xref: /llvm-project/clang/test/OpenMP/target_parallel_if_codegen.cpp (revision b52d33e6de554fbabb6e80c661c4272a900daece)
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 // 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
37 // 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
38 // 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
39 // 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
40 // 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
41 // 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
42 
43 // 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}}"
44 // 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
45 // 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}}"
46 // 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}}"
47 // 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
48 // 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}}"
49 
50 // Test target codegen - host bc file has to be created first.
51 // 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
52 // 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
53 // 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
54 // 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
55 // 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
56 // 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
57 // 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
58 // 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
59 
60 // 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
61 // 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}}"
62 // 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
63 // 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}}"
64 // 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
65 // 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}}"
66 // 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
67 // 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}}"
68 
69 // expected-no-diagnostics
70 #ifndef HEADER
71 #define HEADER
72 
73 
74 
75 
76 // We have 6 target regions
77 
78 
79 
80 // Check target registration is registered as a Ctor.
81 
82 
83 template<typename tx>
84 tx ftemplate(int n) {
85   tx a = 0;
86 
87   #pragma omp target parallel if(parallel: 0)
88   {
89     a += 1;
90   }
91 
92   short b = 1;
93   #pragma omp target parallel if(parallel: 1)
94   {
95     a += b;
96   }
97 
98   return a;
99 }
100 
101 static
102 int fstatic(int n) {
103 
104   #pragma omp target parallel if(n>1)
105   {
106   }
107 
108   #pragma omp target parallel if(target: n-2>2)
109   {
110   }
111 
112   return n+1;
113 }
114 
115 struct S1 {
116   double a;
117 
118   int r1(int n){
119     int b = 1;
120 
121     #pragma omp target parallel if(parallel: n>3)
122     {
123       this->a = (double)b + 1.5;
124     }
125 
126     #pragma omp target parallel if(target: n>4) if(parallel: n>5)
127     {
128       this->a = 2.5;
129     }
130 
131     return (int)a;
132   }
133 };
134 
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 // Check that the offloading functions are emitted and that the parallel function
162 // is appropriately guarded.
163 
164 
165 
166 
167 
168 
169 
170 
171 
172 
173 
174 
175 
176 
177 
178 
179 
180 
181 
182 
183 
184 #endif
185 // CHECK1-LABEL: define {{[^@]+}}@_Z3bari
186 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
187 // CHECK1-NEXT:  entry:
188 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
189 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
190 // CHECK1-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
191 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
192 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
193 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
194 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP0]])
195 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
196 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
197 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
198 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
199 // CHECK1-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP2]])
200 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
201 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
202 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
203 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
204 // CHECK1-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP4]])
205 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
206 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
207 // CHECK1-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
208 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
209 // CHECK1-NEXT:    ret i32 [[TMP6]]
210 //
211 //
212 // CHECK1-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
213 // CHECK1-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
214 // CHECK1-NEXT:  entry:
215 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
216 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
217 // CHECK1-NEXT:    [[B:%.*]] = alloca i32, align 4
218 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
219 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
220 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
221 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
222 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
223 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
224 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i8, align 1
225 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED9:%.*]] = alloca i64, align 8
226 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS14:%.*]] = alloca [2 x i8*], align 8
227 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS15:%.*]] = alloca [2 x i8*], align 8
228 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS16:%.*]] = alloca [2 x i8*], align 8
229 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
230 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
231 // CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
232 // CHECK1-NEXT:    store i32 1, i32* [[B]], align 4
233 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
234 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP0]], 3
235 // CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[CMP]] to i8
236 // CHECK1-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
237 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
238 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
239 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV]], align 4
240 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[B_CASTED]], align 8
241 // CHECK1-NEXT:    [[TMP3:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
242 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP3]] to i1
243 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
244 // CHECK1-NEXT:    [[FROMBOOL3:%.*]] = zext i1 [[TOBOOL]] to i8
245 // CHECK1-NEXT:    store i8 [[FROMBOOL3]], i8* [[CONV2]], align 1
246 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
247 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
248 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
249 // CHECK1-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to %struct.S1**
250 // CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP6]], align 8
251 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
252 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to double**
253 // CHECK1-NEXT:    store double* [[A]], double** [[TMP8]], align 8
254 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
255 // CHECK1-NEXT:    store i8* null, i8** [[TMP9]], align 8
256 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
257 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
258 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP11]], align 8
259 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
260 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
261 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP13]], align 8
262 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
263 // CHECK1-NEXT:    store i8* null, i8** [[TMP14]], align 8
264 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
265 // CHECK1-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
266 // CHECK1-NEXT:    store i64 [[TMP4]], i64* [[TMP16]], align 8
267 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
268 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
269 // CHECK1-NEXT:    store i64 [[TMP4]], i64* [[TMP18]], align 8
270 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
271 // CHECK1-NEXT:    store i8* null, i8** [[TMP19]], align 8
272 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
273 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
274 // CHECK1-NEXT:    [[TMP22:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
275 // CHECK1-NEXT:    [[TOBOOL4:%.*]] = trunc i8 [[TMP22]] to i1
276 // CHECK1-NEXT:    [[TMP23:%.*]] = select i1 [[TOBOOL4]], i32 0, i32 1
277 // CHECK1-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
278 // CHECK1-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
279 // CHECK1-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
280 // CHECK1:       omp_offload.failed:
281 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i64 [[TMP2]], i64 [[TMP4]]) #[[ATTR3:[0-9]+]]
282 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
283 // CHECK1:       omp_offload.cont:
284 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
285 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[TMP26]], 5
286 // CHECK1-NEXT:    [[FROMBOOL7:%.*]] = zext i1 [[CMP6]] to i8
287 // CHECK1-NEXT:    store i8 [[FROMBOOL7]], i8* [[DOTCAPTURE_EXPR_5]], align 1
288 // CHECK1-NEXT:    [[TMP27:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_5]], align 1
289 // CHECK1-NEXT:    [[TOBOOL8:%.*]] = trunc i8 [[TMP27]] to i1
290 // CHECK1-NEXT:    [[CONV10:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED9]] to i8*
291 // CHECK1-NEXT:    [[FROMBOOL11:%.*]] = zext i1 [[TOBOOL8]] to i8
292 // CHECK1-NEXT:    store i8 [[FROMBOOL11]], i8* [[CONV10]], align 1
293 // CHECK1-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED9]], align 8
294 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[N_ADDR]], align 4
295 // CHECK1-NEXT:    [[CMP12:%.*]] = icmp sgt i32 [[TMP29]], 4
296 // CHECK1-NEXT:    br i1 [[CMP12]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
297 // CHECK1:       omp_if.then:
298 // CHECK1-NEXT:    [[A13:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
299 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 0
300 // CHECK1-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to %struct.S1**
301 // CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP31]], align 8
302 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 0
303 // CHECK1-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to double**
304 // CHECK1-NEXT:    store double* [[A13]], double** [[TMP33]], align 8
305 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i64 0, i64 0
306 // CHECK1-NEXT:    store i8* null, i8** [[TMP34]], align 8
307 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 1
308 // CHECK1-NEXT:    [[TMP36:%.*]] = bitcast i8** [[TMP35]] to i64*
309 // CHECK1-NEXT:    store i64 [[TMP28]], i64* [[TMP36]], align 8
310 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 1
