1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
2 // Test host codegen.
3 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
4 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
5 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
6 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
7 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
8 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
9 
10 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
11 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
12 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
13 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
14 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
15 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
16 
17 // Test target codegen - host bc file has to be created first.
18 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
19 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=CHECK9
20 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
21 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
22 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
23 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix=CHECK11
24 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
25 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
26 
27 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
28 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
29 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
30 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
31 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
32 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
33 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
34 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
35 
36 // Test host codegen.
37 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
38 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
39 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
40 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
41 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
42 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
43 
44 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
45 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
46 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
47 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
48 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
49 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
50 
51 // Test target codegen - host bc file has to be created first.
52 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
53 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=CHECK9
54 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
55 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
56 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
57 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix=CHECK11
58 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
59 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
60 
61 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
62 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
63 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
64 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
65 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
66 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
67 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
68 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
69 
70 // expected-no-diagnostics
71 #ifndef HEADER
72 #define HEADER
73 
74 
75 
76 
77 // We have 6 target regions
78 
79 
80 
81 // Check target registration is registered as a Ctor.
82 
83 
84 template<typename tx>
ftemplate(int n)85 tx ftemplate(int n) {
86   tx a = 0;
87 
88   #pragma omp target teams thread_limit(tx(20))
89   {
90   }
91 
92   short b = 1;
93   #pragma omp target teams num_teams(b) thread_limit(1024)
94   {
95     a += b;
96   }
97 
98   return a;
99 }
100 
101 static
fstatic(int n)102 int fstatic(int n) {
103 
104   #pragma omp target teams num_teams(n) thread_limit(n*32)
105   {
106   }
107 
108   #pragma omp target teams thread_limit(32+n)
109   {
110   }
111 
112   return n+1;
113 }
114 
115 struct S1 {
116   double a;
117 
r1S1118   int r1(int n){
119     int b = 1;
120 
121     #pragma omp target teams thread_limit(n-b)
122     {
123       this->a = (double)b + 1.5;
124     }
125 
126     #pragma omp target teams thread_limit(1024)
127     {
128       this->a = 2.5;
129     }
130 
131     return (int)a;
132   }
133 };
134 
bar(int n)135 int bar(int n){
136   int a = 0;
137 
138   S1 S;
139   a += S.r1(n);
140 
141   a += fstatic(n);
142 
143   a += ftemplate<int>(n);
144 
145   return a;
146 }
147 
148 
149 
150 
151 
152 
153 
154 
155 
156 
157 
158 
159 
160 
161 
162 
163 
164 
165 
166 
167 
168 // Check that the offloading functions are emitted and that the parallel function
169 // is appropriately guarded.
170 
171 
172 
173 
174 
175 
176 #endif
177 // CHECK1-LABEL: define {{[^@]+}}@_Z3bari
178 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
179 // CHECK1-NEXT:  entry:
180 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
181 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
182 // CHECK1-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
183 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
184 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
185 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
186 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP0]])
187 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
188 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
189 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
190 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
191 // CHECK1-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP2]])
192 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
193 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
194 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
195 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
196 // CHECK1-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP4]])
197 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
198 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
199 // CHECK1-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
200 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
201 // CHECK1-NEXT:    ret i32 [[TMP6]]
202 //
203 //
204 // CHECK1-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
205 // CHECK1-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
206 // CHECK1-NEXT:  entry:
207 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
208 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
209 // CHECK1-NEXT:    [[B:%.*]] = alloca i32, align 4
210 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
211 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
212 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
213 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
214 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
215 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
216 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
217 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
218 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
219 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
220 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
