [[ATOMIC_TEMP3]] to i8*
2831 // CHECK5-NEXT: [[CALL:%.*]] = call noundef zeroext i1 @__atomic_compare_exchange(i64 noundef 16, i8* noundef [[TMP17]], i8* noundef [[TMP18]], i8* noundef [[TMP19]], i32 noundef signext 0, i32 noundef signext 0) #[[ATTR6]]
2832 // CHECK5-NEXT: br i1 [[CALL]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
2833 // CHECK5: atomic_exit:
2834 // CHECK5-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2835 // CHECK5-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2836 // CHECK5: .omp.reduction.default:
2837 // CHECK5-NEXT: ret void
2838 //
2839 //
2840 // CHECK5-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.11
2841 // CHECK5-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR2]] {
2842 // CHECK5-NEXT: entry:
2843 // CHECK5-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
2844 // CHECK5-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
2845 // CHECK5-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
2846 // CHECK5-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
2847 // CHECK5-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
2848 // CHECK5-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
2849 // CHECK5-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
2850 // CHECK5-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
2851 // CHECK5-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
2852 // CHECK5-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2853 // CHECK5-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i128*
2854 // CHECK5-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
2855 // CHECK5-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
2856 // CHECK5-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i128*
2857 // CHECK5-NEXT: [[TMP12:%.*]] = load i128, i128* [[TMP11]], align 16
2858 // CHECK5-NEXT: [[TMP13:%.*]] = load i128, i128* [[TMP8]], align 16
2859 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i128 [[TMP12]], [[TMP13]]
2860 // CHECK5-NEXT: store i128 [[ADD]], i128* [[TMP11]], align 16
2861 // CHECK5-NEXT: ret void
2862 //
2863 //
2864 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9mapInt128v_l74
2865 // CHECK5-SAME: (i128* noundef nonnull align 16 dereferenceable(16) [[Y:%.*]], i128* noundef nonnull align 16 dereferenceable(16) [[Z:%.*]]) #[[ATTR0]] {
2866 // CHECK5-NEXT: entry:
2867 // CHECK5-NEXT: [[Y_ADDR:%.*]] = alloca i128*, align 8
2868 // CHECK5-NEXT: [[Z_ADDR:%.*]] = alloca i128*, align 8
2869 // CHECK5-NEXT: store i128* [[Y]], i128** [[Y_ADDR]], align 8
2870 // CHECK5-NEXT: store i128* [[Z]], i128** [[Z_ADDR]], align 8
2871 // CHECK5-NEXT: [[TMP0:%.*]] = load i128*, i128** [[Y_ADDR]], align 8
2872 // CHECK5-NEXT: [[TMP1:%.*]] = load i128*, i128** [[Z_ADDR]], align 8
2873 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i128*, i128*)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i128* [[TMP0]], i128* [[TMP1]])
2874 // CHECK5-NEXT: ret void
2875 //
2876 //
2877 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..12
2878 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i128* noundef nonnull align 16 dereferenceable(16) [[Y:%.*]], i128* noundef nonnull align 16 dereferenceable(16) [[Z:%.*]]) #[[ATTR0]] {
2879 // CHECK5-NEXT: entry:
2880 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2881 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2882 // CHECK5-NEXT: [[Y_ADDR:%.*]] = alloca i128*, align 8
2883 // CHECK5-NEXT: [[Z_ADDR:%.*]] = alloca i128*, align 8
2884 // CHECK5-NEXT: [[Y1:%.*]] = alloca i128, align 16
2885 // CHECK5-NEXT: [[X:%.*]] = alloca i128, align 16
2886 // CHECK5-NEXT: [[Z2:%.*]] = alloca i128, align 16
2887 // CHECK5-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
2888 // CHECK5-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i128, align 16
2889 // CHECK5-NEXT: [[ATOMIC_TEMP3:%.*]] = alloca i128, align 16
2890 // CHECK5-NEXT: [[TMP:%.*]] = alloca i128, align 16
2891 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2892 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2893 // CHECK5-NEXT: store i128* [[Y]], i128** [[Y_ADDR]], align 8
2894 // CHECK5-NEXT: store i128* [[Z]], i128** [[Z_ADDR]], align 8
2895 // CHECK5-NEXT: [[TMP0:%.*]] = load i128*, i128** [[Y_ADDR]], align 8
2896 // CHECK5-NEXT: [[TMP1:%.*]] = load i128*, i128** [[Z_ADDR]], align 8
2897 // CHECK5-NEXT: [[TMP2:%.*]] = load i128, i128* [[TMP0]], align 16
2898 // CHECK5-NEXT: store i128 [[TMP2]], i128* [[Y1]], align 16
2899 // CHECK5-NEXT: store i128 0, i128* [[Z2]], align 16
2900 // CHECK5-NEXT: [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
2901 // CHECK5-NEXT: [[TMP4:%.*]] = bitcast i128* [[Z2]] to i8*