311 // CHECK1-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i64*
312 // CHECK1-NEXT:    store i64 [[TMP28]], i64* [[TMP38]], align 8
313 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i64 0, i64 1
314 // CHECK1-NEXT:    store i8* null, i8** [[TMP39]], align 8
315 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 0
316 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 0
317 // CHECK1-NEXT:    [[TMP42:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_5]], align 1
318 // CHECK1-NEXT:    [[TOBOOL17:%.*]] = trunc i8 [[TMP42]] to i1
319 // CHECK1-NEXT:    [[TMP43:%.*]] = select i1 [[TOBOOL17]], i32 0, i32 1
320 // CHECK1-NEXT:    [[TMP44:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 2, i8** [[TMP40]], i8** [[TMP41]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP43]])
321 // CHECK1-NEXT:    [[TMP45:%.*]] = icmp ne i32 [[TMP44]], 0
322 // CHECK1-NEXT:    br i1 [[TMP45]], label [[OMP_OFFLOAD_FAILED18:%.*]], label [[OMP_OFFLOAD_CONT19:%.*]]
323 // CHECK1:       omp_offload.failed18:
324 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]], i64 [[TMP28]]) #[[ATTR3]]
325 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT19]]
326 // CHECK1:       omp_offload.cont19:
327 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
328 // CHECK1:       omp_if.else:
329 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]], i64 [[TMP28]]) #[[ATTR3]]
330 // CHECK1-NEXT:    br label [[OMP_IF_END]]
331 // CHECK1:       omp_if.end:
332 // CHECK1-NEXT:    [[A20:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
333 // CHECK1-NEXT:    [[TMP46:%.*]] = load double, double* [[A20]], align 8
334 // CHECK1-NEXT:    [[CONV21:%.*]] = fptosi double [[TMP46]] to i32
335 // CHECK1-NEXT:    ret i32 [[CONV21]]
336 //
337 //
338 // CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici
339 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
340 // CHECK1-NEXT:  entry:
341 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
342 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
343 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
344 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
345 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
346 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
347 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
348 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
349 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP0]], 1
350 // CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[CMP]] to i8
351 // CHECK1-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
352 // CHECK1-NEXT:    [[TMP1:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
353 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP1]] to i1
354 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
355 // CHECK1-NEXT:    [[FROMBOOL1:%.*]] = zext i1 [[TOBOOL]] to i8
356 // CHECK1-NEXT:    store i8 [[FROMBOOL1]], i8* [[CONV]], align 1
357 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
358 // CHECK1-NEXT:    [[TMP3:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
359 // CHECK1-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP3]] to i1
360 // CHECK1-NEXT:    br i1 [[TOBOOL2]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
361 // CHECK1:       omp_if.then:
362 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
363 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
364 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP5]], align 8
365 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
366 // CHECK1-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
367 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP7]], align 8
368 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
369 // CHECK1-NEXT:    store i8* null, i8** [[TMP8]], align 8
370 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
371 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
372 // CHECK1-NEXT:    [[TMP11:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
373 // CHECK1-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP11]] to i1
374 // CHECK1-NEXT:    [[TMP12:%.*]] = select i1 [[TOBOOL3]], i32 0, i32 1
375 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP9]], i8** [[TMP10]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP12]])
376 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
377 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
378 // CHECK1:       omp_offload.failed:
379 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i64 [[TMP2]]) #[[ATTR3]]
380 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
381 // CHECK1:       omp_offload.cont:
382 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
383 // CHECK1:       omp_if.else:
384 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i64 [[TMP2]]) #[[ATTR3]]
385 // CHECK1-NEXT:    br label [[OMP_IF_END]]
386 // CHECK1:       omp_if.end:
387 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[N_ADDR]], align 4
388 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 2
389 // CHECK1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[SUB]], 2
390 // CHECK1-NEXT:    br i1 [[CMP4]], label [[OMP_IF_THEN5:%.*]], label [[OMP_IF_ELSE8:%.*]]
391 // CHECK1:       omp_if.then5:
392 // CHECK1-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0)
393 // CHECK1-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
394 // CHECK1-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
395 // CHECK1:       omp_offload.failed6:
396 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108() #[[ATTR3]]
397 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
398 // CHECK1:       omp_offload.cont7:
399 // CHECK1-NEXT:    br label [[OMP_IF_END9:%.*]]
400 // CHECK1:       omp_if.else8:
401 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108() #[[ATTR3]]
402 // CHECK1-NEXT:    br label [[OMP_IF_END9]]
403 // CHECK1:       omp_if.end9:
404 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[N_ADDR]], align 4
405 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], 1
406 // CHECK1-NEXT:    ret i32 [[ADD]]
407 //
408 //
409 // CHECK1-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
410 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
411 // CHECK1-NEXT:  entry:
412 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
413 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
414 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
415 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
416 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
417 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
418 // CHECK1-NEXT:    [[B:%.*]] = alloca i16, align 2
419 // CHECK1-NEXT:    [[A_CASTED1:%.*]] = alloca i64, align 8
420 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
421 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [2 x i8*], align 8
422 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [2 x i8*], align 8
423 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [2 x i8*], align 8
424 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
425 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
426 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
427 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
428 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
429 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
430 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
431 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i64*
432 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP3]], align 8
433 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
434 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
435 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP5]], align 8
436 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
437 // CHECK1-NEXT:    store i8* null, i8** [[TMP6]], align 8
438 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
439 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
440 // CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l87.region_id, i32 1, i8** [[TMP7]], i8** [[TMP8]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.9, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1)
441 // CHECK1-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
442 // CHECK1-NEXT:    br i1 [[TMP10]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
443 // CHECK1:       omp_offload.failed:
444 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l87(i64 [[TMP1]]) #[[ATTR3]]
445 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
446 // CHECK1:       omp_offload.cont:
447 // CHECK1-NEXT:    store i16 1, i16* [[B]], align 2
448 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A]], align 4
449 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED1]] to i32*
450 // CHECK1-NEXT:    store i32 [[TMP11]], i32* [[CONV2]], align 4
451 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED1]], align 8
452 // CHECK1-NEXT:    [[TMP13:%.*]] = load i16, i16* [[B]], align 2
453 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i16*
454 // CHECK1-NEXT:    store i16 [[TMP13]], i16* [[CONV3]], align 2
455 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[B_CASTED]], align 8
456 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
457 // CHECK1-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
458 // CHECK1-NEXT:    store i64 [[TMP12]], i64* [[TMP16]], align 8
459 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
460 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
461 // CHECK1-NEXT:    store i64 [[TMP12]], i64* [[TMP18]], align 8
462 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
463 // CHECK1-NEXT:    store i8* null, i8** [[TMP19]], align 8
464 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
465 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
466 // CHECK1-NEXT:    store i64 [[TMP14]], i64* [[TMP21]], align 8
467 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
468 // CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
469 // CHECK1-NEXT:    store i64 [[TMP14]], i64* [[TMP23]], align 8
470 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 1
471 // CHECK1-NEXT:    store i8* null, i8** [[TMP24]], align 8
472 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
473 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
474 // CHECK1-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 2, i8** [[TMP25]], i8** [[TMP26]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
475 // CHECK1-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
476 // CHECK1-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
477 // CHECK1:       omp_offload.failed7:
478 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i64 [[TMP12]], i64 [[TMP14]]) #[[ATTR3]]
479 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
480 // CHECK1:       omp_offload.cont8:
481 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[A]], align 4
482 // CHECK1-NEXT:    ret i32 [[TMP29]]
483 //
484 //
485 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
486 // CHECK1-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
487 // CHECK1-NEXT:  entry:
488 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
489 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
490 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
491 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
492 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