221 // CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
222 // CHECK1-NEXT:    store i32 1, i32* [[B]], align 4
223 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
224 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
225 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
226 // CHECK1-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
227 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
228 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
229 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
230 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
231 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
232 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
233 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV2]], align 4
234 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
235 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
236 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
237 // CHECK1-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
238 // CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 8
239 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
240 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
241 // CHECK1-NEXT:    store double* [[A]], double** [[TMP9]], align 8
242 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
243 // CHECK1-NEXT:    store i8* null, i8** [[TMP10]], align 8
244 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
245 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
246 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP12]], align 8
247 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
248 // CHECK1-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
249 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP14]], align 8
250 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
251 // CHECK1-NEXT:    store i8* null, i8** [[TMP15]], align 8
252 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
253 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
254 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP17]], align 8
255 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
256 // CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
257 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP19]], align 8
258 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
259 // CHECK1-NEXT:    store i8* null, i8** [[TMP20]], align 8
260 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
261 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
262 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
263 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
264 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
265 // CHECK1-NEXT:    store i32 1, i32* [[TMP24]], align 4
266 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
267 // CHECK1-NEXT:    store i32 3, i32* [[TMP25]], align 4
268 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
269 // CHECK1-NEXT:    store i8** [[TMP21]], i8*** [[TMP26]], align 8
270 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
271 // CHECK1-NEXT:    store i8** [[TMP22]], i8*** [[TMP27]], align 8
272 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
273 // CHECK1-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP28]], align 8
274 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
275 // CHECK1-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP29]], align 8
276 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
277 // CHECK1-NEXT:    store i8** null, i8*** [[TMP30]], align 8
278 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
279 // CHECK1-NEXT:    store i8** null, i8*** [[TMP31]], align 8
280 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
281 // CHECK1-NEXT:    store i64 0, i64* [[TMP32]], align 8
282 // CHECK1-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i32 0, i32 [[TMP23]], i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
283 // CHECK1-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
284 // CHECK1-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
285 // CHECK1:       omp_offload.failed:
286 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR2:[0-9]+]]
287 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
288 // CHECK1:       omp_offload.cont:
289 // CHECK1-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
290 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
291 // CHECK1-NEXT:    [[TMP36:%.*]] = bitcast i8** [[TMP35]] to %struct.S1**
292 // CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP36]], align 8
293 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
294 // CHECK1-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to double**
295 // CHECK1-NEXT:    store double* [[A3]], double** [[TMP38]], align 8
296 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
297 // CHECK1-NEXT:    store i8* null, i8** [[TMP39]], align 8
298 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
299 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
300 // CHECK1-NEXT:    [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
301 // CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
302 // CHECK1-NEXT:    store i32 1, i32* [[TMP42]], align 4
303 // CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
304 // CHECK1-NEXT:    store i32 1, i32* [[TMP43]], align 4
305 // CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
306 // CHECK1-NEXT:    store i8** [[TMP40]], i8*** [[TMP44]], align 8
307 // CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
308 // CHECK1-NEXT:    store i8** [[TMP41]], i8*** [[TMP45]], align 8
309 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
310 // CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP46]], align 8
311 // CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
312 // CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP47]], align 8
313 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
314 // CHECK1-NEXT:    store i8** null, i8*** [[TMP48]], align 8
315 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
316 // CHECK1-NEXT:    store i8** null, i8*** [[TMP49]], align 8
317 // CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
318 // CHECK1-NEXT:    store i64 0, i64* [[TMP50]], align 8
319 // CHECK1-NEXT:    [[TMP51:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 1024, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
320 // CHECK1-NEXT:    [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0
321 // CHECK1-NEXT:    br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
322 // CHECK1:       omp_offload.failed8:
323 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR2]]
324 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
325 // CHECK1:       omp_offload.cont9:
326 // CHECK1-NEXT:    [[A10:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
327 // CHECK1-NEXT:    [[TMP53:%.*]] = load double, double* [[A10]], align 8
328 // CHECK1-NEXT:    [[CONV11:%.*]] = fptosi double [[TMP53]] to i32
329 // CHECK1-NEXT:    ret i32 [[CONV11]]
330 //
331 //
332 // CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici
333 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
334 // CHECK1-NEXT:  entry:
335 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
336 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
337 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
338 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