2902 // CHECK5-NEXT: store i8* [[TMP4]], i8** [[TMP3]], align 8
2903 // CHECK5-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2904 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
2905 // CHECK5-NEXT: [[TMP7:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
2906 // CHECK5-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], i32 1, i64 8, i8* [[TMP7]], void (i8*, i8*)* @.omp.reduction.reduction_func.13, [8 x i32]* @.gomp_critical_user_.reduction.var)
2907 // CHECK5-NEXT: switch i32 [[TMP8]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2908 // CHECK5-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2909 // CHECK5-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2910 // CHECK5-NEXT: ]
2911 // CHECK5: .omp.reduction.case1:
2912 // CHECK5-NEXT: [[TMP9:%.*]] = load i128, i128* [[TMP1]], align 16
2913 // CHECK5-NEXT: [[TMP10:%.*]] = load i128, i128* [[Z2]], align 16
2914 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i128 [[TMP9]], [[TMP10]]
2915 // CHECK5-NEXT: store i128 [[ADD]], i128* [[TMP1]], align 16
2916 // CHECK5-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2917 // CHECK5-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2918 // CHECK5: .omp.reduction.case2:
2919 // CHECK5-NEXT: [[TMP11:%.*]] = load i128, i128* [[Z2]], align 16
2920 // CHECK5-NEXT: [[TMP12:%.*]] = bitcast i128* [[TMP1]] to i8*
2921 // CHECK5-NEXT: [[TMP13:%.*]] = bitcast i128* [[ATOMIC_TEMP]] to i8*
2922 // CHECK5-NEXT: call void @__atomic_load(i64 noundef 16, i8* noundef [[TMP12]], i8* noundef [[TMP13]], i32 noundef signext 0) #[[ATTR6]]
2923 // CHECK5-NEXT: br label [[ATOMIC_CONT:%.*]]
2924 // CHECK5: atomic_cont:
2925 // CHECK5-NEXT: [[TMP14:%.*]] = load i128, i128* [[ATOMIC_TEMP]], align 16
2926 // CHECK5-NEXT: store i128 [[TMP14]], i128* [[TMP]], align 16
2927 // CHECK5-NEXT: [[TMP15:%.*]] = load i128, i128* [[TMP]], align 16
2928 // CHECK5-NEXT: [[TMP16:%.*]] = load i128, i128* [[Z2]], align 16
2929 // CHECK5-NEXT: [[ADD4:%.*]] = add nsw i128 [[TMP15]], [[TMP16]]
2930 // CHECK5-NEXT: store i128 [[ADD4]], i128* [[ATOMIC_TEMP3]], align 16
2931 // CHECK5-NEXT: [[TMP17:%.*]] = bitcast i128* [[TMP1]] to i8*
2932 // CHECK5-NEXT: [[TMP18:%.*]] = bitcast i128* [[ATOMIC_TEMP]] to i8*
2933 // CHECK5-NEXT: [[TMP19:%.*]] = bitcast i128* [[ATOMIC_TEMP3]] to i8*
2934 // CHECK5-NEXT: [[CALL:%.*]] = call noundef zeroext i1 @__atomic_compare_exchange(i64 noundef 16, i8* noundef [[TMP17]], i8* noundef [[TMP18]], i8* noundef [[TMP19]], i32 noundef signext 0, i32 noundef signext 0) #[[ATTR6]]
2935 // CHECK5-NEXT: br i1 [[CALL]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
2936 // CHECK5: atomic_exit:
2937 // CHECK5-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2938 // CHECK5-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
2939 // CHECK5: .omp.reduction.default:
2940 // CHECK5-NEXT: ret void
2941 //
2942 //
2943 // CHECK5-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.13
2944 // CHECK5-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR2]] {
2945 // CHECK5-NEXT: entry:
2946 // CHECK5-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
2947 // CHECK5-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
2948 // CHECK5-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
2949 // CHECK5-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
2950 // CHECK5-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
2951 // CHECK5-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
2952 // CHECK5-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
2953 // CHECK5-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
2954 // CHECK5-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
2955 // CHECK5-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2956 // CHECK5-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i128*
2957 // CHECK5-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
2958 // CHECK5-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
2959 // CHECK5-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i128*
2960 // CHECK5-NEXT: [[TMP12:%.*]] = load i128, i128* [[TMP11]], align 16
2961 // CHECK5-NEXT: [[TMP13:%.*]] = load i128, i128* [[TMP8]], align 16
2962 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i128 [[TMP12]], [[TMP13]]
2963 // CHECK5-NEXT: store i128 [[ADD]], i128* [[TMP11]], align 16
2964 // CHECK5-NEXT: ret void
2965 //
2966 //
2967 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z14mapWithPrivatev_l27
2968 // CHECK7-SAME: () #[[ATTR0:[0-9]+]] {
2969 // CHECK7-NEXT: entry:
2970 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
2971 // CHECK7-NEXT: ret void
2972 //
2973 //
2974 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined.