493 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
494 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
495 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
496 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
497 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
498 // CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
499 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
500 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
501 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
502 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
503 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[CONV2]], align 4
504 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
505 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8, i8* [[CONV1]], align 1
506 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP4]] to i1
507 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
508 // CHECK1:       omp_if.then:
509 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%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 [[TMP3]])
510 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
511 // CHECK1:       omp_if.else:
512 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
513 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
514 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
515 // CHECK1-NEXT:    call void @.omp_outlined.(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], %struct.S1* [[TMP1]], i64 [[TMP3]]) #[[ATTR3]]
516 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
517 // CHECK1-NEXT:    br label [[OMP_IF_END]]
518 // CHECK1:       omp_if.end:
519 // CHECK1-NEXT:    ret void
520 //
521 //
522 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
523 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]]) #[[ATTR2:[0-9]+]] {
524 // CHECK1-NEXT:  entry:
525 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
526 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
527 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
528 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
529 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
530 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
531 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
532 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
533 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
534 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
535 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
536 // CHECK1-NEXT:    [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
537 // CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[CONV1]], 1.500000e+00
538 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
539 // CHECK1-NEXT:    store double [[ADD]], double* [[A]], align 8
540 // CHECK1-NEXT:    ret void
541 //
542 //
543 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
544 // CHECK1-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
545 // CHECK1-NEXT:  entry:
546 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
547 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
548 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
549 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
550 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
551 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
552 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
553 // CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
554 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
555 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
556 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
557 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
558 // CHECK1:       omp_if.then:
559 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%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]])
560 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
561 // CHECK1:       omp_if.else:
562 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
563 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
564 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
565 // CHECK1-NEXT:    call void @.omp_outlined..1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], %struct.S1* [[TMP1]]) #[[ATTR3]]
566 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
567 // CHECK1-NEXT:    br label [[OMP_IF_END]]
568 // CHECK1:       omp_if.end:
569 // CHECK1-NEXT:    ret void
570 //
571 //
572 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
573 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]]) #[[ATTR2]] {
574 // CHECK1-NEXT:  entry:
575 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
576 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
577 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
578 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
579 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
580 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
581 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
582 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
583 // CHECK1-NEXT:    store double 2.500000e+00, double* [[A]], align 8
584 // CHECK1-NEXT:    ret void
585 //
586 //
587 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
588 // CHECK1-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
589 // CHECK1-NEXT:  entry:
590 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
591 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
592 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
593 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
594 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
595 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
596 // CHECK1-NEXT:    [[TMP1:%.*]] = load i8, i8* [[CONV]], align 1
597 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP1]] to i1
598 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
599 // CHECK1:       omp_if.then:
600 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
601 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
602 // CHECK1:       omp_if.else:
603 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
604 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
605 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
606 // CHECK1-NEXT:    call void @.omp_outlined..4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]]) #[[ATTR3]]
607 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
608 // CHECK1-NEXT:    br label [[OMP_IF_END]]
609 // CHECK1:       omp_if.end:
610 // CHECK1-NEXT:    ret void
611 //
612 //
613 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
614 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
615 // CHECK1-NEXT:  entry:
616 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
617 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
618 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
619 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
620 // CHECK1-NEXT:    ret void
621 //
622 //
623 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
624 // CHECK1-SAME: () #[[ATTR1]] {
625 // CHECK1-NEXT:  entry:
626 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..7 to void (i32*, i32*, ...)*))
627 // CHECK1-NEXT:    ret void
628 //
629 //
630 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
631 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
632 // CHECK1-NEXT:  entry:
633 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
634 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
635 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
636 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
637 // CHECK1-NEXT:    ret void
638 //
639 //
640 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l87
641 // CHECK1-SAME: (i64 noundef [[A:%.*]]) #[[ATTR1]] {
642 // CHECK1-NEXT:  entry:
643 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
644 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
645 // CHECK1-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
646 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
647 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
648 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
649 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
650 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
651 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
652 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV1]], align 4
653 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
654 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
655 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
656 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
657 // CHECK1-NEXT:    call void @.omp_outlined..8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP2]]) #[[ATTR3]]
658 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
659 // CHECK1-NEXT:    ret void
660 //
661 //
662 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..8
663 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR2]] {
664 // CHECK1-NEXT:  entry:
665 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
666 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
667 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
668 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
669 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
670 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
671 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
672 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
673 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
674 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
675 // CHECK1-NEXT:    ret void
676 //
677 //
678 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
679 // CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR1]] {
680 // CHECK1-NEXT:  entry:
681 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
682 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
683 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
684 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
685 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
686 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
687 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
688 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
689 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
690 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
691 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
692 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
693 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
694 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i16*
695 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
696 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
697 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
698 // CHECK1-NEXT:    ret void
699 //
700 //
701 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
702 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR2]] {
703 // CHECK1-NEXT:  entry:
704 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
705 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
706 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
707 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
708 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
709 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
710 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