339 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8
340 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 8
341 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 8
342 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 8
343 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
344 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED5:%.*]] = alloca i64, align 8
345 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [1 x i8*], align 8
346 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [1 x i8*], align 8
347 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [1 x i8*], align 8
348 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
349 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
350 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
351 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
352 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP1]], 32
353 // CHECK1-NEXT:    store i32 [[MUL]], i32* [[DOTCAPTURE_EXPR_1]], align 4
354 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
355 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
356 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
357 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
358 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
359 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED2]] to i32*
360 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
361 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED2]], align 8
362 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
363 // CHECK1-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
364 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP7]], align 8
365 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
366 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
367 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP9]], align 8
368 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
369 // CHECK1-NEXT:    store i8* null, i8** [[TMP10]], align 8
370 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
371 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
372 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP12]], align 8
373 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
374 // CHECK1-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
375 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP14]], align 8
376 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
377 // CHECK1-NEXT:    store i8* null, i8** [[TMP15]], align 8
378 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
379 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
380 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
381 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
382 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
383 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
384 // CHECK1-NEXT:    store i32 1, i32* [[TMP20]], align 4
385 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
386 // CHECK1-NEXT:    store i32 2, i32* [[TMP21]], align 4
387 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
388 // CHECK1-NEXT:    store i8** [[TMP16]], i8*** [[TMP22]], align 8
389 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
390 // CHECK1-NEXT:    store i8** [[TMP17]], i8*** [[TMP23]], align 8
391 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
392 // CHECK1-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64** [[TMP24]], align 8
393 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
394 // CHECK1-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i64** [[TMP25]], align 8
395 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
396 // CHECK1-NEXT:    store i8** null, i8*** [[TMP26]], align 8
397 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
398 // CHECK1-NEXT:    store i8** null, i8*** [[TMP27]], align 8
399 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
400 // CHECK1-NEXT:    store i64 0, i64* [[TMP28]], align 8
401 // CHECK1-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 [[TMP18]], i32 [[TMP19]], i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
402 // CHECK1-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
403 // CHECK1-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
404 // CHECK1:       omp_offload.failed:
405 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR2]]
406 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
407 // CHECK1:       omp_offload.cont:
408 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[N_ADDR]], align 4
409 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP31]]
410 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_4]], align 4
411 // CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
412 // CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED5]] to i32*
413 // CHECK1-NEXT:    store i32 [[TMP32]], i32* [[CONV6]], align 4
414 // CHECK1-NEXT:    [[TMP33:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED5]], align 8
415 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
416 // CHECK1-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i64*
417 // CHECK1-NEXT:    store i64 [[TMP33]], i64* [[TMP35]], align 8
418 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
419 // CHECK1-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
420 // CHECK1-NEXT:    store i64 [[TMP33]], i64* [[TMP37]], align 8
421 // CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
422 // CHECK1-NEXT:    store i8* null, i8** [[TMP38]], align 8
423 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
424 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
425 // CHECK1-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
426 // CHECK1-NEXT:    [[KERNEL_ARGS10:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
427 // CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS10]], i32 0, i32 0
428 // CHECK1-NEXT:    store i32 1, i32* [[TMP42]], align 4
429 // CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS10]], i32 0, i32 1
430 // CHECK1-NEXT:    store i32 1, i32* [[TMP43]], align 4
431 // CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS10]], i32 0, i32 2
432 // CHECK1-NEXT:    store i8** [[TMP39]], i8*** [[TMP44]], align 8
433 // CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS10]], i32 0, i32 3
434 // CHECK1-NEXT:    store i8** [[TMP40]], i8*** [[TMP45]], align 8
435 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS10]], i32 0, i32 4
436 // CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP46]], align 8
437 // CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS10]], i32 0, i32 5
438 // CHECK1-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP47]], align 8
439 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS10]], i32 0, i32 6
440 // CHECK1-NEXT:    store i8** null, i8*** [[TMP48]], align 8
441 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS10]], i32 0, i32 7
442 // CHECK1-NEXT:    store i8** null, i8*** [[TMP49]], align 8
443 // CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS10]], i32 0, i32 8
444 // CHECK1-NEXT:    store i64 0, i64* [[TMP50]], align 8
445 // CHECK1-NEXT:    [[TMP51:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 [[TMP41]], i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS10]])
446 // CHECK1-NEXT:    [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0
447 // CHECK1-NEXT:    br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
448 // CHECK1:       omp_offload.failed11:
449 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i64 [[TMP33]]) #[[ATTR2]]
450 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT12]]
451 // CHECK1:       omp_offload.cont12:
452 // CHECK1-NEXT:    [[TMP53:%.*]] = load i32, i32* [[N_ADDR]], align 4
453 // CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP53]], 1
454 // CHECK1-NEXT:    ret i32 [[ADD13]]
455 //
456 //
457 // CHECK1-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
458 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
459 // CHECK1-NEXT:  entry:
460 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
461 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
462 // CHECK1-NEXT:    [[B:%.*]] = alloca i16, align 2
463 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
464 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
465 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
466 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
467 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
468 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
469 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
470 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
471 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
472 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
473 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
474 // CHECK1-NEXT:    store i32 1, i32* [[TMP0]], align 4
475 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
476 // CHECK1-NEXT:    store i32 0, i32* [[TMP1]], align 4
477 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
478 // CHECK1-NEXT:    store i8** null, i8*** [[TMP2]], align 8
479 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
480 // CHECK1-NEXT:    store i8** null, i8*** [[TMP3]], align 8
481 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
482 // CHECK1-NEXT:    store i64* null, i64** [[TMP4]], align 8
483 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
484 // CHECK1-NEXT:    store i64* null, i64** [[TMP5]], align 8
485 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
486 // CHECK1-NEXT:    store i8** null, i8*** [[TMP6]], align 8
487 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
488 // CHECK1-NEXT:    store i8** null, i8*** [[TMP7]], align 8
489 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
490 // CHECK1-NEXT:    store i64 0, i64* [[TMP8]], align 8
491 // CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 20, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
492 // CHECK1-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
493 // CHECK1-NEXT:    br i1 [[TMP10]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
494 // CHECK1:       omp_offload.failed:
495 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
496 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
497 // CHECK1:       omp_offload.cont:
498 // CHECK1-NEXT:    store i16 1, i16* [[B]], align 2
499 // CHECK1-NEXT:    [[TMP11:%.*]] = load i16, i16* [[B]], align 2
500 // CHECK1-NEXT:    store i16 [[TMP11]], i16* [[DOTCAPTURE_EXPR_]], align 2
501 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[A]], align 4
502 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
503 // CHECK1-NEXT:    store i32 [[TMP12]], i32* [[CONV]], align 4
504 // CHECK1-NEXT:    [[TMP13:%.*]] = load i64, i64* [[A_CASTED]], align 8
505 // CHECK1-NEXT:    [[TMP14:%.*]] = load i16, i16* [[B]], align 2
506 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_CASTED]] to i16*
507 // CHECK1-NEXT:    store i16 [[TMP14]], i16* [[CONV1]], align 2
508 // CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[B_CASTED]], align 8
509 // CHECK1-NEXT:    [[TMP16:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
510 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i16*
511 // CHECK1-NEXT:    store i16 [[TMP16]], i16* [[CONV2]], align 2
512 // CHECK1-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
513 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
514 // CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
515 // CHECK1-NEXT:    store i64 [[TMP13]], i64* [[TMP19]], align 8
516 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
517 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
518 // CHECK1-NEXT:    store i64 [[TMP13]], i64* [[TMP21]], align 8
519 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
520 // CHECK1-NEXT:    store i8* null, i8** [[TMP22]], align 8
521 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
522 // CHECK1-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
523 // CHECK1-NEXT:    store i64 [[TMP15]], i64* [[TMP24]], align 8
524 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
525 // CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
526 // CHECK1-NEXT:    store i64 [[TMP15]], i64* [[TMP26]], align 8
527 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
528 // CHECK1-NEXT:    store i8* null, i8** [[TMP27]], align 8
529 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
530 // CHECK1-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
531 // CHECK1-NEXT:    store i64 [[TMP17]], i64* [[TMP29]], align 8
532 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
533 // CHECK1-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
534 // CHECK1-NEXT:    store i64 [[TMP17]], i64* [[TMP31]], align 8
535 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
536 // CHECK1-NEXT:    store i8* null, i8** [[TMP32]], align 8
537 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
538 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
539 // CHECK1-NEXT:    [[TMP35:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
540 // CHECK1-NEXT:    [[TMP36:%.*]] = sext i16 [[TMP35]] to i32
541 // CHECK1-NEXT:    [[KERNEL_ARGS3:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
542 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 0
543 // CHECK1-NEXT:    store i32 1, i32* [[TMP37]], align 4
544 // CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 1
545 // CHECK1-NEXT:    store i32 3, i32* [[TMP38]], align 4
546 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 2
547 // CHECK1-NEXT:    store i8** [[TMP33]], i8*** [[TMP39]], align 8
548 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 3
549 // CHECK1-NEXT:    store i8** [[TMP34]], i8*** [[TMP40]], align 8
550 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 4
551 // CHECK1-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP41]], align 8
552 // CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 5
553 // CHECK1-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP42]], align 8
554 // CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 6
555 // CHECK1-NEXT:    store i8** null, i8*** [[TMP43]], align 8
556 // CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 7
557 // CHECK1-NEXT:    store i8** null, i8*** [[TMP44]], align 8
558 // CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]], i32 0, i32 8
559 // CHECK1-NEXT:    store i64 0, i64* [[TMP45]], align 8
560 // CHECK1-NEXT:    [[TMP46:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 [[TMP36]], i32 1024, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS3]])
561 // CHECK1-NEXT:    [[TMP47:%.*]] = icmp ne i32 [[TMP46]], 0
562 // CHECK1-NEXT:    br i1 [[TMP47]], label [[OMP_OFFLOAD_FAILED4:%.*]], label [[OMP_OFFLOAD_CONT5:%.*]]
563 // CHECK1:       omp_offload.failed4:
564 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i64 [[TMP13]], i64 [[TMP15]], i64 [[TMP17]]) #[[ATTR2]]