2975 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
2976 // CHECK7-NEXT: entry:
2977 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2978 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2979 // CHECK7-NEXT: [[X:%.*]] = alloca i32, align 4
2980 // CHECK7-NEXT: [[Y:%.*]] = alloca i32, align 4
2981 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2982 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2983 // CHECK7-NEXT: ret void
2984 //
2985 //
2986 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z19mapWithFirstprivatev_l33
2987 // CHECK7-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[Y:%.*]]) #[[ATTR0]] {
2988 // CHECK7-NEXT: entry:
2989 // CHECK7-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 4
2990 // CHECK7-NEXT: [[Y_ADDR:%.*]] = alloca i32*, align 4
2991 // CHECK7-NEXT: [[X_CASTED:%.*]] = alloca i32, align 4
2992 // CHECK7-NEXT: [[Y_CASTED:%.*]] = alloca i32, align 4
2993 // CHECK7-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 4
2994 // CHECK7-NEXT: store i32* [[Y]], i32** [[Y_ADDR]], align 4
2995 // CHECK7-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 4
2996 // CHECK7-NEXT: [[TMP1:%.*]] = load i32*, i32** [[Y_ADDR]], align 4
2997 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
2998 // CHECK7-NEXT: store i32 [[TMP2]], i32* [[X_CASTED]], align 4
2999 // CHECK7-NEXT: [[TMP3:%.*]] = load i32, i32* [[X_CASTED]], align 4
3000 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP1]], align 4
3001 // CHECK7-NEXT: store i32 [[TMP4]], i32* [[Y_CASTED]], align 4
3002 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[Y_CASTED]], align 4
3003 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP3]], i32 [[TMP5]])
3004 // CHECK7-NEXT: ret void
3005 //
3006 //
3007 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..1
3008 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[X:%.*]], i32 noundef [[Y:%.*]]) #[[ATTR0]] {
3009 // CHECK7-NEXT: entry:
3010 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3011 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3012 // CHECK7-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4
3013 // CHECK7-NEXT: [[Y_ADDR:%.*]] = alloca i32, align 4
3014 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3015 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3016 // CHECK7-NEXT: store i32 [[X]], i32* [[X_ADDR]], align 4
3017 // CHECK7-NEXT: store i32 [[Y]], i32* [[Y_ADDR]], align 4
3018 // CHECK7-NEXT: ret void
3019 //
3020 //
3021 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z16mapWithReductionv_l39
3022 // CHECK7-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[Y:%.*]]) #[[ATTR0]] {
3023 // CHECK7-NEXT: entry:
3024 // CHECK7-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 4
3025 // CHECK7-NEXT: [[Y_ADDR:%.*]] = alloca i32*, align 4
3026 // CHECK7-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 4
3027 // CHECK7-NEXT: store i32* [[Y]], i32** [[Y_ADDR]], align 4
3028 // CHECK7-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 4
3029 // CHECK7-NEXT: [[TMP1:%.*]] = load i32*, i32** [[Y_ADDR]], align 4
3030 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[TMP0]], i32* [[TMP1]])
3031 // CHECK7-NEXT: ret void
3032 //
3033 //
3034 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..2
3035 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[Y:%.*]]) #[[ATTR0]] {
3036 // CHECK7-NEXT: entry:
3037 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3038 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3039 // CHECK7-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 4
3040 // CHECK7-NEXT: [[Y_ADDR:%.*]] = alloca i32*, align 4
3041 // CHECK7-NEXT: [[X1:%.*]] = alloca i32, align 4
3042 // CHECK7-NEXT: [[Y2:%.*]] = alloca i32, align 4
3043 // CHECK7-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [2 x i8*], align 4
3044 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3045 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3046 // CHECK7-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 4
3047 // CHECK7-NEXT: store i32* [[Y]], i32** [[Y_ADDR]], align 4
3048 // CHECK7-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 4
3049 // CHECK7-NEXT: [[TMP1:%.*]] = load i32*, i32** [[Y_ADDR]], align 4
3050 // CHECK7-NEXT: store i32 0, i32* [[X1]], align 4
3051 // CHECK7-NEXT: store i32 0, i32* [[Y2]], align 4
3052 // CHECK7-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
3053 // CHECK7-NEXT: [[TMP3:%.*]] = bitcast i32* [[X1]] to i8*
3054 // CHECK7-NEXT: store i8* [[TMP3]], i8** [[TMP2]], align 4
3055 // CHECK7-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 1
3056 // CHECK7-NEXT: [[TMP5:%.*]] = bitcast i32* [[Y2]] to i8*
3057 // CHECK7-NEXT: store i8* [[TMP5]], i8** [[TMP4]], align 4
3058 // CHECK7-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3059 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
3060 // CHECK7-NEXT: [[TMP8:%.*]] = bitcast [2 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
3061 // CHECK7-NEXT: [[TMP9:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP7]], i32 2, i32 8, i8* [[TMP8]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
3062 // CHECK7-NEXT: switch i32 [[TMP9]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3063 // CHECK7-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3064 // CHECK7-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3065 // CHECK7-NEXT: ]