711 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
712 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
713 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
714 // CHECK1-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 2
715 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
716 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
717 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
718 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
719 // CHECK1-NEXT:    ret void
720 //
721 //
722 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
723 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
724 // CHECK1-NEXT:  entry:
725 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
726 // CHECK1-NEXT:    ret void
727 //
728 //
729 // CHECK3-LABEL: define {{[^@]+}}@_Z3bari
730 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
731 // CHECK3-NEXT:  entry:
732 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
733 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
734 // CHECK3-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
735 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
736 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
737 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
738 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP0]])
739 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
740 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
741 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
742 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
743 // CHECK3-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP2]])
744 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
745 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
746 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
747 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
748 // CHECK3-NEXT:    [[CALL3:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP4]])
749 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
750 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
751 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
752 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
753 // CHECK3-NEXT:    ret i32 [[TMP6]]
754 //
755 //
756 // CHECK3-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
757 // CHECK3-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
758 // CHECK3-NEXT:  entry:
759 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
760 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
761 // CHECK3-NEXT:    [[B:%.*]] = alloca i32, align 4
762 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
763 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
764 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
765 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
766 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
767 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
768 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i8, align 1
769 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED8:%.*]] = alloca i32, align 4
770 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [2 x i8*], align 4
771 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [2 x i8*], align 4
772 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [2 x i8*], align 4
773 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
774 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
775 // CHECK3-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
776 // CHECK3-NEXT:    store i32 1, i32* [[B]], align 4
777 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
778 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP0]], 3
779 // CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[CMP]] to i8
780 // CHECK3-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
781 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
782 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[B_CASTED]], align 4
783 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B_CASTED]], align 4
784 // CHECK3-NEXT:    [[TMP3:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
785 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP3]] to i1
786 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i8*
787 // CHECK3-NEXT:    [[FROMBOOL2:%.*]] = zext i1 [[TOBOOL]] to i8
788 // CHECK3-NEXT:    store i8 [[FROMBOOL2]], i8* [[CONV]], align 1
789 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
790 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
791 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
792 // CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to %struct.S1**
793 // CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP6]], align 4
794 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
795 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to double**
796 // CHECK3-NEXT:    store double* [[A]], double** [[TMP8]], align 4
797 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
798 // CHECK3-NEXT:    store i8* null, i8** [[TMP9]], align 4
799 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
800 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
801 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[TMP11]], align 4
802 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
803 // CHECK3-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
804 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[TMP13]], align 4
805 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
806 // CHECK3-NEXT:    store i8* null, i8** [[TMP14]], align 4
807 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
808 // CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
809 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[TMP16]], align 4
810 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
811 // CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
812 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[TMP18]], align 4
813 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
814 // CHECK3-NEXT:    store i8* null, i8** [[TMP19]], align 4
815 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
816 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
817 // CHECK3-NEXT:    [[TMP22:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
818 // CHECK3-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP22]] to i1
819 // CHECK3-NEXT:    [[TMP23:%.*]] = select i1 [[TOBOOL3]], i32 0, i32 1
820 // CHECK3-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
821 // CHECK3-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
822 // CHECK3-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
823 // CHECK3:       omp_offload.failed:
824 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i32 [[TMP2]], i32 [[TMP4]]) #[[ATTR3:[0-9]+]]
825 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
826 // CHECK3:       omp_offload.cont:
827 // CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
828 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP26]], 5
829 // CHECK3-NEXT:    [[FROMBOOL6:%.*]] = zext i1 [[CMP5]] to i8
830 // CHECK3-NEXT:    store i8 [[FROMBOOL6]], i8* [[DOTCAPTURE_EXPR_4]], align 1
831 // CHECK3-NEXT:    [[TMP27:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_4]], align 1
832 // CHECK3-NEXT:    [[TOBOOL7:%.*]] = trunc i8 [[TMP27]] to i1
833 // CHECK3-NEXT:    [[CONV9:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED8]] to i8*
834 // CHECK3-NEXT:    [[FROMBOOL10:%.*]] = zext i1 [[TOBOOL7]] to i8
835 // CHECK3-NEXT:    store i8 [[FROMBOOL10]], i8* [[CONV9]], align 1
836 // CHECK3-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED8]], align 4
837 // CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[N_ADDR]], align 4
838 // CHECK3-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], 4
839 // CHECK3-NEXT:    br i1 [[CMP11]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
840 // CHECK3:       omp_if.then:
841 // CHECK3-NEXT:    [[A12:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
842 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
843 // CHECK3-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to %struct.S1**
844 // CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP31]], align 4
845 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
846 // CHECK3-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to double**
847 // CHECK3-NEXT:    store double* [[A12]], double** [[TMP33]], align 4
848 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
849 // CHECK3-NEXT:    store i8* null, i8** [[TMP34]], align 4
850 // CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 1
851 // CHECK3-NEXT:    [[TMP36:%.*]] = bitcast i8** [[TMP35]] to i32*
852 // CHECK3-NEXT:    store i32 [[TMP28]], i32* [[TMP36]], align 4
853 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 1
854 // CHECK3-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
855 // CHECK3-NEXT:    store i32 [[TMP28]], i32* [[TMP38]], align 4
856 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 1
857 // CHECK3-NEXT:    store i8* null, i8** [[TMP39]], align 4
858 // CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
859 // CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
860 // CHECK3-NEXT:    [[TMP42:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_4]], align 1
861 // CHECK3-NEXT:    [[TOBOOL16:%.*]] = trunc i8 [[TMP42]] to i1
862 // CHECK3-NEXT:    [[TMP43:%.*]] = select i1 [[TOBOOL16]], i32 0, i32 1
863 // CHECK3-NEXT:    [[TMP44:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 2, i8** [[TMP40]], i8** [[TMP41]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP43]])
864 // CHECK3-NEXT:    [[TMP45:%.*]] = icmp ne i32 [[TMP44]], 0
865 // CHECK3-NEXT:    br i1 [[TMP45]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
866 // CHECK3:       omp_offload.failed17:
867 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]], i32 [[TMP28]]) #[[ATTR3]]
868 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
869 // CHECK3:       omp_offload.cont18:
870 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
871 // CHECK3:       omp_if.else:
872 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]], i32 [[TMP28]]) #[[ATTR3]]
873 // CHECK3-NEXT:    br label [[OMP_IF_END]]
874 // CHECK3:       omp_if.end:
875 // CHECK3-NEXT:    [[A19:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
876 // CHECK3-NEXT:    [[TMP46:%.*]] = load double, double* [[A19]], align 4
877 // CHECK3-NEXT:    [[CONV20:%.*]] = fptosi double [[TMP46]] to i32
878 // CHECK3-NEXT:    ret i32 [[CONV20]]
879 //
880 //
881 // CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici
882 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
883 // CHECK3-NEXT:  entry:
884 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
885 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
886 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
887 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
888 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
889 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
890 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
891 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
892 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP0]], 1
893 // CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[CMP]] to i8
894 // CHECK3-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
