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 -verify -fopenmp -fopenmp-version=45 -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 -fopenmp -fopenmp-version=45 -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 -fopenmp -fopenmp-version=45 -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=45 -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=45 -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=45 -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 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
11 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
12 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -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 --implicit-check-not="{{__kmpc|__tgt}}"
13 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
14 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
15 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -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 --implicit-check-not="{{__kmpc|__tgt}}"
16
17 // Test target codegen - host bc file has to be created first.
18 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
19 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -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=CHECK9
20 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -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
21 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -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=CHECK9
22 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
23 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -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=CHECK11
24 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -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
25 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -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=CHECK11
26
27 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
28 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -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 --implicit-check-not="{{__kmpc|__tgt}}"
29 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -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
30 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -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 --implicit-check-not="{{__kmpc|__tgt}}"
31 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
32 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -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 --implicit-check-not="{{__kmpc|__tgt}}"
33 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -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
34 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -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 --implicit-check-not="{{__kmpc|__tgt}}"
35
36 // Test host codegen.
37 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
38 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
39 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -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
40 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
41 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
42 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -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
43
44 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
45 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
46 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -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 --implicit-check-not="{{__kmpc|__tgt}}"
47 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
48 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
49 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -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 --implicit-check-not="{{__kmpc|__tgt}}"
50
51 // Test target codegen - host bc file has to be created first.
52 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
53 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -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=CHECK9
54 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -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
55 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -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=CHECK9
56 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
57 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -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=CHECK11
58 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -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
59 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -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=CHECK11
60
61 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
62 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -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 --implicit-check-not="{{__kmpc|__tgt}}"
63 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -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
64 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -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 --implicit-check-not="{{__kmpc|__tgt}}"
65 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
66 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -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 --implicit-check-not="{{__kmpc|__tgt}}"
67 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -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
68 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -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 --implicit-check-not="{{__kmpc|__tgt}}"
69
70 // expected-no-diagnostics
71 #ifndef HEADER
72 #define HEADER
73
74
75
76
77 // We have 6 target regions
78
79
80
81 // Check target registration is registered as a Ctor.
82
83
84 template<typename tx>
ftemplate(int n)85 tx ftemplate(int n) {
86 tx a = 0;
87
88 #pragma omp target teams num_teams(tx(20))
89 {
90 }
91
92 short b = 1;
93 #pragma omp target teams num_teams(b)
94 {
95 a += b;
96 }
97
98 return a;
99 }
100
101 static
fstatic(int n)102 int fstatic(int n) {
103
104 #pragma omp target teams num_teams(n)
105 {
106 }
107
108 #pragma omp target teams num_teams(32+n)
109 {
110 }
111
112 return n+1;
113 }
114
115 struct S1 {
116 double a;
117
r1S1118 int r1(int n){
119 int b = 1;
120
121 #pragma omp target teams num_teams(n-b)
122 {
123 this->a = (double)b + 1.5;
124 }
125
126 #pragma omp target teams num_teams(1024)
127 {
128 this->a = 2.5;
129 }
130
131 return (int)a;
132 }
133 };
134
bar(int n)135 int bar(int n){
136 int a = 0;
137
138 S1 S;
139 a += S.r1(n);
140
141 a += fstatic(n);
142
143 a += ftemplate<int>(n);
144
145 return a;
146 }
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168 // Check that the offloading functions are emitted and that the parallel function
169 // is appropriately guarded.
170
171
172
173
174
175
176 #endif
177 // CHECK1-LABEL: define {{[^@]+}}@_Z3bari
178 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
179 // CHECK1-NEXT: entry:
180 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
181 // CHECK1-NEXT: [[A:%.*]] = alloca i32, align 4
182 // CHECK1-NEXT: [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
183 // CHECK1-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
184 // CHECK1-NEXT: store i32 0, i32* [[A]], align 4
185 // CHECK1-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
186 // CHECK1-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP0]])
187 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[A]], align 4
188 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
189 // CHECK1-NEXT: store i32 [[ADD]], i32* [[A]], align 4
190 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
191 // CHECK1-NEXT: [[CALL1:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP2]])
192 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[A]], align 4
193 // CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
194 // CHECK1-NEXT: store i32 [[ADD2]], i32* [[A]], align 4
195 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
196 // CHECK1-NEXT: [[CALL3:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP4]])
197 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[A]], align 4
198 // CHECK1-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
199 // CHECK1-NEXT: store i32 [[ADD4]], i32* [[A]], align 4
200 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[A]], align 4
201 // CHECK1-NEXT: ret i32 [[TMP6]]
202 //
203 //
204 // CHECK1-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
205 // CHECK1-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
206 // CHECK1-NEXT: entry:
207 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
208 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
209 // CHECK1-NEXT: [[B:%.*]] = alloca i32, align 4
210 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
211 // CHECK1-NEXT: [[B_CASTED:%.*]] = alloca i64, align 8
212 // CHECK1-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
213 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
214 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
215 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
216 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
217 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
218 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
219 // CHECK1-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
220 // CHECK1-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
221 // CHECK1-NEXT: [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
222 // CHECK1-NEXT: store i32 1, i32* [[B]], align 4
223 // CHECK1-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
224 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[B]], align 4
225 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
226 // CHECK1-NEXT: store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
227 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[B]], align 4
228 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
229 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[CONV]], align 4
230 // CHECK1-NEXT: [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
231 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
232 // CHECK1-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
233 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[CONV2]], align 4
234 // CHECK1-NEXT: [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
235 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
236 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
237 // CHECK1-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
238 // CHECK1-NEXT: store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 8
239 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
240 // CHECK1-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
241 // CHECK1-NEXT: store double* [[A]], double** [[TMP9]], align 8
242 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
243 // CHECK1-NEXT: store i8* null, i8** [[TMP10]], align 8
244 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
245 // CHECK1-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
246 // CHECK1-NEXT: store i64 [[TMP3]], i64* [[TMP12]], align 8
247 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
248 // CHECK1-NEXT: [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
249 // CHECK1-NEXT: store i64 [[TMP3]], i64* [[TMP14]], align 8
250 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
251 // CHECK1-NEXT: store i8* null, i8** [[TMP15]], align 8
252 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
253 // CHECK1-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
254 // CHECK1-NEXT: store i64 [[TMP5]], i64* [[TMP17]], align 8
255 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