3066 // CHECK7: .omp.reduction.case1:
3067 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP0]], align 4
3068 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[X1]], align 4
3069 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
3070 // CHECK7-NEXT: store i32 [[ADD]], i32* [[TMP0]], align 4
3071 // CHECK7-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP1]], align 4
3072 // CHECK7-NEXT: [[TMP13:%.*]] = load i32, i32* [[Y2]], align 4
3073 // CHECK7-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3074 // CHECK7-NEXT: store i32 [[ADD3]], i32* [[TMP1]], align 4
3075 // CHECK7-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP7]], [8 x i32]* @.gomp_critical_user_.reduction.var)
3076 // CHECK7-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
3077 // CHECK7: .omp.reduction.case2:
3078 // CHECK7-NEXT: [[TMP14:%.*]] = load i32, i32* [[X1]], align 4
3079 // CHECK7-NEXT: [[TMP15:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP14]] monotonic, align 4
3080 // CHECK7-NEXT: [[TMP16:%.*]] = load i32, i32* [[Y2]], align 4
3081 // CHECK7-NEXT: [[TMP17:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP16]] monotonic, align 4
3082 // CHECK7-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP7]], [8 x i32]* @.gomp_critical_user_.reduction.var)
3083 // CHECK7-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
3084 // CHECK7: .omp.reduction.default:
3085 // CHECK7-NEXT: ret void
3086 //
3087 //
3088 // CHECK7-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
3089 // CHECK7-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] {
3090 // CHECK7-NEXT: entry:
3091 // CHECK7-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4
3092 // CHECK7-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4
3093 // CHECK7-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4
3094 // CHECK7-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4
3095 // CHECK7-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4
3096 // CHECK7-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [2 x i8*]*
3097 // CHECK7-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4
3098 // CHECK7-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [2 x i8*]*
3099 // CHECK7-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 0
3100 // CHECK7-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
3101 // CHECK7-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
3102 // CHECK7-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i32 0, i32 0
3103 // CHECK7-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
3104 // CHECK7-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
3105 // CHECK7-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP5]], i32 0, i32 1
3106 // CHECK7-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 4
3107 // CHECK7-NEXT: [[TMP14:%.*]] = bitcast i8* [[TMP13]] to i32*
3108 // CHECK7-NEXT: [[TMP15:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[TMP3]], i32 0, i32 1
3109 // CHECK7-NEXT: [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 4
3110 // CHECK7-NEXT: [[TMP17:%.*]] = bitcast i8* [[TMP16]] to i32*
3111 // CHECK7-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP11]], align 4
3112 // CHECK7-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP8]], align 4
3113 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
3114 // CHECK7-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
3115 // CHECK7-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP17]], align 4
3116 // CHECK7-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP14]], align 4
3117 // CHECK7-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
3118 // CHECK7-NEXT: store i32 [[ADD2]], i32* [[TMP17]], align 4
3119 // CHECK7-NEXT: ret void
3120 //
3121 //
3122 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z7mapFromv_l45
3123 // CHECK7-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
3124 // CHECK7-NEXT: entry:
3125 // CHECK7-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 4
3126 // CHECK7-NEXT: [[X_CASTED:%.*]] = alloca i32, align 4
3127 // CHECK7-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 4
3128 // CHECK7-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 4
3129 // CHECK7-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3130 // CHECK7-NEXT: store i32 [[TMP1]], i32* [[X_CASTED]], align 4
3131 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[X_CASTED]], align 4
3132 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP2]])
3133 // CHECK7-NEXT: ret void
3134 //
3135 //
3136 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..3
3137 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[X:%.*]]) #[[ATTR0]] {
3138 // CHECK7-NEXT: entry:
3139 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3140 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3141 // CHECK7-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4
3142 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3143 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3144 // CHECK7-NEXT: store i32 [[X]], i32* [[X_ADDR]], align 4
3145 // CHECK7-NEXT: ret void
3146 //
3147 //
3148 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5mapTov_l51
3149 // CHECK7-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
3150 // CHECK7-NEXT: entry:
3151 // CHECK7-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 4
3152 // CHECK7-NEXT: [[X_CASTED:%.*]] = alloca i32, align 4