895 // CHECK3-NEXT:    [[TMP1:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
896 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP1]] to i1
897 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i8*
898 // CHECK3-NEXT:    [[FROMBOOL1:%.*]] = zext i1 [[TOBOOL]] to i8
899 // CHECK3-NEXT:    store i8 [[FROMBOOL1]], i8* [[CONV]], align 1
900 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
901 // CHECK3-NEXT:    [[TMP3:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
902 // CHECK3-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP3]] to i1
903 // CHECK3-NEXT:    br i1 [[TOBOOL2]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
904 // CHECK3:       omp_if.then:
905 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
906 // CHECK3-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32*
907 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[TMP5]], align 4
908 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
909 // CHECK3-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
910 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[TMP7]], align 4
911 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
912 // CHECK3-NEXT:    store i8* null, i8** [[TMP8]], align 4
913 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
914 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
915 // CHECK3-NEXT:    [[TMP11:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
916 // CHECK3-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP11]] to i1
917 // CHECK3-NEXT:    [[TMP12:%.*]] = select i1 [[TOBOOL3]], i32 0, i32 1
918 // CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP9]], i8** [[TMP10]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP12]])
919 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
920 // CHECK3-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
921 // CHECK3:       omp_offload.failed:
922 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i32 [[TMP2]]) #[[ATTR3]]
923 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
924 // CHECK3:       omp_offload.cont:
925 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
926 // CHECK3:       omp_if.else:
927 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i32 [[TMP2]]) #[[ATTR3]]
928 // CHECK3-NEXT:    br label [[OMP_IF_END]]
929 // CHECK3:       omp_if.end:
930 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[N_ADDR]], align 4
931 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 2
932 // CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[SUB]], 2
933 // CHECK3-NEXT:    br i1 [[CMP4]], label [[OMP_IF_THEN5:%.*]], label [[OMP_IF_ELSE8:%.*]]
934 // CHECK3:       omp_if.then5:
935 // CHECK3-NEXT:    [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0)
936 // CHECK3-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
937 // CHECK3-NEXT:    br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
938 // CHECK3:       omp_offload.failed6:
939 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108() #[[ATTR3]]
940 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
941 // CHECK3:       omp_offload.cont7:
942 // CHECK3-NEXT:    br label [[OMP_IF_END9:%.*]]
943 // CHECK3:       omp_if.else8:
944 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108() #[[ATTR3]]
945 // CHECK3-NEXT:    br label [[OMP_IF_END9]]
946 // CHECK3:       omp_if.end9:
947 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[N_ADDR]], align 4
948 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], 1
949 // CHECK3-NEXT:    ret i32 [[ADD]]
950 //
951 //
952 // CHECK3-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
953 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
954 // CHECK3-NEXT:  entry:
955 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
956 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
957 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
958 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
959 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
960 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
961 // CHECK3-NEXT:    [[B:%.*]] = alloca i16, align 2
962 // CHECK3-NEXT:    [[A_CASTED1:%.*]] = alloca i32, align 4
963 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
964 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS2:%.*]] = alloca [2 x i8*], align 4
965 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS3:%.*]] = alloca [2 x i8*], align 4
966 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS4:%.*]] = alloca [2 x i8*], align 4
967 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
968 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
969 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
970 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
971 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
972 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
973 // CHECK3-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32*
974 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP3]], align 4
975 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
976 // CHECK3-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32*
977 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP5]], align 4
978 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
979 // CHECK3-NEXT:    store i8* null, i8** [[TMP6]], align 4
980 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
981 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
982 // CHECK3-NEXT:    [[TMP9:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l87.region_id, i32 1, i8** [[TMP7]], i8** [[TMP8]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.9, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1)
983 // CHECK3-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
984 // CHECK3-NEXT:    br i1 [[TMP10]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
985 // CHECK3:       omp_offload.failed:
986 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l87(i32 [[TMP1]]) #[[ATTR3]]
987 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
988 // CHECK3:       omp_offload.cont:
989 // CHECK3-NEXT:    store i16 1, i16* [[B]], align 2
990 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A]], align 4
991 // CHECK3-NEXT:    store i32 [[TMP11]], i32* [[A_CASTED1]], align 4
992 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[A_CASTED1]], align 4
993 // CHECK3-NEXT:    [[TMP13:%.*]] = load i16, i16* [[B]], align 2
994 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
995 // CHECK3-NEXT:    store i16 [[TMP13]], i16* [[CONV]], align 2
996 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[B_CASTED]], align 4
997 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS2]], i32 0, i32 0
998 // CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
999 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP16]], align 4
1000 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS3]], i32 0, i32 0
1001 // CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
1002 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP18]], align 4
1003 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS4]], i32 0, i32 0
1004 // CHECK3-NEXT:    store i8* null, i8** [[TMP19]], align 4
1005 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS2]], i32 0, i32 1
1006 // CHECK3-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
1007 // CHECK3-NEXT:    store i32 [[TMP14]], i32* [[TMP21]], align 4
1008 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS3]], i32 0, i32 1
1009 // CHECK3-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
1010 // CHECK3-NEXT:    store i32 [[TMP14]], i32* [[TMP23]], align 4
1011 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS4]], i32 0, i32 1
1012 // CHECK3-NEXT:    store i8* null, i8** [[TMP24]], align 4
1013 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS2]], i32 0, i32 0
1014 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS3]], i32 0, i32 0
1015 // CHECK3-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 2, i8** [[TMP25]], i8** [[TMP26]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
1016 // CHECK3-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
1017 // CHECK3-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
1018 // CHECK3:       omp_offload.failed5:
1019 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i32 [[TMP12]], i32 [[TMP14]]) #[[ATTR3]]
1020 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
1021 // CHECK3:       omp_offload.cont6:
1022 // CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[A]], align 4
1023 // CHECK3-NEXT:    ret i32 [[TMP29]]
1024 //
1025 //
1026 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
1027 // CHECK3-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
1028 // CHECK3-NEXT:  entry:
1029 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1030 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1031 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1032 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1033 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
1034 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1035 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1036 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1037 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1038 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1039 // CHECK3-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1040 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8*
1041 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B_ADDR]], align 4
1042 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[B_CASTED]], align 4
1043 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
1044 // CHECK3-NEXT:    [[TMP4:%.*]] = load i8, i8* [[CONV]], align 1
1045 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP4]] to i1
1046 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1047 // CHECK3:       omp_if.then:
1048 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%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 [[TMP3]])
1049 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
1050 // CHECK3:       omp_if.else:
1051 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1052 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
1053 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1054 // CHECK3-NEXT:    call void @.omp_outlined.(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], %struct.S1* [[TMP1]], i32 [[TMP3]]) #[[ATTR3]]
1055 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1056 // CHECK3-NEXT:    br label [[OMP_IF_END]]
1057 // CHECK3:       omp_if.end:
1058 // CHECK3-NEXT:    ret void
1059 //
1060 //
1061 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
1062 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]]) #[[ATTR2:[0-9]+]] {
1063 // CHECK3-NEXT:  entry:
1064 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1065 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1066 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1067 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1068 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1069 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1070 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1071 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1072 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1073 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