256 // CHECK1-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
257 // CHECK1-NEXT: store i64 [[TMP5]], i64* [[TMP19]], align 8
258 // CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
259 // CHECK1-NEXT: store i8* null, i8** [[TMP20]], align 8
260 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
261 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
262 // CHECK1-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
263 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
264 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
265 // CHECK1-NEXT: store i32 1, i32* [[TMP24]], align 4
266 // CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
267 // CHECK1-NEXT: store i32 3, i32* [[TMP25]], align 4
268 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
269 // CHECK1-NEXT: store i8** [[TMP21]], i8*** [[TMP26]], align 8
270 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
271 // CHECK1-NEXT: store i8** [[TMP22]], i8*** [[TMP27]], align 8
272 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
273 // CHECK1-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP28]], align 8
274 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
275 // CHECK1-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP29]], align 8
276 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
277 // CHECK1-NEXT: store i8** null, i8*** [[TMP30]], align 8
278 // CHECK1-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
279 // CHECK1-NEXT: store i8** null, i8*** [[TMP31]], align 8
280 // CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
281 // CHECK1-NEXT: store i64 0, i64* [[TMP32]], align 8
282 // CHECK1-NEXT: [[TMP33:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i32 [[TMP23]], i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
283 // CHECK1-NEXT: [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
284 // CHECK1-NEXT: br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
285 // CHECK1: omp_offload.failed:
286 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR2:[0-9]+]]
287 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
288 // CHECK1: omp_offload.cont:
289 // CHECK1-NEXT: [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
290 // CHECK1-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
291 // CHECK1-NEXT: [[TMP36:%.*]] = bitcast i8** [[TMP35]] to %struct.S1**
292 // CHECK1-NEXT: store %struct.S1* [[THIS1]], %struct.S1** [[TMP36]], align 8
293 // CHECK1-NEXT: [[TMP37:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
294 // CHECK1-NEXT: [[TMP38:%.*]] = bitcast i8** [[TMP37]] to double**
295 // CHECK1-NEXT: store double* [[A3]], double** [[TMP38]], align 8
296 // CHECK1-NEXT: [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
297 // CHECK1-NEXT: store i8* null, i8** [[TMP39]], align 8
298 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
299 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
300 // CHECK1-NEXT: [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
301 // CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
302 // CHECK1-NEXT: store i32 1, i32* [[TMP42]], align 4
303 // CHECK1-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
304 // CHECK1-NEXT: store i32 1, i32* [[TMP43]], align 4
305 // CHECK1-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
306 // CHECK1-NEXT: store i8** [[TMP40]], i8*** [[TMP44]], align 8
307 // CHECK1-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
308 // CHECK1-NEXT: store i8** [[TMP41]], i8*** [[TMP45]], align 8
309 // CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
310 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP46]], align 8
311 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
312 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP47]], align 8
313 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
314 // CHECK1-NEXT: store i8** null, i8*** [[TMP48]], align 8
315 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
316 // CHECK1-NEXT: store i8** null, i8*** [[TMP49]], align 8
317 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
318 // CHECK1-NEXT: store i64 0, i64* [[TMP50]], align 8
319 // CHECK1-NEXT: [[TMP51:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 1024, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
320 // CHECK1-NEXT: [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0
321 // CHECK1-NEXT: br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
322 // CHECK1: omp_offload.failed8:
323 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR2]]
324 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT9]]
325 // CHECK1: omp_offload.cont9:
326 // CHECK1-NEXT: [[A10:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
327 // CHECK1-NEXT: [[TMP53:%.*]] = load double, double* [[A10]], align 8
328 // CHECK1-NEXT: [[CONV11:%.*]] = fptosi double [[TMP53]] to i32
329 // CHECK1-NEXT: ret i32 [[CONV11]]
330 //
331 //
332 // CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici
333 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
334 // CHECK1-NEXT: entry:
335 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
336 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
337 // CHECK1-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
338 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
339 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
340 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
341 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
342 // CHECK1-NEXT: [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8
343 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
344 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
345 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
346 // CHECK1-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
347 // CHECK1-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
348 // CHECK1-NEXT: store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
349 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
350 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
351 // CHECK1-NEXT: store i32 [[TMP1]], i32* [[CONV]], align 4
352 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
353 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
354 // CHECK1-NEXT: [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i64*
355 // CHECK1-NEXT: store i64 [[TMP2]], i64* [[TMP4]], align 8
356 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
357 // CHECK1-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
358 // CHECK1-NEXT: store i64 [[TMP2]], i64* [[TMP6]], align 8
359 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
360 // CHECK1-NEXT: store i8* null, i8** [[TMP7]], align 8
361 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
362 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
363 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
364 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
365 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
366 // CHECK1-NEXT: store i32 1, i32* [[TMP11]], align 4
367 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
368 // CHECK1-NEXT: store i32 1, i32* [[TMP12]], align 4
369 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
370 // CHECK1-NEXT: store i8** [[TMP8]], i8*** [[TMP13]], align 8
371 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
372 // CHECK1-NEXT: store i8** [[TMP9]], i8*** [[TMP14]], align 8
373 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
374 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64** [[TMP15]], align 8
375 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
376 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i64** [[TMP16]], align 8
377 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
378 // CHECK1-NEXT: store i8** null, i8*** [[TMP17]], align 8
379 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
380 // CHECK1-NEXT: store i8** null, i8*** [[TMP18]], align 8
381 // CHECK1-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
382 // CHECK1-NEXT: store i64 0, i64* [[TMP19]], align 8
383 // CHECK1-NEXT: [[TMP20:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 [[TMP10]], i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
384 // CHECK1-NEXT: [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
385 // CHECK1-NEXT: br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
386 // CHECK1: omp_offload.failed:
387 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i64 [[TMP2]]) #[[ATTR2]]
388 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
389 // CHECK1: omp_offload.cont:
390 // CHECK1-NEXT: [[TMP22:%.*]] = load i32, i32* [[N_ADDR]], align 4
391 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 32, [[TMP22]]
392 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
393 // CHECK1-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
394 // CHECK1-NEXT: [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED2]] to i32*
395 // CHECK1-NEXT: store i32 [[TMP23]], i32* [[CONV3]], align 4
396 // CHECK1-NEXT: [[TMP24:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED2]], align 8
397 // CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
398 // CHECK1-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
399 // CHECK1-NEXT: store i64 [[TMP24]], i64* [[TMP26]], align 8
400 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
401 // CHECK1-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
402 // CHECK1-NEXT: store i64 [[TMP24]], i64* [[TMP28]], align 8
403 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
404 // CHECK1-NEXT: store i8* null, i8** [[TMP29]], align 8
405 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
406 // CHECK1-NEXT: [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
407 // CHECK1-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
408 // CHECK1-NEXT: [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
409 // CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
410 // CHECK1-NEXT: store i32 1, i32* [[TMP33]], align 4
411 // CHECK1-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
412 // CHECK1-NEXT: store i32 1, i32* [[TMP34]], align 4
413 // CHECK1-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
414 // CHECK1-NEXT: store i8** [[TMP30]], i8*** [[TMP35]], align 8
415 // CHECK1-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
416 // CHECK1-NEXT: store i8** [[TMP31]], i8*** [[TMP36]], align 8
417 // CHECK1-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
418 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP37]], align 8
419 // CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
420 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP38]], align 8
421 // CHECK1-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
422 // CHECK1-NEXT: store i8** null, i8*** [[TMP39]], align 8
423 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
424 // CHECK1-NEXT: store i8** null, i8*** [[TMP40]], align 8
425 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
426 // CHECK1-NEXT: store i64 0, i64* [[TMP41]], align 8
427 // CHECK1-NEXT: [[TMP42:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 [[TMP32]], i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
428 // CHECK1-NEXT: [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
429 // CHECK1-NEXT: br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
430 // CHECK1: omp_offload.failed8:
431 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i64 [[TMP24]]) #[[ATTR2]]
432 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT9]]
433 // CHECK1: omp_offload.cont9:
434 // CHECK1-NEXT: [[TMP44:%.*]] = load i32, i32* [[N_ADDR]], align 4
435 // CHECK1-NEXT: [[ADD10:%.*]] = add nsw i32 [[TMP44]], 1
436 // CHECK1-NEXT: ret i32 [[ADD10]]
437 //
438 //
439 // CHECK1-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
440 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
441 // CHECK1-NEXT: entry:
442 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