3153 // CHECK7-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 4
3154 // CHECK7-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 4
3155 // CHECK7-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3156 // CHECK7-NEXT: store i32 [[TMP1]], i32* [[X_CASTED]], align 4
3157 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[X_CASTED]], align 4
3158 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP2]])
3159 // CHECK7-NEXT: ret void
3160 //
3161 //
3162 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..4
3163 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[X:%.*]]) #[[ATTR0]] {
3164 // CHECK7-NEXT: entry:
3165 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3166 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3167 // CHECK7-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4
3168 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3169 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3170 // CHECK7-NEXT: store i32 [[X]], i32* [[X_ADDR]], align 4
3171 // CHECK7-NEXT: ret void
3172 //
3173 //
3174 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapAllocv_l57
3175 // CHECK7-SAME: (i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR0]] {
3176 // CHECK7-NEXT: entry:
3177 // CHECK7-NEXT: [[X_ADDR:%.*]] = alloca i32*, align 4
3178 // CHECK7-NEXT: [[X_CASTED:%.*]] = alloca i32, align 4
3179 // CHECK7-NEXT: store i32* [[X]], i32** [[X_ADDR]], align 4
3180 // CHECK7-NEXT: [[TMP0:%.*]] = load i32*, i32** [[X_ADDR]], align 4
3181 // CHECK7-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3182 // CHECK7-NEXT: store i32 [[TMP1]], i32* [[X_CASTED]], align 4
3183 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[X_CASTED]], align 4
3184 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32 [[TMP2]])
3185 // CHECK7-NEXT: ret void
3186 //
3187 //
3188 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..5
3189 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[X:%.*]]) #[[ATTR0]] {
3190 // CHECK7-NEXT: entry:
3191 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3192 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3193 // CHECK7-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4
3194 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3195 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3196 // CHECK7-NEXT: store i32 [[X]], i32* [[X_ADDR]], align 4
3197 // CHECK7-NEXT: ret void
3198 //
3199 //
3200 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l63
3201 // CHECK7-SAME: ([88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR0]] {
3202 // CHECK7-NEXT: entry:
3203 // CHECK7-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 4
3204 // CHECK7-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 4
3205 // CHECK7-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 4
3206 // CHECK7-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 4
3207 // CHECK7-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 4
3208 // CHECK7-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 4
3209 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [88 x i32]*, [99 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), [88 x i32]* [[TMP0]], [99 x i32]* [[TMP1]])
3210 // CHECK7-NEXT: ret void
3211 //
3212 //
3213 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..6
3214 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR0]] {
3215 // CHECK7-NEXT: entry:
3216 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3217 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3218 // CHECK7-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 4
3219 // CHECK7-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 4
3220 // CHECK7-NEXT: [[Y1:%.*]] = alloca [88 x i32], align 4
3221 // CHECK7-NEXT: [[X:%.*]] = alloca [77 x i32], align 4
3222 // CHECK7-NEXT: [[Z2:%.*]] = alloca [99 x i32], align 4
3223 // CHECK7-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
3224 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3225 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3226 // CHECK7-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 4
3227 // CHECK7-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 4
3228 // CHECK7-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 4
3229 // CHECK7-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 4
3230 // CHECK7-NEXT: [[TMP2:%.*]] = bitcast [88 x i32]* [[Y1]] to i8*
3231 // CHECK7-NEXT: [[TMP3:%.*]] = bitcast [88 x i32]* [[TMP0]] to i8*
3232 // CHECK7-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP2]], i8* align 4 [[TMP3]], i32 352, i1 false)
3233 // CHECK7-NEXT: [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [99 x i32], [99 x i32]* [[Z2]], i32 0, i32 0
3234 // CHECK7-NEXT: [[TMP4:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i32 99
3235 // CHECK7-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP4]]
3236 // CHECK7-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
3237 // CHECK7: omp.arrayinit.body:
3238 // CHECK7-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
3239 // CHECK7-NEXT: store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
3240 // CHECK7-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
3241 // CHECK7-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
3242 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
3243 // CHECK7: omp.arrayinit.done:
3244 // CHECK7-NEXT: [[LHS_BEGIN:%.*]] = bitcast [99 x i32]* [[TMP1]] to i32*
3245 // CHECK7-NEXT: [[RHS_BEGIN:%.*]] = bitcast [99 x i32]* [[Z2]] to i32*
3246 // CHECK7-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
3247 // CHECK7-NEXT: [[TMP6:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
3248 // CHECK7-NEXT: store i8* [[TMP6]], i8** [[TMP5]], align 4
3249 // CHECK7-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3250 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3251 // CHECK7-NEXT: [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
3252 // CHECK7-NEXT: [[TMP10:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 1, i32 4, i8* [[TMP9]], void (i8*, i8*)* @.omp.reduction.reduction_func.7, [8 x i32]* @.gomp_critical_user_.reduction.var)
3253 // CHECK7-NEXT: switch i32 [[TMP10]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3254 // CHECK7-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3255 // CHECK7-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3256 // CHECK7-NEXT: ]
3257 // CHECK7: .omp.reduction.case1:
3258 // CHECK7-NEXT: [[TMP11:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i32 99
3259 // CHECK7-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP11]]
3260 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
3261 // CHECK7: omp.arraycpy.body:
3262 // CHECK7-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
3263 // CHECK7-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST3:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT4:%.*]], [[OMP_ARRAYCPY_BODY]] ]
3264 // CHECK7-NEXT: [[TMP12:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
3265 // CHECK7-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
3266 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3267 // CHECK7-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
3268 // CHECK7-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT4]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], i32 1
3269 // CHECK7-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
3270 // CHECK7-NEXT: [[OMP_ARRAYCPY_DONE5:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT4]], [[TMP11]]
3271 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
3272 // CHECK7: omp.arraycpy.done6:
3273 // CHECK7-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
3274 // CHECK7-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
3275 // CHECK7: .omp.reduction.case2:
3276 // CHECK7-NEXT: [[TMP14:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i32 99
3277 // CHECK7-NEXT: [[OMP_ARRAYCPY_ISEMPTY7:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP14]]
3278 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY7]], label [[OMP_ARRAYCPY_DONE14:%.*]], label [[OMP_ARRAYCPY_BODY8:%.*]]
3279 // CHECK7: omp.arraycpy.body8:
3280 // CHECK7-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST9:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT12:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
3281 // CHECK7-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST10:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT11:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
3282 // CHECK7-NEXT: [[TMP15:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], align 4
3283 // CHECK7-NEXT: [[TMP16:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 [[TMP15]] monotonic, align 4
3284 // CHECK7-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT11]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 1
3285 // CHECK7-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT12]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], i32 1
3286 // CHECK7-NEXT: [[OMP_ARRAYCPY_DONE13:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT11]], [[TMP14]]
3287 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_DONE13]], label [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_BODY8]]
3288 // CHECK7: omp.arraycpy.done14:
3289 // CHECK7-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
3290 // CHECK7-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
3291 // CHECK7: .omp.reduction.default:
3292 // CHECK7-NEXT: ret void
3293 //
3294 //
3295 // CHECK7-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.7
3296 // CHECK7-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR2]] {
3297 // CHECK7-NEXT: entry:
3298 // CHECK7-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4
3299 // CHECK7-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4
3300 // CHECK7-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4
3301 // CHECK7-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4
3302 // CHECK7-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4
3303 // CHECK7-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
3304 // CHECK7-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4
3305 // CHECK7-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
3306 // CHECK7-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
3307 // CHECK7-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
3308 // CHECK7-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
3309 // CHECK7-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0
3310 // CHECK7-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
3311 // CHECK7-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
3312 // CHECK7-NEXT: [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i32 99
3313 // CHECK7-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