1074 // CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to double
1075 // CHECK3-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
1076 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1077 // CHECK3-NEXT:    store double [[ADD]], double* [[A]], align 4
1078 // CHECK3-NEXT:    ret void
1079 //
1080 //
1081 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
1082 // CHECK3-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1083 // CHECK3-NEXT:  entry:
1084 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1085 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1086 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
1087 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1088 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1089 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1090 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1091 // CHECK3-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1092 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8*
1093 // CHECK3-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
1094 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
1095 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1096 // CHECK3:       omp_if.then:
1097 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%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]])
1098 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
1099 // CHECK3:       omp_if.else:
1100 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1101 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
1102 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1103 // CHECK3-NEXT:    call void @.omp_outlined..1(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], %struct.S1* [[TMP1]]) #[[ATTR3]]
1104 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1105 // CHECK3-NEXT:    br label [[OMP_IF_END]]
1106 // CHECK3:       omp_if.end:
1107 // CHECK3-NEXT:    ret void
1108 //
1109 //
1110 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
1111 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]]) #[[ATTR2]] {
1112 // CHECK3-NEXT:  entry:
1113 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1114 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1115 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1116 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1117 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1118 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1119 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1120 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1121 // CHECK3-NEXT:    store double 2.500000e+00, double* [[A]], align 4
1122 // CHECK3-NEXT:    ret void
1123 //
1124 //
1125 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
1126 // CHECK3-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1127 // CHECK3-NEXT:  entry:
1128 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1129 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
1130 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1131 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1132 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1133 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8*
1134 // CHECK3-NEXT:    [[TMP1:%.*]] = load i8, i8* [[CONV]], align 1
1135 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP1]] to i1
1136 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1137 // CHECK3:       omp_if.then:
1138 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
1139 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
1140 // CHECK3:       omp_if.else:
1141 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1142 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
1143 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1144 // CHECK3-NEXT:    call void @.omp_outlined..4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]]) #[[ATTR3]]
1145 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1146 // CHECK3-NEXT:    br label [[OMP_IF_END]]
1147 // CHECK3:       omp_if.end:
1148 // CHECK3-NEXT:    ret void
1149 //
1150 //
1151 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
1152 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
1153 // CHECK3-NEXT:  entry:
1154 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1155 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1156 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1157 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1158 // CHECK3-NEXT:    ret void
1159 //
1160 //
1161 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
1162 // CHECK3-SAME: () #[[ATTR1]] {
1163 // CHECK3-NEXT:  entry:
1164 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..7 to void (i32*, i32*, ...)*))
1165 // CHECK3-NEXT:    ret void
1166 //
1167 //
1168 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
1169 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
1170 // CHECK3-NEXT:  entry:
1171 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1172 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1173 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1174 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1175 // CHECK3-NEXT:    ret void
1176 //
1177 //
1178 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l87
1179 // CHECK3-SAME: (i32 noundef [[A:%.*]]) #[[ATTR1]] {
1180 // CHECK3-NEXT:  entry:
1181 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1182 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
1183 // CHECK3-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
1184 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1185 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1186 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
1187 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
1188 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
1189 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
1190 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1191 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
1192 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1193 // CHECK3-NEXT:    call void @.omp_outlined..8(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i32 [[TMP2]]) #[[ATTR3]]
1194 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1195 // CHECK3-NEXT:    ret void
1196 //
1197 //
1198 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..8
1199 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR2]] {
1200 // CHECK3-NEXT:  entry:
1201 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1202 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1203 // CHECK3-NEXT:    [[A_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:    store i32 [[A]], i32* [[A_ADDR]], align 4
1207 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
1208 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
1209 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
1210 // CHECK3-NEXT:    ret void
1211 //
1212 //
1213 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
1214 // CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR1]] {
1215 // CHECK3-NEXT:  entry:
1216 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1217 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1218 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
1219 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1220 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
1221 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1222 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1223 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
1224 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
1225 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
1226 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
1227 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[B_CASTED]] to i16*
1228 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
1229 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
1230 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
1231 // CHECK3-NEXT:    ret void
1232 //
1233 //
1234 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
1235 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR2]] {
1236 // CHECK3-NEXT:  entry:
1237 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1238 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1239 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1240 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1241 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1242 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1243 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
1244 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1245 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1246 // CHECK3-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
1247 // CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
1248 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
1249 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
1250 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
1251 // CHECK3-NEXT:    ret void
1252 //
1253 //
1254 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1255 // CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
1256 // CHECK3-NEXT:  entry:
1257 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
1258 // CHECK3-NEXT:    ret void
1259 //
1260 //
1261 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
1262 // CHECK9-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
1263 // CHECK9-NEXT:  entry:
1264 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1265 // CHECK9-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
1266 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1267 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
1268 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1269 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
1270 // CHECK9-NEXT:    [[TMP1:%.*]] = load i8, i8* [[CONV]], align 1
1271 // CHECK9-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP1]] to i1
1272 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1273 // CHECK9:       omp_if.then:
1274 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
1275 // CHECK9-NEXT:    br label [[OMP_IF_END:%.*]]
1276 // CHECK9:       omp_if.else:
1277 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1278 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
1279 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1280 // CHECK9-NEXT:    call void @.omp_outlined.(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]]) #[[ATTR2:[0-9]+]]
1281 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1282 // CHECK9-NEXT:    br label [[OMP_IF_END]]
1283 // CHECK9:       omp_if.end:
1284 // CHECK9-NEXT:    ret void
1285 //
1286 //
1287 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
1288 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
1289 // CHECK9-NEXT:  entry:
1290 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1291 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1292 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1293 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1294 // CHECK9-NEXT:    ret void
1295 //
1296 //
1297 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
1298 // CHECK9-SAME: () #[[ATTR0]] {
1299 // CHECK9-NEXT:  entry:
1300 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
1301 // CHECK9-NEXT:    ret void
1302 //
1303 //