443 // CHECK1-NEXT: [[A:%.*]] = alloca i32, align 4
444 // CHECK1-NEXT: [[B:%.*]] = alloca i16, align 2
445 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
446 // CHECK1-NEXT: [[A_CASTED:%.*]] = alloca i64, align 8
447 // CHECK1-NEXT: [[B_CASTED:%.*]] = alloca i64, align 8
448 // CHECK1-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
449 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
450 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
451 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
452 // CHECK1-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
453 // CHECK1-NEXT: store i32 0, i32* [[A]], align 4
454 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
455 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
456 // CHECK1-NEXT: store i32 1, i32* [[TMP0]], align 4
457 // CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
458 // CHECK1-NEXT: store i32 0, i32* [[TMP1]], align 4
459 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
460 // CHECK1-NEXT: store i8** null, i8*** [[TMP2]], align 8
461 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
462 // CHECK1-NEXT: store i8** null, i8*** [[TMP3]], align 8
463 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
464 // CHECK1-NEXT: store i64* null, i64** [[TMP4]], align 8
465 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
466 // CHECK1-NEXT: store i64* null, i64** [[TMP5]], align 8
467 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
468 // CHECK1-NEXT: store i8** null, i8*** [[TMP6]], align 8
469 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
470 // CHECK1-NEXT: store i8** null, i8*** [[TMP7]], align 8
471 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
472 // CHECK1-NEXT: store i64 0, i64* [[TMP8]], align 8
473 // CHECK1-NEXT: [[TMP9:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 20, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
474 // CHECK1-NEXT: [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
475 // CHECK1-NEXT: br i1 [[TMP10]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
476 // CHECK1: omp_offload.failed:
477 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
478 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
479 // CHECK1: omp_offload.cont:
480 // CHECK1-NEXT: store i16 1, i16* [[B]], align 2
481 // CHECK1-NEXT: [[TMP11:%.*]] = load i16, i16* [[B]], align 2
482 // CHECK1-NEXT: store i16 [[TMP11]], i16* [[DOTCAPTURE_EXPR_]], align 2
483 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[A]], align 4
484 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
485 // CHECK1-NEXT: store i32 [[TMP12]], i32* [[CONV]], align 4
486 // CHECK1-NEXT: [[TMP13:%.*]] = load i64, i64* [[A_CASTED]], align 8
487 // CHECK1-NEXT: [[TMP14:%.*]] = load i16, i16* [[B]], align 2
488 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[B_CASTED]] to i16*
489 // CHECK1-NEXT: store i16 [[TMP14]], i16* [[CONV1]], align 2
490 // CHECK1-NEXT: [[TMP15:%.*]] = load i64, i64* [[B_CASTED]], align 8
491 // CHECK1-NEXT: [[TMP16:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
492 // CHECK1-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i16*
493 // CHECK1-NEXT: store i16 [[TMP16]], i16* [[CONV2]], align 2
494 // CHECK1-NEXT: [[TMP17:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
495 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
496 // CHECK1-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
497 // CHECK1-NEXT: store i64 [[TMP13]], i64* [[TMP19]], align 8
498 // CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
499 // CHECK1-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
500 // CHECK1-NEXT: store i64 [[TMP13]], i64* [[TMP21]], align 8
501 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
502 // CHECK1-NEXT: store i8* null, i8** [[TMP22]], align 8
503 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
504 // CHECK1-NEXT: [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
505 // CHECK1-NEXT: store i64 [[TMP15]], i64* [[TMP24]], align 8
506 // CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
507 // CHECK1-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
508 // CHECK1-NEXT: store i64 [[TMP15]], i64* [[TMP26]], align 8
509 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
510 // CHECK1-NEXT: store i8* null, i8** [[TMP27]], align 8
511 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
512 // CHECK1-NEXT: [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
513 // CHECK1-NEXT: store i64 [[TMP17]], i64* [[TMP29]], align 8
514 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
515 // CHECK1-NEXT: [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
516 // CHECK1-NEXT: store i64 [[TMP17]], i64* [[TMP31]], align 8
517 // CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
518 // CHECK1-NEXT: store i8* null, i8** [[TMP32]], align 8
519 // CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
520 // CHECK1-NEXT: [[TMP34:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
521 // CHECK1-NEXT: [[TMP35:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
522 // CHECK1-NEXT: [[TMP36:%.*]] = sext i16 [[TMP35]] to i32
523 // CHECK1-NEXT: [[KERNEL_ARGS3:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
524 // CHECK1-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 0
525 // CHECK1-NEXT: store i32 1, i32* [[TMP37]], align 4
526 // CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 1
527 // CHECK1-NEXT: store i32 3, i32* [[TMP38]], align 4
528 // CHECK1-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 2
529 // CHECK1-NEXT: store i8** [[TMP33]], i8*** [[TMP39]], align 8
530 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 3
531 // CHECK1-NEXT: store i8** [[TMP34]], i8*** [[TMP40]], align 8
532 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 4
533 // CHECK1-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP41]], align 8
534 // CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 5
535 // CHECK1-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP42]], align 8
536 // CHECK1-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 6
537 // CHECK1-NEXT: store i8** null, i8*** [[TMP43]], align 8
538 // CHECK1-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 7
539 // CHECK1-NEXT: store i8** null, i8*** [[TMP44]], align 8
540 // CHECK1-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 8
541 // CHECK1-NEXT: store i64 0, i64* [[TMP45]], align 8
542 // CHECK1-NEXT: [[TMP46:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 [[TMP36]], i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]])
543 // CHECK1-NEXT: [[TMP47:%.*]] = icmp ne i32 [[TMP46]], 0
544 // CHECK1-NEXT: br i1 [[TMP47]], label [[OMP_OFFLOAD_FAILED4:%.*]], label [[OMP_OFFLOAD_CONT5:%.*]]
545 // CHECK1: omp_offload.failed4:
546 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]]) #[[ATTR2]]
547 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT5]]
548 // CHECK1: omp_offload.cont5:
549 // CHECK1-NEXT: [[TMP48:%.*]] = load i32, i32* [[A]], align 4
550 // CHECK1-NEXT: ret i32 [[TMP48]]
551 //
552 //
553 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
554 // CHECK1-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
555 // CHECK1-NEXT: entry:
556 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
557 // CHECK1-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8
558 // CHECK1-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
559 // CHECK1-NEXT: [[B_CASTED:%.*]] = alloca i64, align 8
560 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
561 // CHECK1-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
562 // CHECK1-NEXT: store i64 [[B]], i64* [[B_ADDR]], align 8
563 // CHECK1-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
564 // CHECK1-NEXT: [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
565 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
566 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
567 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 4
568 // CHECK1-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 0)
569 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
570 // CHECK1-NEXT: [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
571 // CHECK1-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
572 // CHECK1-NEXT: [[TMP4:%.*]] = load i64, i64* [[B_CASTED]], align 8
573 // 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*, %struct.S1*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP4]])
574 // CHECK1-NEXT: ret void
575 //
576 //
577 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
578 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]]) #[[ATTR1]] {
579 // CHECK1-NEXT: entry:
580 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
581 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
582 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
583 // CHECK1-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8
584 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
585 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
586 // CHECK1-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
587 // CHECK1-NEXT: store i64 [[B]], i64* [[B_ADDR]], align 8
588 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
589 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
590 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
591 // CHECK1-NEXT: [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
592 // CHECK1-NEXT: [[ADD:%.*]] = fadd double [[CONV1]], 1.500000e+00
593 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
594 // CHECK1-NEXT: store double [[ADD]], double* [[A]], align 8
595 // CHECK1-NEXT: ret void
596 //
597 //
598 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
599 // CHECK1-SAME: (%struct.S1* noundef [[THIS:%.*]]) #[[ATTR3:[0-9]+]] {
600 // CHECK1-NEXT: entry:
601 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
602 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
603 // CHECK1-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
604 // CHECK1-NEXT: [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
605 // CHECK1-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024, i32 0)
606 // 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*, %struct.S1*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
607 // CHECK1-NEXT: ret void
608 //
609 //
610 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
611 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]]) #[[ATTR1]] {
612 // CHECK1-NEXT: entry:
613 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
614 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
615 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
616 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
617 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
618 // CHECK1-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
619 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
620 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
621 // CHECK1-NEXT: store double 2.500000e+00, double* [[A]], align 8
622 // CHECK1-NEXT: ret void
623 //
624 //
625 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
626 // CHECK1-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
627 // CHECK1-NEXT: entry:
628 // CHECK1-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
629 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
630 // CHECK1-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
631 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
632 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
633 // CHECK1-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 0)
634 // 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..4 to void (i32*, i32*, ...)*))
635 // CHECK1-NEXT: ret void
636 //
637 //
638 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
639 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
640 // CHECK1-NEXT: entry:
641 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
642 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
643 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
644 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
645 // CHECK1-NEXT: ret void
646 //
647 //
648 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
649 // CHECK1-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
650 // CHECK1-NEXT: entry:
651 // CHECK1-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
652 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
653 // CHECK1-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
654 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
655 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
656 // CHECK1-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 0)
657 // 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..7 to void (i32*, i32*, ...)*))
658 // CHECK1-NEXT: ret void
659 //
660 //
661 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
662 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
663 // CHECK1-NEXT: entry:
664 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
665 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
666 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
667 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
668 // CHECK1-NEXT: ret void
669 //
670 //
671 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
672 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
673 // CHECK1-NEXT: entry:
674 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
675 // CHECK1-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20, i32 0)
676 // 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..10 to void (i32*, i32*, ...)*))
677 // CHECK1-NEXT: ret void
678 //
679 //
680 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..10
681 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
682 // CHECK1-NEXT: entry:
683 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
684 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
685 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
686 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
687 // CHECK1-NEXT: ret void
688 //
689 //
690 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
691 // CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
692 // CHECK1-NEXT: entry:
693 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca i64, align 8
694 // CHECK1-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8
695 // CHECK1-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
696 // CHECK1-NEXT: [[A_CASTED:%.*]] = alloca i64, align 8
697 // CHECK1-NEXT: [[B_CASTED:%.*]] = alloca i64, align 8
698 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
699 // CHECK1-NEXT: store i64 [[A]], i64* [[A_ADDR]], align 8
700 // CHECK1-NEXT: store i64 [[B]], i64* [[B_ADDR]], align 8
701 // CHECK1-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
702 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
703 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
704 // CHECK1-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
705 // CHECK1-NEXT: [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 2
706 // CHECK1-NEXT: [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
707 // CHECK1-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 0)
708 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
709 // CHECK1-NEXT: [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
710 // CHECK1-NEXT: store i32 [[TMP3]], i32* [[CONV3]], align 4
711 // CHECK1-NEXT: [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
712 // CHECK1-NEXT: [[TMP5:%.*]] = load i16, i16* [[CONV1]], align 2
713 // CHECK1-NEXT: [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i16*
714 // CHECK1-NEXT: store i16 [[TMP5]], i16* [[CONV4]], align 2
715 // CHECK1-NEXT: [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
716 // 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..11 to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]])
717 // CHECK1-NEXT: ret void
718 //
719 //
720 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
721 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR1]] {
722 // CHECK1-NEXT: entry:
723 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
724 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
725 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca i64, align 8
726 // CHECK1-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8
727 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
728 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
729 // CHECK1-NEXT: store i64 [[A]], i64* [[A_ADDR]], align 8
730 // CHECK1-NEXT: store i64 [[B]], i64* [[B_ADDR]], align 8
731 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
732 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
733 // CHECK1-NEXT: [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 2
734 // CHECK1-NEXT: [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
735 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
736 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
737 // CHECK1-NEXT: store i32 [[ADD]], i32* [[CONV]], align 4
738 // CHECK1-NEXT: ret void
739 //
740 //
741 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
742 // CHECK1-SAME: () #[[ATTR5:[0-9]+]] {
743 // CHECK1-NEXT: entry:
744 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
745 // CHECK1-NEXT: ret void
746 //
747 //
748 // CHECK3-LABEL: define {{[^@]+}}@_Z3bari
749 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
750 // CHECK3-NEXT: entry:
751 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
752 // CHECK3-NEXT: [[A:%.*]] = alloca i32, align 4
753 // CHECK3-NEXT: [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
754 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
755 // CHECK3-NEXT: store i32 0, i32* [[A]], align 4
756 // CHECK3-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
757 // CHECK3-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP0]])
758 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[A]], align 4
759 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
760 // CHECK3-NEXT: store i32 [[ADD]], i32* [[A]], align 4
761 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
762 // CHECK3-NEXT: [[CALL1:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP2]])
763 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[A]], align 4
764 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
765 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[A]], align 4
766 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
767 // CHECK3-NEXT: [[CALL3:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP4]])
768 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[A]], align 4
769 // CHECK3-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
770 // CHECK3-NEXT: store i32 [[ADD4]], i32* [[A]], align 4
771 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[A]], align 4
772 // CHECK3-NEXT: ret i32 [[TMP6]]
773 //
774 //
775 // CHECK3-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
776 // CHECK3-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
777 // CHECK3-NEXT: entry:
778 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
779 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
780 // CHECK3-NEXT: [[B:%.*]] = alloca i32, align 4
781 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
782 // CHECK3-NEXT: [[B_CASTED:%.*]] = alloca i32, align 4
783 // CHECK3-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
784 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
785 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
786 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
787 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
788 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
789 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
790 // CHECK3-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
791 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
792 // CHECK3-NEXT: [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
793 // CHECK3-NEXT: store i32 1, i32* [[B]], align 4
794 // CHECK3-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
795 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[B]], align 4
796 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
797 // CHECK3-NEXT: store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
798 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[B]], align 4
799 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[B_CASTED]], align 4
800 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
801 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
802 // CHECK3-NEXT: store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
803 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
804 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
805 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
806 // CHECK3-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
807 // CHECK3-NEXT: store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 4
808 // CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
809 // CHECK3-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
810 // CHECK3-NEXT: store double* [[A]], double** [[TMP9]], align 4
811 // CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
812 // CHECK3-NEXT: store i8* null, i8** [[TMP10]], align 4
813 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
814 // CHECK3-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
815 // CHECK3-NEXT: store i32 [[TMP3]], i32* [[TMP12]], align 4
816 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
817 // CHECK3-NEXT: [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
818 // CHECK3-NEXT: store i32 [[TMP3]], i32* [[TMP14]], align 4
819 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
820 // CHECK3-NEXT: store i8* null, i8** [[TMP15]], align 4
821 // CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
822 // CHECK3-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
823 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[TMP17]], align 4
824 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
825 // CHECK3-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
826 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[TMP19]], align 4
827 // CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
828 // CHECK3-NEXT: store i8* null, i8** [[TMP20]], align 4
829 // CHECK3-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
830 // CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
831 // CHECK3-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
832 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
833 // CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
834 // CHECK3-NEXT: store i32 1, i32* [[TMP24]], align 4
835 // CHECK3-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
836 // CHECK3-NEXT: store i32 3, i32* [[TMP25]], align 4
837 // CHECK3-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
838 // CHECK3-NEXT: store i8** [[TMP21]], i8*** [[TMP26]], align 4
839 // CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
840 // CHECK3-NEXT: store i8** [[TMP22]], i8*** [[TMP27]], align 4
841 // CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
842 // CHECK3-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP28]], align 4
843 // CHECK3-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
844 // CHECK3-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP29]], align 4
845 // CHECK3-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
846 // CHECK3-NEXT: store i8** null, i8*** [[TMP30]], align 4
847 // CHECK3-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
848 // CHECK3-NEXT: store i8** null, i8*** [[TMP31]], align 4
849 // CHECK3-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
850 // CHECK3-NEXT: store i64 0, i64* [[TMP32]], align 8