565 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT5]]
566 // CHECK1:       omp_offload.cont5:
567 // CHECK1-NEXT:    [[TMP48:%.*]] = load i32, i32* [[A]], align 4
568 // CHECK1-NEXT:    ret i32 [[TMP48]]
569 //
570 //
571 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
572 // CHECK1-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
573 // CHECK1-NEXT:  entry:
574 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
575 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
576 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
577 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
578 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
579 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
580 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
581 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
582 // CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
583 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
584 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
585 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 4
586 // CHECK1-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 [[TMP2]])
587 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
588 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
589 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
590 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[B_CASTED]], align 8
591 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP4]])
592 // CHECK1-NEXT:    ret void
593 //
594 //
595 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
596 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]]) #[[ATTR1]] {
597 // CHECK1-NEXT:  entry:
598 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
599 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
600 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
601 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
602 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
603 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
604 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
605 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
606 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
607 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
608 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
609 // CHECK1-NEXT:    [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
610 // CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[CONV1]], 1.500000e+00
611 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
612 // CHECK1-NEXT:    store double [[ADD]], double* [[A]], align 8
613 // CHECK1-NEXT:    ret void
614 //
615 //
616 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
617 // CHECK1-SAME: (%struct.S1* noundef [[THIS:%.*]]) #[[ATTR3:[0-9]+]] {
618 // CHECK1-NEXT:  entry:
619 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
620 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
621 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
622 // CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
623 // CHECK1-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 1024)
624 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
625 // CHECK1-NEXT:    ret void
626 //
627 //
628 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
629 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]]) #[[ATTR1]] {
630 // CHECK1-NEXT:  entry:
631 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
632 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
633 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
634 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
635 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
636 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
637 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
638 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
639 // CHECK1-NEXT:    store double 2.500000e+00, double* [[A]], align 8
640 // CHECK1-NEXT:    ret void
641 //
642 //
643 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
644 // CHECK1-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1]] {
645 // CHECK1-NEXT:  entry:
646 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
647 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
648 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
649 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
650 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
651 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
652 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
653 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
654 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV3]], align 4
655 // CHECK1-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
656 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
657 // CHECK1-NEXT:    ret void
658 //
659 //
660 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
661 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
662 // CHECK1-NEXT:  entry:
663 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
664 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
665 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
666 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
667 // CHECK1-NEXT:    ret void
668 //
669 //
670 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
671 // CHECK1-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
672 // CHECK1-NEXT:  entry:
673 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
674 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
675 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
676 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
677 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
678 // CHECK1-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 [[TMP1]])
679 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..7 to void (i32*, i32*, ...)*))
680 // CHECK1-NEXT:    ret void
681 //
682 //
683 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
684 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
685 // CHECK1-NEXT:  entry:
686 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
687 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
688 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
689 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
690 // CHECK1-NEXT:    ret void
691 //
692 //
693 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
694 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
695 // CHECK1-NEXT:  entry:
696 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
697 // CHECK1-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 20)
698 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*))
699 // CHECK1-NEXT:    ret void
700 //
701 //
702 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..10
703 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
704 // CHECK1-NEXT:  entry:
705 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
706 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
707 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
708 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
709 // CHECK1-NEXT:    ret void
710 //
711 //
712 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
713 // CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
714 // CHECK1-NEXT:  entry:
715 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
716 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