1304 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
1305 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1306 // CHECK9-NEXT:  entry:
1307 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1308 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1309 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1310 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1311 // CHECK9-NEXT:    ret void
1312 //
1313 //
1314 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
1315 // CHECK9-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1316 // CHECK9-NEXT:  entry:
1317 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1318 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1319 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1320 // CHECK9-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
1321 // CHECK9-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
1322 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1323 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1324 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1325 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1326 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1327 // CHECK9-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1328 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1329 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
1330 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
1331 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
1332 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[CONV2]], align 4
1333 // CHECK9-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
1334 // CHECK9-NEXT:    [[TMP4:%.*]] = load i8, i8* [[CONV1]], align 1
1335 // CHECK9-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP4]] to i1
1336 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1337 // CHECK9:       omp_if.then:
1338 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%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 [[TMP3]])
1339 // CHECK9-NEXT:    br label [[OMP_IF_END:%.*]]
1340 // CHECK9:       omp_if.else:
1341 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1342 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
1343 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1344 // CHECK9-NEXT:    call void @.omp_outlined..2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], %struct.S1* [[TMP1]], i64 [[TMP3]]) #[[ATTR2]]
1345 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1346 // CHECK9-NEXT:    br label [[OMP_IF_END]]
1347 // CHECK9:       omp_if.end:
1348 // CHECK9-NEXT:    ret void
1349 //
1350 //
1351 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
1352 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]]) #[[ATTR1]] {
1353 // CHECK9-NEXT:  entry:
1354 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1355 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1356 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1357 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1358 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1359 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1360 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1361 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1362 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1363 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1364 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1365 // CHECK9-NEXT:    [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
1366 // CHECK9-NEXT:    [[ADD:%.*]] = fadd double [[CONV1]], 1.500000e+00
1367 // CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1368 // CHECK9-NEXT:    store double [[ADD]], double* [[A]], align 8
1369 // CHECK9-NEXT:    ret void
1370 //
1371 //
1372 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
1373 // CHECK9-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1374 // CHECK9-NEXT:  entry:
1375 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1376 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1377 // CHECK9-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
1378 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1379 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1380 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1381 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1382 // CHECK9-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1383 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
1384 // CHECK9-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
1385 // CHECK9-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
1386 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1387 // CHECK9:       omp_if.then:
1388 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%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]])
1389 // CHECK9-NEXT:    br label [[OMP_IF_END:%.*]]
1390 // CHECK9:       omp_if.else:
1391 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1392 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
1393 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1394 // CHECK9-NEXT:    call void @.omp_outlined..3(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], %struct.S1* [[TMP1]]) #[[ATTR2]]
1395 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1396 // CHECK9-NEXT:    br label [[OMP_IF_END]]
1397 // CHECK9:       omp_if.end:
1398 // CHECK9-NEXT:    ret void
1399 //
1400 //
1401 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
1402 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]]) #[[ATTR1]] {
1403 // CHECK9-NEXT:  entry:
1404 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1405 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1406 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1407 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1408 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1409 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1410 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1411 // CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1412 // CHECK9-NEXT:    store double 2.500000e+00, double* [[A]], align 8
1413 // CHECK9-NEXT:    ret void
1414 //
1415 //
1416 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l87
1417 // CHECK9-SAME: (i64 noundef [[A:%.*]]) #[[ATTR0]] {
1418 // CHECK9-NEXT:  entry:
1419 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1420 // CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1421 // CHECK9-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
1422 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1423 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1424 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1425 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1426 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1427 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1428 // CHECK9-NEXT:    store i32 [[TMP1]], i32* [[CONV1]], align 4
1429 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
1430 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1431 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
1432 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1433 // CHECK9-NEXT:    call void @.omp_outlined..4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP2]]) #[[ATTR2]]
1434 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1435 // CHECK9-NEXT:    ret void
1436 //
1437 //
1438 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
1439 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR1]] {
1440 // CHECK9-NEXT:  entry:
1441 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1442 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1443 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1444 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1445 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1446 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1447 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1448 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
1449 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
1450 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
1451 // CHECK9-NEXT:    ret void
1452 //
1453 //
1454 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
1455 // CHECK9-SAME: (i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
1456 // CHECK9-NEXT:  entry:
1457 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1458 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1459 // CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1460 // CHECK9-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
1461 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1462 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1463 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1464 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
1465 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
1466 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1467 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
1468 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
1469 // CHECK9-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
1470 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i16*
1471 // CHECK9-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
1472 // CHECK9-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
1473 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
1474 // CHECK9-NEXT:    ret void
1475 //
1476 //
1477 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
1478 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR1]] {
1479 // CHECK9-NEXT:  entry:
1480 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1481 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1482 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1483 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1484 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1485 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1486 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1487 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1488 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1489 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
1490 // CHECK9-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 2
1491 // CHECK9-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
1492 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1493 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
1494 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
1495 // CHECK9-NEXT:    ret void
1496 //
1497 //
1498 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
1499 // CHECK11-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
1500 // CHECK11-NEXT:  entry:
1501 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1502 // CHECK11-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
1503 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1504 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
1505 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1506 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8*
1507 // CHECK11-NEXT:    [[TMP1:%.*]] = load i8, i8* [[CONV]], align 1
1508 // CHECK11-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP1]] to i1
1509 // CHECK11-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1510 // CHECK11:       omp_if.then:
1511 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