851 // CHECK3-NEXT: [[TMP33:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i32 [[TMP23]], i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
852 // CHECK3-NEXT: [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
853 // CHECK3-NEXT: br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
854 // CHECK3: omp_offload.failed:
855 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR2:[0-9]+]]
856 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
857 // CHECK3: omp_offload.cont:
858 // CHECK3-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
859 // CHECK3-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
860 // CHECK3-NEXT: [[TMP36:%.*]] = bitcast i8** [[TMP35]] to %struct.S1**
861 // CHECK3-NEXT: store %struct.S1* [[THIS1]], %struct.S1** [[TMP36]], align 4
862 // CHECK3-NEXT: [[TMP37:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
863 // CHECK3-NEXT: [[TMP38:%.*]] = bitcast i8** [[TMP37]] to double**
864 // CHECK3-NEXT: store double* [[A2]], double** [[TMP38]], align 4
865 // CHECK3-NEXT: [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
866 // CHECK3-NEXT: store i8* null, i8** [[TMP39]], align 4
867 // CHECK3-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
868 // CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
869 // CHECK3-NEXT: [[KERNEL_ARGS6:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
870 // CHECK3-NEXT: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 0
871 // CHECK3-NEXT: store i32 1, i32* [[TMP42]], align 4
872 // CHECK3-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 1
873 // CHECK3-NEXT: store i32 1, i32* [[TMP43]], align 4
874 // CHECK3-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 2
875 // CHECK3-NEXT: store i8** [[TMP40]], i8*** [[TMP44]], align 4
876 // CHECK3-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 3
877 // CHECK3-NEXT: store i8** [[TMP41]], i8*** [[TMP45]], align 4
878 // CHECK3-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 4
879 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP46]], align 4
880 // CHECK3-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 5
881 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP47]], align 4
882 // CHECK3-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 6
883 // CHECK3-NEXT: store i8** null, i8*** [[TMP48]], align 4
884 // CHECK3-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 7
885 // CHECK3-NEXT: store i8** null, i8*** [[TMP49]], align 4
886 // CHECK3-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 8
887 // CHECK3-NEXT: store i64 0, i64* [[TMP50]], align 8
888 // CHECK3-NEXT: [[TMP51:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 1024, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]])
889 // CHECK3-NEXT: [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0
890 // CHECK3-NEXT: br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
891 // CHECK3: omp_offload.failed7:
892 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR2]]
893 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT8]]
894 // CHECK3: omp_offload.cont8:
895 // CHECK3-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
896 // CHECK3-NEXT: [[TMP53:%.*]] = load double, double* [[A9]], align 4
897 // CHECK3-NEXT: [[CONV:%.*]] = fptosi double [[TMP53]] to i32
898 // CHECK3-NEXT: ret i32 [[CONV]]
899 //
900 //
901 // CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici
902 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
903 // CHECK3-NEXT: entry:
904 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
905 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
906 // CHECK3-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
907 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
908 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
909 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
910 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
911 // CHECK3-NEXT: [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
912 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
913 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
914 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
915 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
916 // CHECK3-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
917 // CHECK3-NEXT: store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
918 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
919 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
920 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
921 // CHECK3-NEXT: [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
922 // CHECK3-NEXT: [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
923 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[TMP4]], align 4
924 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
925 // CHECK3-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
926 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[TMP6]], align 4
927 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
928 // CHECK3-NEXT: store i8* null, i8** [[TMP7]], align 4
929 // CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
930 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
931 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
932 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
933 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
934 // CHECK3-NEXT: store i32 1, i32* [[TMP11]], align 4
935 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
936 // CHECK3-NEXT: store i32 1, i32* [[TMP12]], align 4
937 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
938 // CHECK3-NEXT: store i8** [[TMP8]], i8*** [[TMP13]], align 4
939 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
940 // CHECK3-NEXT: store i8** [[TMP9]], i8*** [[TMP14]], align 4
941 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
942 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64** [[TMP15]], align 4
943 // CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
944 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i64** [[TMP16]], align 4
945 // CHECK3-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
946 // CHECK3-NEXT: store i8** null, i8*** [[TMP17]], align 4
947 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
948 // CHECK3-NEXT: store i8** null, i8*** [[TMP18]], align 4
949 // CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
950 // CHECK3-NEXT: store i64 0, i64* [[TMP19]], align 8
951 // CHECK3-NEXT: [[TMP20:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 [[TMP10]], i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
952 // CHECK3-NEXT: [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
953 // CHECK3-NEXT: br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
954 // CHECK3: omp_offload.failed:
955 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i32 [[TMP2]]) #[[ATTR2]]
956 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
957 // CHECK3: omp_offload.cont:
958 // CHECK3-NEXT: [[TMP22:%.*]] = load i32, i32* [[N_ADDR]], align 4
959 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 32, [[TMP22]]
960 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
961 // CHECK3-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
962 // CHECK3-NEXT: store i32 [[TMP23]], i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
963 // CHECK3-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
964 // CHECK3-NEXT: [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
965 // CHECK3-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
966 // CHECK3-NEXT: store i32 [[TMP24]], i32* [[TMP26]], align 4
967 // CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
968 // CHECK3-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
969 // CHECK3-NEXT: store i32 [[TMP24]], i32* [[TMP28]], align 4
970 // CHECK3-NEXT: [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
971 // CHECK3-NEXT: store i8* null, i8** [[TMP29]], align 4
972 // CHECK3-NEXT: [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
973 // CHECK3-NEXT: [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
974 // CHECK3-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
975 // CHECK3-NEXT: [[KERNEL_ARGS6:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
976 // CHECK3-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 0
977 // CHECK3-NEXT: store i32 1, i32* [[TMP33]], align 4
978 // CHECK3-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 1
979 // CHECK3-NEXT: store i32 1, i32* [[TMP34]], align 4
980 // CHECK3-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 2
981 // CHECK3-NEXT: store i8** [[TMP30]], i8*** [[TMP35]], align 4
982 // CHECK3-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 3
983 // CHECK3-NEXT: store i8** [[TMP31]], i8*** [[TMP36]], align 4
984 // CHECK3-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 4
985 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP37]], align 4
986 // CHECK3-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 5
987 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP38]], align 4
988 // CHECK3-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 6
989 // CHECK3-NEXT: store i8** null, i8*** [[TMP39]], align 4
990 // CHECK3-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 7
991 // CHECK3-NEXT: store i8** null, i8*** [[TMP40]], align 4
992 // CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 8
993 // CHECK3-NEXT: store i64 0, i64* [[TMP41]], align 8
994 // CHECK3-NEXT: [[TMP42:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 [[TMP32]], i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]])
995 // CHECK3-NEXT: [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
996 // CHECK3-NEXT: br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
997 // CHECK3: omp_offload.failed7:
998 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i32 [[TMP24]]) #[[ATTR2]]
999 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT8]]
1000 // CHECK3: omp_offload.cont8:
1001 // CHECK3-NEXT: [[TMP44:%.*]] = load i32, i32* [[N_ADDR]], align 4
1002 // CHECK3-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP44]], 1
1003 // CHECK3-NEXT: ret i32 [[ADD9]]
1004 //
1005 //
1006 // CHECK3-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
1007 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
1008 // CHECK3-NEXT: entry:
1009 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
1010 // CHECK3-NEXT: [[A:%.*]] = alloca i32, align 4
1011 // CHECK3-NEXT: [[B:%.*]] = alloca i16, align 2
1012 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
1013 // CHECK3-NEXT: [[A_CASTED:%.*]] = alloca i32, align 4
1014 // CHECK3-NEXT: [[B_CASTED:%.*]] = alloca i32, align 4
1015 // CHECK3-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
1016 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
1017 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
1018 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
1019 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
1020 // CHECK3-NEXT: store i32 0, i32* [[A]], align 4
1021 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1022 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1023 // CHECK3-NEXT: store i32 1, i32* [[TMP0]], align 4
1024 // CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1025 // CHECK3-NEXT: store i32 0, i32* [[TMP1]], align 4
1026 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1027 // CHECK3-NEXT: store i8** null, i8*** [[TMP2]], align 4
1028 // CHECK3-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1029 // CHECK3-NEXT: store i8** null, i8*** [[TMP3]], align 4
1030 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1031 // CHECK3-NEXT: store i64* null, i64** [[TMP4]], align 4
1032 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1033 // CHECK3-NEXT: store i64* null, i64** [[TMP5]], align 4