3314 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
3315 // CHECK7: omp.arraycpy.body:
3316 // CHECK7-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
3317 // CHECK7-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
3318 // CHECK7-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
3319 // CHECK7-NEXT: [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
3320 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
3321 // CHECK7-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
3322 // CHECK7-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
3323 // CHECK7-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
3324 // CHECK7-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
3325 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
3326 // CHECK7: omp.arraycpy.done2:
3327 // CHECK7-NEXT: ret void
3328 //
3329 //
3330 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z8mapArrayv_l65
3331 // CHECK7-SAME: ([88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR0]] {
3332 // CHECK7-NEXT: entry:
3333 // CHECK7-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 4
3334 // CHECK7-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 4
3335 // CHECK7-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 4
3336 // CHECK7-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 4
3337 // CHECK7-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 4
3338 // CHECK7-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 4
3339 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [88 x i32]*, [99 x i32]*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), [88 x i32]* [[TMP0]], [99 x i32]* [[TMP1]])
3340 // CHECK7-NEXT: ret void
3341 //
3342 //
3343 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..8
3344 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [88 x i32]* noundef nonnull align 4 dereferenceable(352) [[Y:%.*]], [99 x i32]* noundef nonnull align 4 dereferenceable(396) [[Z:%.*]]) #[[ATTR0]] {
3345 // CHECK7-NEXT: entry:
3346 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3347 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3348 // CHECK7-NEXT: [[Y_ADDR:%.*]] = alloca [88 x i32]*, align 4
3349 // CHECK7-NEXT: [[Z_ADDR:%.*]] = alloca [99 x i32]*, align 4
3350 // CHECK7-NEXT: [[Y1:%.*]] = alloca [88 x i32], align 4
3351 // CHECK7-NEXT: [[X:%.*]] = alloca [77 x i32], align 4
3352 // CHECK7-NEXT: [[Z2:%.*]] = alloca [99 x i32], align 4
3353 // CHECK7-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4
3354 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3355 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3356 // CHECK7-NEXT: store [88 x i32]* [[Y]], [88 x i32]** [[Y_ADDR]], align 4
3357 // CHECK7-NEXT: store [99 x i32]* [[Z]], [99 x i32]** [[Z_ADDR]], align 4
3358 // CHECK7-NEXT: [[TMP0:%.*]] = load [88 x i32]*, [88 x i32]** [[Y_ADDR]], align 4
3359 // CHECK7-NEXT: [[TMP1:%.*]] = load [99 x i32]*, [99 x i32]** [[Z_ADDR]], align 4
3360 // CHECK7-NEXT: [[TMP2:%.*]] = bitcast [88 x i32]* [[Y1]] to i8*
3361 // CHECK7-NEXT: [[TMP3:%.*]] = bitcast [88 x i32]* [[TMP0]] to i8*
3362 // CHECK7-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP2]], i8* align 4 [[TMP3]], i32 352, i1 false)
3363 // CHECK7-NEXT: [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [99 x i32], [99 x i32]* [[Z2]], i32 0, i32 0
3364 // CHECK7-NEXT: [[TMP4:%.*]] = getelementptr i32, i32* [[ARRAY_BEGIN]], i32 99
3365 // CHECK7-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i32* [[ARRAY_BEGIN]], [[TMP4]]
3366 // CHECK7-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
3367 // CHECK7: omp.arrayinit.body:
3368 // CHECK7-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
3369 // CHECK7-NEXT: store i32 0, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
3370 // CHECK7-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
3371 // CHECK7-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
3372 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
3373 // CHECK7: omp.arrayinit.done:
3374 // CHECK7-NEXT: [[LHS_BEGIN:%.*]] = bitcast [99 x i32]* [[TMP1]] to i32*
3375 // CHECK7-NEXT: [[RHS_BEGIN:%.*]] = bitcast [99 x i32]* [[Z2]] to i32*
3376 // CHECK7-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
3377 // CHECK7-NEXT: [[TMP6:%.*]] = bitcast i32* [[RHS_BEGIN]] to i8*
3378 // CHECK7-NEXT: store i8* [[TMP6]], i8** [[TMP5]], align 4
3379 // CHECK7-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3380 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3381 // CHECK7-NEXT: [[TMP9:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
3382 // CHECK7-NEXT: [[TMP10:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 1, i32 4, i8* [[TMP9]], void (i8*, i8*)* @.omp.reduction.reduction_func.9, [8 x i32]* @.gomp_critical_user_.reduction.var)
3383 // CHECK7-NEXT: switch i32 [[TMP10]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3384 // CHECK7-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3385 // CHECK7-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3386 // CHECK7-NEXT: ]
3387 // CHECK7: .omp.reduction.case1:
3388 // CHECK7-NEXT: [[TMP11:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i32 99
3389 // CHECK7-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP11]]
3390 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
3391 // CHECK7: omp.arraycpy.body:
3392 // CHECK7-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