717 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
718 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
719 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
720 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
721 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
722 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
723 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
724 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
725 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
726 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
727 // CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 2
728 // CHECK1-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
729 // CHECK1-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 1024)
730 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
731 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
732 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV3]], align 4
733 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
734 // CHECK1-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV1]], align 2
735 // CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i16*
736 // CHECK1-NEXT:    store i16 [[TMP5]], i16* [[CONV4]], align 2
737 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
738 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]])
739 // CHECK1-NEXT:    ret void
740 //
741 //
742 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
743 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR1]] {
744 // CHECK1-NEXT:  entry:
745 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
746 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
747 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
748 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
749 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
750 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
751 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
752 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
753 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
754 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
755 // CHECK1-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 2
756 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
757 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
758 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
759 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
760 // CHECK1-NEXT:    ret void
761 //
762 //
763 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
764 // CHECK1-SAME: () #[[ATTR5:[0-9]+]] {
765 // CHECK1-NEXT:  entry:
766 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
767 // CHECK1-NEXT:    ret void
768 //
769 //
770 // CHECK3-LABEL: define {{[^@]+}}@_Z3bari
771 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
772 // CHECK3-NEXT:  entry:
773 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
774 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
775 // CHECK3-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
776 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
777 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
778 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
779 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP0]])
780 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
781 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
782 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
783 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
784 // CHECK3-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP2]])
785 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
786 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
787 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
788 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
789 // CHECK3-NEXT:    [[CALL3:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP4]])
790 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
791 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
792 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
793 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
794 // CHECK3-NEXT:    ret i32 [[TMP6]]
795 //
796 //
797 // CHECK3-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
798 // CHECK3-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
799 // CHECK3-NEXT:  entry:
800 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
801 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
802 // CHECK3-NEXT:    [[B:%.*]] = alloca i32, align 4
803 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
804 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
805 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
806 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
807 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
808 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
809 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
810 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
811 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
812 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
813 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
814 // CHECK3-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
815 // CHECK3-NEXT:    store i32 1, i32* [[B]], align 4
816 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
817 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
818 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
819 // CHECK3-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
820 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
821 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[B_CASTED]], align 4
822 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
823 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
824 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
825 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
826 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
827 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
828 // CHECK3-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
829 // CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 4
830 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
831 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
832 // CHECK3-NEXT:    store double* [[A]], double** [[TMP9]], align 4
833 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
834 // CHECK3-NEXT:    store i8* null, i8** [[TMP10]], align 4
835 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
836 // CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
837 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP12]], align 4
838 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
839 // CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
840 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP14]], align 4
841 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
842 // CHECK3-NEXT:    store i8* null, i8** [[TMP15]], align 4
843 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
844 // CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
845 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
846 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
847 // CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
848 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
849 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