1512 // CHECK11-NEXT:    br label [[OMP_IF_END:%.*]]
1513 // CHECK11:       omp_if.else:
1514 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1515 // CHECK11-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
1516 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1517 // CHECK11-NEXT:    call void @.omp_outlined.(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]]) #[[ATTR2:[0-9]+]]
1518 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1519 // CHECK11-NEXT:    br label [[OMP_IF_END]]
1520 // CHECK11:       omp_if.end:
1521 // CHECK11-NEXT:    ret void
1522 //
1523 //
1524 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
1525 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
1526 // CHECK11-NEXT:  entry:
1527 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1528 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1529 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1530 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1531 // CHECK11-NEXT:    ret void
1532 //
1533 //
1534 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
1535 // CHECK11-SAME: () #[[ATTR0]] {
1536 // CHECK11-NEXT:  entry:
1537 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
1538 // CHECK11-NEXT:    ret void
1539 //
1540 //
1541 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
1542 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1543 // CHECK11-NEXT:  entry:
1544 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1545 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1546 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1547 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1548 // CHECK11-NEXT:    ret void
1549 //
1550 //
1551 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
1552 // CHECK11-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1553 // CHECK11-NEXT:  entry:
1554 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1555 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1556 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1557 // CHECK11-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1558 // CHECK11-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
1559 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1560 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1561 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1562 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1563 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1564 // CHECK11-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1565 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8*
1566 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B_ADDR]], align 4
1567 // CHECK11-NEXT:    store i32 [[TMP2]], i32* [[B_CASTED]], align 4
1568 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
1569 // CHECK11-NEXT:    [[TMP4:%.*]] = load i8, i8* [[CONV]], align 1
1570 // CHECK11-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP4]] to i1
1571 // CHECK11-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1572 // CHECK11:       omp_if.then:
1573 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%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 [[TMP3]])
1574 // CHECK11-NEXT:    br label [[OMP_IF_END:%.*]]
1575 // CHECK11:       omp_if.else:
1576 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1577 // CHECK11-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
1578 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1579 // CHECK11-NEXT:    call void @.omp_outlined..2(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], %struct.S1* [[TMP1]], i32 [[TMP3]]) #[[ATTR2]]
1580 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1581 // CHECK11-NEXT:    br label [[OMP_IF_END]]
1582 // CHECK11:       omp_if.end:
1583 // CHECK11-NEXT:    ret void
1584 //
1585 //
1586 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
1587 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]]) #[[ATTR1]] {
1588 // CHECK11-NEXT:  entry:
1589 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1590 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1591 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1592 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1593 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1594 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1595 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1596 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1597 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1598 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
1599 // CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to double
1600 // CHECK11-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
1601 // CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1602 // CHECK11-NEXT:    store double [[ADD]], double* [[A]], align 4
1603 // CHECK11-NEXT:    ret void
1604 //
1605 //
1606 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
1607 // CHECK11-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1608 // CHECK11-NEXT:  entry:
1609 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1610 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1611 // CHECK11-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
1612 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1613 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1614 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1615 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1616 // CHECK11-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1617 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8*
1618 // CHECK11-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
1619 // CHECK11-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
1620 // CHECK11-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1621 // CHECK11:       omp_if.then:
1622 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%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]])
1623 // CHECK11-NEXT:    br label [[OMP_IF_END:%.*]]
1624 // CHECK11:       omp_if.else:
1625 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1626 // CHECK11-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
1627 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1628 // CHECK11-NEXT:    call void @.omp_outlined..3(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], %struct.S1* [[TMP1]]) #[[ATTR2]]
1629 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1630 // CHECK11-NEXT:    br label [[OMP_IF_END]]
1631 // CHECK11:       omp_if.end:
1632 // CHECK11-NEXT:    ret void
1633 //
1634 //
1635 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
1636 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]]) #[[ATTR1]] {
1637 // CHECK11-NEXT:  entry:
1638 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1639 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1640 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1641 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1642 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1643 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1644 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1645 // CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1646 // CHECK11-NEXT:    store double 2.500000e+00, double* [[A]], align 4
1647 // CHECK11-NEXT:    ret void
1648 //
1649 //
1650 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l87
1651 // CHECK11-SAME: (i32 noundef [[A:%.*]]) #[[ATTR0]] {
1652 // CHECK11-NEXT:  entry:
1653 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1654 // CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
1655 // CHECK11-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
1656 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1657 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1658 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
1659 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
1660 // CHECK11-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
1661 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
1662 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1663 // CHECK11-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
1664 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1665 // CHECK11-NEXT:    call void @.omp_outlined..4(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], i32 [[TMP2]]) #[[ATTR2]]
1666 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
1667 // CHECK11-NEXT:    ret void
1668 //
1669 //
1670 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
1671 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR1]] {
1672 // CHECK11-NEXT:  entry:
1673 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1674 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1675 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1676 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1677 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1678 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
1679 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
1680 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
1681 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
1682 // CHECK11-NEXT:    ret void
1683 //
1684 //
1685 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
1686 // CHECK11-SAME: (i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR0]] {
1687 // CHECK11-NEXT:  entry:
1688 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1689 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1690 // CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
1691 // CHECK11-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1692 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
1693 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1694 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1695 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
1696 // CHECK11-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
1697 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
1698 // CHECK11-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
1699 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[B_CASTED]] to i16*
1700 // CHECK11-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
1701 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
1702 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
1703 // CHECK11-NEXT:    ret void
1704 //
1705 //
1706 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
1707 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR1]] {
1708 // CHECK11-NEXT:  entry:
1709 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1710 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1711 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1712 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1713 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1714 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1715 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
1716 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1717 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1718 // CHECK11-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
1719 // CHECK11-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
1720 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
1721 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
1722 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
1723 // CHECK11-NEXT:    ret void
1724 //
1725