3393 // CHECK7-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST3:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT4:%.*]], [[OMP_ARRAYCPY_BODY]] ]
3394 // CHECK7-NEXT: [[TMP12:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
3395 // CHECK7-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
3396 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3397 // CHECK7-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], align 4
3398 // CHECK7-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT4]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST3]], i32 1
3399 // CHECK7-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
3400 // CHECK7-NEXT: [[OMP_ARRAYCPY_DONE5:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT4]], [[TMP11]]
3401 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
3402 // CHECK7: omp.arraycpy.done6:
3403 // CHECK7-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
3404 // CHECK7-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
3405 // CHECK7: .omp.reduction.case2:
3406 // CHECK7-NEXT: [[TMP14:%.*]] = getelementptr i32, i32* [[LHS_BEGIN]], i32 99
3407 // CHECK7-NEXT: [[OMP_ARRAYCPY_ISEMPTY7:%.*]] = icmp eq i32* [[LHS_BEGIN]], [[TMP14]]
3408 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY7]], label [[OMP_ARRAYCPY_DONE14:%.*]], label [[OMP_ARRAYCPY_BODY8:%.*]]
3409 // CHECK7: omp.arraycpy.body8:
3410 // CHECK7-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST9:%.*]] = phi i32* [ [[RHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT12:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
3411 // CHECK7-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST10:%.*]] = phi i32* [ [[LHS_BEGIN]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT11:%.*]], [[OMP_ARRAYCPY_BODY8]] ]
3412 // CHECK7-NEXT: [[TMP15:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], align 4
3413 // CHECK7-NEXT: [[TMP16:%.*]] = atomicrmw add i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 [[TMP15]] monotonic, align 4
3414 // CHECK7-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT11]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST10]], i32 1
3415 // CHECK7-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT12]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST9]], i32 1
3416 // CHECK7-NEXT: [[OMP_ARRAYCPY_DONE13:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT11]], [[TMP14]]
3417 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_DONE13]], label [[OMP_ARRAYCPY_DONE14]], label [[OMP_ARRAYCPY_BODY8]]
3418 // CHECK7: omp.arraycpy.done14:
3419 // CHECK7-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], [8 x i32]* @.gomp_critical_user_.reduction.var)
3420 // CHECK7-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
3421 // CHECK7: .omp.reduction.default:
3422 // CHECK7-NEXT: ret void
3423 //
3424 //
3425 // CHECK7-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.9
3426 // CHECK7-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR2]] {
3427 // CHECK7-NEXT: entry:
3428 // CHECK7-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4
3429 // CHECK7-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4
3430 // CHECK7-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4
3431 // CHECK7-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4
3432 // CHECK7-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4
3433 // CHECK7-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
3434 // CHECK7-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4
3435 // CHECK7-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
3436 // CHECK7-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0
3437 // CHECK7-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
3438 // CHECK7-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
3439 // CHECK7-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0
3440 // CHECK7-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4
3441 // CHECK7-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
3442 // CHECK7-NEXT: [[TMP12:%.*]] = getelementptr i32, i32* [[TMP11]], i32 99
3443 // CHECK7-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i32* [[TMP11]], [[TMP12]]
3444 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
3445 // CHECK7: omp.arraycpy.body:
3446 // CHECK7-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i32* [ [[TMP8]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
3447 // CHECK7-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i32* [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
3448 // CHECK7-NEXT: [[TMP13:%.*]] = load i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
3449 // CHECK7-NEXT: [[TMP14:%.*]] = load i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
3450 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
3451 // CHECK7-NEXT: store i32 [[ADD]], i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
3452 // CHECK7-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
3453 // CHECK7-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, i32* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
3454 // CHECK7-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i32* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
3455 // CHECK7-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
3456 // CHECK7: omp.arraycpy.done2:
3457 // CHECK7-NEXT: ret void
3458 //
3459