1034 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1035 // CHECK3-NEXT: store i8** null, i8*** [[TMP6]], align 4
1036 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1037 // CHECK3-NEXT: store i8** null, i8*** [[TMP7]], align 4
1038 // CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1039 // CHECK3-NEXT: store i64 0, i64* [[TMP8]], align 8
1040 // CHECK3-NEXT: [[TMP9:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 20, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1041 // CHECK3-NEXT: [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
1042 // CHECK3-NEXT: br i1 [[TMP10]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1043 // CHECK3: omp_offload.failed:
1044 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
1045 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
1046 // CHECK3: omp_offload.cont:
1047 // CHECK3-NEXT: store i16 1, i16* [[B]], align 2
1048 // CHECK3-NEXT: [[TMP11:%.*]] = load i16, i16* [[B]], align 2
1049 // CHECK3-NEXT: store i16 [[TMP11]], i16* [[DOTCAPTURE_EXPR_]], align 2
1050 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[A]], align 4
1051 // CHECK3-NEXT: store i32 [[TMP12]], i32* [[A_CASTED]], align 4
1052 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[A_CASTED]], align 4
1053 // CHECK3-NEXT: [[TMP14:%.*]] = load i16, i16* [[B]], align 2
1054 // CHECK3-NEXT: [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
1055 // CHECK3-NEXT: store i16 [[TMP14]], i16* [[CONV]], align 2
1056 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[B_CASTED]], align 4
1057 // CHECK3-NEXT: [[TMP16:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
1058 // CHECK3-NEXT: [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i16*
1059 // CHECK3-NEXT: store i16 [[TMP16]], i16* [[CONV1]], align 2
1060 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
1061 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1062 // CHECK3-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
1063 // CHECK3-NEXT: store i32 [[TMP13]], i32* [[TMP19]], align 4
1064 // CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1065 // CHECK3-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
1066 // CHECK3-NEXT: store i32 [[TMP13]], i32* [[TMP21]], align 4
1067 // CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1068 // CHECK3-NEXT: store i8* null, i8** [[TMP22]], align 4
1069 // CHECK3-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1070 // CHECK3-NEXT: [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
1071 // CHECK3-NEXT: store i32 [[TMP15]], i32* [[TMP24]], align 4
1072 // CHECK3-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1073 // CHECK3-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
1074 // CHECK3-NEXT: store i32 [[TMP15]], i32* [[TMP26]], align 4
1075 // CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1076 // CHECK3-NEXT: store i8* null, i8** [[TMP27]], align 4
1077 // CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1078 // CHECK3-NEXT: [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
1079 // CHECK3-NEXT: store i32 [[TMP17]], i32* [[TMP29]], align 4
1080 // CHECK3-NEXT: [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1081 // CHECK3-NEXT: [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i32*
1082 // CHECK3-NEXT: store i32 [[TMP17]], i32* [[TMP31]], align 4
1083 // CHECK3-NEXT: [[TMP32:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1084 // CHECK3-NEXT: store i8* null, i8** [[TMP32]], align 4
1085 // CHECK3-NEXT: [[TMP33:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1086 // CHECK3-NEXT: [[TMP34:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1087 // CHECK3-NEXT: [[TMP35:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
1088 // CHECK3-NEXT: [[TMP36:%.*]] = sext i16 [[TMP35]] to i32
1089 // CHECK3-NEXT: [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1090 // CHECK3-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 0
1091 // CHECK3-NEXT: store i32 1, i32* [[TMP37]], align 4
1092 // CHECK3-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 1
1093 // CHECK3-NEXT: store i32 3, i32* [[TMP38]], align 4
1094 // CHECK3-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 2
1095 // CHECK3-NEXT: store i8** [[TMP33]], i8*** [[TMP39]], align 4
1096 // CHECK3-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 3
1097 // CHECK3-NEXT: store i8** [[TMP34]], i8*** [[TMP40]], align 4
1098 // CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 4
1099 // CHECK3-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP41]], align 4
1100 // CHECK3-NEXT: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 5
1101 // CHECK3-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP42]], align 4
1102 // CHECK3-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 6
1103 // CHECK3-NEXT: store i8** null, i8*** [[TMP43]], align 4
1104 // CHECK3-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 7
1105 // CHECK3-NEXT: store i8** null, i8*** [[TMP44]], align 4
1106 // CHECK3-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 8
1107 // CHECK3-NEXT: store i64 0, i64* [[TMP45]], align 8
1108 // CHECK3-NEXT: [[TMP46:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 [[TMP36]], i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]])
1109 // CHECK3-NEXT: [[TMP47:%.*]] = icmp ne i32 [[TMP46]], 0
1110 // CHECK3-NEXT: br i1 [[TMP47]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
1111 // CHECK3: omp_offload.failed3:
1112 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i32 [[TMP13]], i32 [[TMP15]], i32 [[TMP17]]) #[[ATTR2]]
1113 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT4]]
1114 // CHECK3: omp_offload.cont4:
1115 // CHECK3-NEXT: [[TMP48:%.*]] = load i32, i32* [[A]], align 4
1116 // CHECK3-NEXT: ret i32 [[TMP48]]
1117 //
1118 //
1119 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
1120 // CHECK3-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
1121 // CHECK3-NEXT: entry:
1122 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1123 // CHECK3-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4
1124 // CHECK3-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1125 // CHECK3-NEXT: [[B_CASTED:%.*]] = alloca i32, align 4
1126 // CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1127 // CHECK3-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1128 // CHECK3-NEXT: store i32 [[B]], i32* [[B_ADDR]], align 4
1129 // CHECK3-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1130 // CHECK3-NEXT: [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1131 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1132 // CHECK3-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 0)
1133 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[B_ADDR]], align 4
1134 // CHECK3-NEXT: store i32 [[TMP3]], i32* [[B_CASTED]], align 4
1135 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[B_CASTED]], align 4
1136 // 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*, %struct.S1*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP4]])
1137 // CHECK3-NEXT: ret void
1138 //
1139 //
1140 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
1141 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]]) #[[ATTR1]] {
1142 // CHECK3-NEXT: entry:
1143 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1144 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1145 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1146 // CHECK3-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4
1147 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1148 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1149 // CHECK3-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1150 // CHECK3-NEXT: store i32 [[B]], i32* [[B_ADDR]], align 4
1151 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1152 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
1153 // CHECK3-NEXT: [[CONV:%.*]] = sitofp i32 [[TMP1]] to double
1154 // CHECK3-NEXT: [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
1155 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1156 // CHECK3-NEXT: store double [[ADD]], double* [[A]], align 4
1157 // CHECK3-NEXT: ret void
1158 //
1159 //
1160 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
1161 // CHECK3-SAME: (%struct.S1* noundef [[THIS:%.*]]) #[[ATTR3:[0-9]+]] {
1162 // CHECK3-NEXT: entry:
1163 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1164 // CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1165 // CHECK3-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1166 // CHECK3-NEXT: [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1167 // CHECK3-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024, i32 0)
1168 // 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*, %struct.S1*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
1169 // CHECK3-NEXT: ret void
1170 //
1171 //
1172 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
1173 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]]) #[[ATTR1]] {
1174 // CHECK3-NEXT: entry:
1175 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1176 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1177 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1178 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1179 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1180 // CHECK3-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1181 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1182 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1183 // CHECK3-NEXT: store double 2.500000e+00, double* [[A]], align 4
1184 // CHECK3-NEXT: ret void
1185 //
1186 //
1187 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
1188 // CHECK3-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1189 // CHECK3-NEXT: entry:
1190 // CHECK3-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1191 // CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1192 // CHECK3-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1193 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1194 // CHECK3-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 0)
1195 // 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..4 to void (i32*, i32*, ...)*))
1196 // CHECK3-NEXT: ret void
1197 //
1198 //
1199 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
1200 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1201 // CHECK3-NEXT: entry:
1202 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1203 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1204 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1205 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1206 // CHECK3-NEXT: ret void
1207 //
1208 //
1209 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
1210 // CHECK3-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1211 // CHECK3-NEXT: entry:
1212 // CHECK3-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1213 // CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1214 // CHECK3-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1215 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1216 // CHECK3-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 0)
1217 // 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..7 to void (i32*, i32*, ...)*))
1218 // CHECK3-NEXT: ret void
1219 //
1220 //
1221 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
1222 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1223 // CHECK3-NEXT: entry:
1224 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1225 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1226 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1227 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1228 // CHECK3-NEXT: ret void
1229 //
1230 //
1231 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
1232 // CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
1233 // CHECK3-NEXT: entry:
1234 // CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1235 // CHECK3-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20, i32 0)