850 // CHECK3-NEXT:    store i8* null, i8** [[TMP20]], align 4
851 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
852 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
853 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
854 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
855 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
856 // CHECK3-NEXT:    store i32 1, i32* [[TMP24]], align 4
857 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
858 // CHECK3-NEXT:    store i32 3, i32* [[TMP25]], align 4
859 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
860 // CHECK3-NEXT:    store i8** [[TMP21]], i8*** [[TMP26]], align 4
861 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
862 // CHECK3-NEXT:    store i8** [[TMP22]], i8*** [[TMP27]], align 4
863 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
864 // CHECK3-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP28]], align 4
865 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
866 // CHECK3-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP29]], align 4
867 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
868 // CHECK3-NEXT:    store i8** null, i8*** [[TMP30]], align 4
869 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
870 // CHECK3-NEXT:    store i8** null, i8*** [[TMP31]], align 4
871 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
872 // CHECK3-NEXT:    store i64 0, i64* [[TMP32]], align 8
873 // CHECK3-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i32 0, i32 [[TMP23]], i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
874 // CHECK3-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
875 // CHECK3-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
876 // CHECK3:       omp_offload.failed:
877 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR2:[0-9]+]]
878 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
879 // CHECK3:       omp_offload.cont:
880 // CHECK3-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
881 // CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
882 // CHECK3-NEXT:    [[TMP36:%.*]] = bitcast i8** [[TMP35]] to %struct.S1**
883 // CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP36]], align 4
884 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
885 // CHECK3-NEXT:    [[TMP38:%.*]] = bitcast i8** [[TMP37]] to double**
886 // CHECK3-NEXT:    store double* [[A2]], double** [[TMP38]], align 4
887 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
888 // CHECK3-NEXT:    store i8* null, i8** [[TMP39]], align 4
889 // CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
890 // CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
891 // CHECK3-NEXT:    [[KERNEL_ARGS6:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
892 // CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 0
893 // CHECK3-NEXT:    store i32 1, i32* [[TMP42]], align 4
894 // CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 1
895 // CHECK3-NEXT:    store i32 1, i32* [[TMP43]], align 4
896 // CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 2
897 // CHECK3-NEXT:    store i8** [[TMP40]], i8*** [[TMP44]], align 4
898 // CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 3
899 // CHECK3-NEXT:    store i8** [[TMP41]], i8*** [[TMP45]], align 4
900 // CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 4
901 // CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP46]], align 4
902 // CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 5
903 // CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP47]], align 4
904 // CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 6
905 // CHECK3-NEXT:    store i8** null, i8*** [[TMP48]], align 4
906 // CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 7
907 // CHECK3-NEXT:    store i8** null, i8*** [[TMP49]], align 4
908 // CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]], i32 0, i32 8
909 // CHECK3-NEXT:    store i64 0, i64* [[TMP50]], align 8
910 // CHECK3-NEXT:    [[TMP51:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 1024, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS6]])
911 // CHECK3-NEXT:    [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0
912 // CHECK3-NEXT:    br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
913 // CHECK3:       omp_offload.failed7:
914 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR2]]
915 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
916 // CHECK3:       omp_offload.cont8:
917 // CHECK3-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
918 // CHECK3-NEXT:    [[TMP53:%.*]] = load double, double* [[A9]], align 4
919 // CHECK3-NEXT:    [[CONV:%.*]] = fptosi double [[TMP53]] to i32
920 // CHECK3-NEXT:    ret i32 [[CONV]]
921 //
922 //
923 // CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici
924 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
925 // CHECK3-NEXT:  entry:
926 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
927 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
928 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
929 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
930 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
931 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 4
932 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 4
933 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 4
934 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
935 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED4:%.*]] = alloca i32, align 4
936 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [1 x i8*], align 4
937 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [1 x i8*], align 4
938 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [1 x i8*], align 4
939 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
940 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
941 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
942 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
943 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP1]], 32
944 // CHECK3-NEXT:    store i32 [[MUL]], i32* [[DOTCAPTURE_EXPR_1]], align 4
945 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
946 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
947 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
948 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
949 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
950 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
951 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
952 // CHECK3-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
953 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP7]], align 4
954 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
955 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
956 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP9]], align 4
957 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
958 // CHECK3-NEXT:    store i8* null, i8** [[TMP10]], align 4
959 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
960 // CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
961 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP12]], align 4
962 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
963 // CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
964 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP14]], align 4
965 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
966 // CHECK3-NEXT:    store i8* null, i8** [[TMP15]], align 4