1236 // 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..10 to void (i32*, i32*, ...)*))
1237 // CHECK3-NEXT: ret void
1238 //
1239 //
1240 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..10
1241 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1242 // CHECK3-NEXT: entry:
1243 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1244 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1245 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1246 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1247 // CHECK3-NEXT: ret void
1248 //
1249 //
1250 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
1251 // CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1252 // CHECK3-NEXT: entry:
1253 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4
1254 // CHECK3-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4
1255 // CHECK3-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1256 // CHECK3-NEXT: [[A_CASTED:%.*]] = alloca i32, align 4
1257 // CHECK3-NEXT: [[B_CASTED:%.*]] = alloca i32, align 4
1258 // CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1259 // CHECK3-NEXT: store i32 [[A]], i32* [[A_ADDR]], align 4
1260 // CHECK3-NEXT: store i32 [[B]], i32* [[B_ADDR]], align 4
1261 // CHECK3-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1262 // CHECK3-NEXT: [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1263 // CHECK3-NEXT: [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
1264 // CHECK3-NEXT: [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 2
1265 // CHECK3-NEXT: [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
1266 // CHECK3-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 0)
1267 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
1268 // CHECK3-NEXT: store i32 [[TMP3]], i32* [[A_CASTED]], align 4
1269 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
1270 // CHECK3-NEXT: [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
1271 // CHECK3-NEXT: [[CONV2:%.*]] = bitcast i32* [[B_CASTED]] to i16*
1272 // CHECK3-NEXT: store i16 [[TMP5]], i16* [[CONV2]], align 2
1273 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
1274 // 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..11 to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]])
1275 // CHECK3-NEXT: ret void
1276 //
1277 //
1278 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
1279 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR1]] {
1280 // CHECK3-NEXT: entry:
1281 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1282 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1283 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4
1284 // CHECK3-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4
1285 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1286 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1287 // CHECK3-NEXT: store i32 [[A]], i32* [[A_ADDR]], align 4
1288 // CHECK3-NEXT: store i32 [[B]], i32* [[B_ADDR]], align 4
1289 // CHECK3-NEXT: [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1290 // CHECK3-NEXT: [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
1291 // CHECK3-NEXT: [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
1292 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
1293 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
1294 // CHECK3-NEXT: store i32 [[ADD]], i32* [[A_ADDR]], align 4
1295 // CHECK3-NEXT: ret void
1296 //
1297 //
1298 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1299 // CHECK3-SAME: () #[[ATTR5:[0-9]+]] {
1300 // CHECK3-NEXT: entry:
1301 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
1302 // CHECK3-NEXT: ret void
1303 //
1304 //
1305 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
1306 // CHECK9-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
1307 // CHECK9-NEXT: entry:
1308 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1309 // CHECK9-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
1310 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1311 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1312 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1313 // CHECK9-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 0)
1314 // CHECK9-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*, ...)*))
1315 // CHECK9-NEXT: ret void
1316 //
1317 //
1318 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
1319 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
1320 // CHECK9-NEXT: entry:
1321 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1322 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1323 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1324 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1325 // CHECK9-NEXT: ret void
1326 //
1327 //
1328 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
1329 // CHECK9-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1330 // CHECK9-NEXT: entry:
1331 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1332 // CHECK9-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1333 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1334 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1335 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1336 // CHECK9-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 0)
1337 // CHECK9-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..1 to void (i32*, i32*, ...)*))
1338 // CHECK9-NEXT: ret void
1339 //
1340 //
1341 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
1342 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
1343 // CHECK9-NEXT: entry:
1344 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1345 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1346 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1347 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1348 // CHECK9-NEXT: ret void
1349 //
1350 //
1351 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
1352 // CHECK9-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1353 // CHECK9-NEXT: entry:
1354 // CHECK9-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1355 // CHECK9-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8
1356 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1357 // CHECK9-NEXT: [[B_CASTED:%.*]] = alloca i64, align 8
1358 // CHECK9-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1359 // CHECK9-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1360 // CHECK9-NEXT: store i64 [[B]], i64* [[B_ADDR]], align 8
1361 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1362 // CHECK9-NEXT: [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1363 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1364 // CHECK9-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1365 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 4
1366 // CHECK9-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 0)
1367 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
1368 // CHECK9-NEXT: [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
1369 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
1370 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[B_CASTED]], align 8
1371 // CHECK9-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*, %struct.S1*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP4]])
1372 // CHECK9-NEXT: ret void
1373 //
1374 //
1375 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
1376 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
1377 // CHECK9-NEXT: entry:
1378 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1379 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1380 // CHECK9-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1381 // CHECK9-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8
1382 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1383 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1384 // CHECK9-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1385 // CHECK9-NEXT: store i64 [[B]], i64* [[B_ADDR]], align 8
1386 // CHECK9-NEXT: [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1387 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1388 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1389 // CHECK9-NEXT: [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
1390 // CHECK9-NEXT: [[ADD:%.*]] = fadd double [[CONV1]], 1.500000e+00
1391 // CHECK9-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1392 // CHECK9-NEXT: store double [[ADD]], double* [[A]], align 8
1393 // CHECK9-NEXT: ret void
1394 //
1395 //
1396 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
1397 // CHECK9-SAME: (%struct.S1* noundef [[THIS:%.*]]) #[[ATTR2:[0-9]+]] {
1398 // CHECK9-NEXT: entry:
1399 // CHECK9-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1400 // CHECK9-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1401 // CHECK9-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1402 // CHECK9-NEXT: [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1403 // CHECK9-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024, i32 0)
1404 // CHECK9-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*, %struct.S1*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
1405 // CHECK9-NEXT: ret void
1406 //
1407 //
1408 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
1409 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]]) #[[ATTR0]] {
1410 // CHECK9-NEXT: entry:
1411 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1412 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1413 // CHECK9-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1414 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1415 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1416 // CHECK9-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1417 // CHECK9-NEXT: [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1418 // CHECK9-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1419 // CHECK9-NEXT: store double 2.500000e+00, double* [[A]], align 8
1420 // CHECK9-NEXT: ret void
1421 //
1422 //
1423 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
1424 // CHECK9-SAME: () #[[ATTR3:[0-9]+]] {
1425 // CHECK9-NEXT: entry:
1426 // CHECK9-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1427 // CHECK9-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20, i32 0)
1428 // CHECK9-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..4 to void (i32*, i32*, ...)*))
1429 // CHECK9-NEXT: ret void
1430 //
1431 //
1432 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
1433 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
1434 // CHECK9-NEXT: entry:
1435 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1436 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1437 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1438 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1439 // CHECK9-NEXT: ret void
1440 //
1441 //
1442 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
1443 // CHECK9-SAME: (i64 noundef [[A:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1444 // CHECK9-NEXT: entry:
1445 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i64, align 8
1446 // CHECK9-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8
1447 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1448 // CHECK9-NEXT: [[A_CASTED:%.*]] = alloca i64, align 8
1449 // CHECK9-NEXT: [[B_CASTED:%.*]] = alloca i64, align 8
1450 // CHECK9-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1451 // CHECK9-NEXT: store i64 [[A]], i64* [[A_ADDR]], align 8
1452 // CHECK9-NEXT: store i64 [[B]], i64* [[B_ADDR]], align 8
1453 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1454 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1455 // CHECK9-NEXT: [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
1456 // CHECK9-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
1457 // CHECK9-NEXT: [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 2
1458 // CHECK9-NEXT: [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
1459 // CHECK9-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 0)
1460 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
1461 // CHECK9-NEXT: [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1462 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV3]], align 4
1463 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
1464 // CHECK9-NEXT: [[TMP5:%.*]] = load i16, i16* [[CONV1]], align 2
1465 // CHECK9-NEXT: [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i16*
1466 // CHECK9-NEXT: store i16 [[TMP5]], i16* [[CONV4]], align 2
1467 // CHECK9-NEXT: [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
1468 // CHECK9-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..5 to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]])
1469 // CHECK9-NEXT: ret void
1470 //
1471 //
1472 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
1473 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
1474 // CHECK9-NEXT: entry:
1475 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1476 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1477 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i64, align 8
1478 // CHECK9-NEXT: [[B_ADDR:%.*]] = alloca i64, align 8
1479 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1480 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1481 // CHECK9-NEXT: store i64 [[A]], i64* [[A_ADDR]], align 8
1482 // CHECK9-NEXT: store i64 [[B]], i64* [[B_ADDR]], align 8
1483 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1484 // CHECK9-NEXT: [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
1485 // CHECK9-NEXT: [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 2
1486 // CHECK9-NEXT: [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
1487 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1488 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
1489 // CHECK9-NEXT: store i32 [[ADD]], i32* [[CONV]], align 4
1490 // CHECK9-NEXT: ret void
1491 //
1492 //
1493 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
1494 // CHECK11-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
1495 // CHECK11-NEXT: entry:
1496 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1497 // CHECK11-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
1498 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1499 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1500 // CHECK11-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 0)
1501 // CHECK11-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*, ...)*))
1502 // CHECK11-NEXT: ret void
1503 //
1504 //
1505 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
1506 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
1507 // CHECK11-NEXT: entry:
1508 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1509 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1510 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1511 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1512 // CHECK11-NEXT: ret void
1513 //
1514 //
1515 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
1516 // CHECK11-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1517 // CHECK11-NEXT: entry:
1518 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1519 // CHECK11-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1520 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1521 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1522 // CHECK11-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 0)
1523 // CHECK11-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..1 to void (i32*, i32*, ...)*))
1524 // CHECK11-NEXT: ret void
1525 //
1526 //
1527 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
1528 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
1529 // CHECK11-NEXT: entry:
1530 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1531 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1532 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1533 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1534 // CHECK11-NEXT: ret void
1535 //
1536 //
1537 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
1538 // CHECK11-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1539 // CHECK11-NEXT: entry:
1540 // CHECK11-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1541 // CHECK11-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4
1542 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1543 // CHECK11-NEXT: [[B_CASTED:%.*]] = alloca i32, align 4
1544 // CHECK11-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1545 // CHECK11-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1546 // CHECK11-NEXT: store i32 [[B]], i32* [[B_ADDR]], align 4
1547 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1548 // CHECK11-NEXT: [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1549 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1550 // CHECK11-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 0)
1551 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[B_ADDR]], align 4
1552 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[B_CASTED]], align 4
1553 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[B_CASTED]], align 4
1554 // CHECK11-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*, %struct.S1*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP4]])
1555 // CHECK11-NEXT: ret void
1556 //
1557 //
1558 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
1559 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]]) #[[ATTR0]] {
1560 // CHECK11-NEXT: entry:
1561 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1562 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1563 // CHECK11-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1564 // CHECK11-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4
1565 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1566 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1567 // CHECK11-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1568 // CHECK11-NEXT: store i32 [[B]], i32* [[B_ADDR]], align 4
1569 // CHECK11-NEXT: [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1570 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
1571 // CHECK11-NEXT: [[CONV:%.*]] = sitofp i32 [[TMP1]] to double
1572 // CHECK11-NEXT: [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
1573 // CHECK11-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1574 // CHECK11-NEXT: store double [[ADD]], double* [[A]], align 4
1575 // CHECK11-NEXT: ret void
1576 //
1577 //
1578 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
1579 // CHECK11-SAME: (%struct.S1* noundef [[THIS:%.*]]) #[[ATTR2:[0-9]+]] {
1580 // CHECK11-NEXT: entry:
1581 // CHECK11-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1582 // CHECK11-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1583 // CHECK11-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1584 // CHECK11-NEXT: [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1585 // CHECK11-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024, i32 0)
1586 // CHECK11-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*, %struct.S1*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
1587 // CHECK11-NEXT: ret void
1588 //
1589 //
1590 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
1591 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]]) #[[ATTR0]] {
1592 // CHECK11-NEXT: entry:
1593 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1594 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1595 // CHECK11-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1596 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1597 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1598 // CHECK11-NEXT: store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1599 // CHECK11-NEXT: [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1600 // CHECK11-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1601 // CHECK11-NEXT: store double 2.500000e+00, double* [[A]], align 4
1602 // CHECK11-NEXT: ret void
1603 //
1604 //
1605 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
1606 // CHECK11-SAME: () #[[ATTR3:[0-9]+]] {
1607 // CHECK11-NEXT: entry:
1608 // CHECK11-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1609 // CHECK11-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20, i32 0)
1610 // CHECK11-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..4 to void (i32*, i32*, ...)*))
1611 // CHECK11-NEXT: ret void
1612 //
1613 //
1614 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
1615 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
1616 // CHECK11-NEXT: entry:
1617 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1618 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1619 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1620 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1621 // CHECK11-NEXT: ret void
1622 //
1623 //
1624 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
1625 // CHECK11-SAME: (i32 noundef [[A:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1626 // CHECK11-NEXT: entry:
1627 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4
1628 // CHECK11-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4
1629 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1630 // CHECK11-NEXT: [[A_CASTED:%.*]] = alloca i32, align 4
1631 // CHECK11-NEXT: [[B_CASTED:%.*]] = alloca i32, align 4
1632 // CHECK11-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1633 // CHECK11-NEXT: store i32 [[A]], i32* [[A_ADDR]], align 4
1634 // CHECK11-NEXT: store i32 [[B]], i32* [[B_ADDR]], align 4
1635 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1636 // CHECK11-NEXT: [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1637 // CHECK11-NEXT: [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
1638 // CHECK11-NEXT: [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 2
1639 // CHECK11-NEXT: [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
1640 // CHECK11-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 0)
1641 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
1642 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[A_CASTED]], align 4
1643 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
1644 // CHECK11-NEXT: [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
1645 // CHECK11-NEXT: [[CONV2:%.*]] = bitcast i32* [[B_CASTED]] to i16*
1646 // CHECK11-NEXT: store i16 [[TMP5]], i16* [[CONV2]], align 2
1647 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
1648 // CHECK11-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..5 to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]])
1649 // CHECK11-NEXT: ret void
1650 //
1651 //
1652 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
1653 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR0]] {
1654 // CHECK11-NEXT: entry:
1655 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1656 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1657 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4
1658 // CHECK11-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4
1659 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1660 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1661 // CHECK11-NEXT: store i32 [[A]], i32* [[A_ADDR]], align 4
1662 // CHECK11-NEXT: store i32 [[B]], i32* [[B_ADDR]], align 4
1663 // CHECK11-NEXT: [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1664 // CHECK11-NEXT: [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
1665 // CHECK11-NEXT: [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
1666 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
1667 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
1668 // CHECK11-NEXT: store i32 [[ADD]], i32* [[A_ADDR]], align 4
1669 // CHECK11-NEXT: ret void
1670 //
1671