967 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
968 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
969 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
970 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
971 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
972 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
973 // CHECK3-NEXT:    store i32 1, i32* [[TMP20]], align 4
974 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
975 // CHECK3-NEXT:    store i32 2, i32* [[TMP21]], align 4
976 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
977 // CHECK3-NEXT:    store i8** [[TMP16]], i8*** [[TMP22]], align 4
978 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
979 // CHECK3-NEXT:    store i8** [[TMP17]], i8*** [[TMP23]], align 4
980 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
981 // CHECK3-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64** [[TMP24]], align 4
982 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
983 // CHECK3-NEXT:    store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i64** [[TMP25]], align 4
984 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
985 // CHECK3-NEXT:    store i8** null, i8*** [[TMP26]], align 4
986 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
987 // CHECK3-NEXT:    store i8** null, i8*** [[TMP27]], align 4
988 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
989 // CHECK3-NEXT:    store i64 0, i64* [[TMP28]], align 8
990 // CHECK3-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 [[TMP18]], i32 [[TMP19]], i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
991 // CHECK3-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
992 // CHECK3-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
993 // CHECK3:       omp_offload.failed:
994 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR2]]
995 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
996 // CHECK3:       omp_offload.cont:
997 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[N_ADDR]], align 4
998 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP31]]
999 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_3]], align 4
1000 // CHECK3-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
1001 // CHECK3-NEXT:    store i32 [[TMP32]], i32* [[DOTCAPTURE_EXPR__CASTED4]], align 4
1002 // CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED4]], align 4
1003 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
1004 // CHECK3-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i32*
1005 // CHECK3-NEXT:    store i32 [[TMP33]], i32* [[TMP35]], align 4
1006 // CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
1007 // CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i32*
1008 // CHECK3-NEXT:    store i32 [[TMP33]], i32* [[TMP37]], align 4
1009 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
1010 // CHECK3-NEXT:    store i8* null, i8** [[TMP38]], align 4
1011 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
1012 // CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
1013 // CHECK3-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
1014 // CHECK3-NEXT:    [[KERNEL_ARGS8:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1015 // CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS8]], i32 0, i32 0
1016 // CHECK3-NEXT:    store i32 1, i32* [[TMP42]], align 4
1017 // CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS8]], i32 0, i32 1
1018 // CHECK3-NEXT:    store i32 1, i32* [[TMP43]], align 4
1019 // CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS8]], i32 0, i32 2
1020 // CHECK3-NEXT:    store i8** [[TMP39]], i8*** [[TMP44]], align 4
1021 // CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS8]], i32 0, i32 3
1022 // CHECK3-NEXT:    store i8** [[TMP40]], i8*** [[TMP45]], align 4
1023 // CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS8]], i32 0, i32 4
1024 // CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP46]], align 4
1025 // CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS8]], i32 0, i32 5
1026 // CHECK3-NEXT:    store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP47]], align 4
1027 // CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS8]], i32 0, i32 6
1028 // CHECK3-NEXT:    store i8** null, i8*** [[TMP48]], align 4
1029 // CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS8]], i32 0, i32 7
1030 // CHECK3-NEXT:    store i8** null, i8*** [[TMP49]], align 4
1031 // CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS8]], i32 0, i32 8
1032 // CHECK3-NEXT:    store i64 0, i64* [[TMP50]], align 8
1033 // CHECK3-NEXT:    [[TMP51:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 [[TMP41]], i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS8]])
1034 // CHECK3-NEXT:    [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0
1035 // CHECK3-NEXT:    br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED9:%.*]], label [[OMP_OFFLOAD_CONT10:%.*]]
1036 // CHECK3:       omp_offload.failed9:
1037 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i32 [[TMP33]]) #[[ATTR2]]
1038 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT10]]
1039 // CHECK3:       omp_offload.cont10:
1040 // CHECK3-NEXT:    [[TMP53:%.*]] = load i32, i32* [[N_ADDR]], align 4
1041 // CHECK3-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP53]], 1
1042 // CHECK3-NEXT:    ret i32 [[ADD11]]
1043 //
1044 //
1045 // CHECK3-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
1046 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
1047 // CHECK3-NEXT:  entry:
1048 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1049 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
1050 // CHECK3-NEXT:    [[B:%.*]] = alloca i16, align 2
1051 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
1052 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
1053 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1054 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
1055 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
1056 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
1057 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
1058 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1059 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
1060 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1061 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1062 // CHECK3-NEXT:    store i32 1, i32* [[TMP0]], align 4
1063 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1064 // CHECK3-NEXT:    store i32 0, i32* [[TMP1]], align 4
1065 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1066 // CHECK3-NEXT:    store i8** null, i8*** [[TMP2]], align 4
1067 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1068 // CHECK3-NEXT:    store i8** null, i8*** [[TMP3]], align 4
1069 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1070 // CHECK3-NEXT:    store i64* null, i64** [[TMP4]], align 4
1071 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1072 // CHECK3-NEXT:    store i64* null, i64** [[TMP5]], align 4
1073 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1074 // CHECK3-NEXT:    store i8** null, i8*** [[TMP6]], align 4
1075 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1076 // CHECK3-NEXT:    store i8** null, i8*** [[TMP7]], align 4
1077 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1078 // CHECK3-NEXT:    store i64 0, i64* [[TMP8]], align 8
1079 // CHECK3-NEXT:    [[TMP9:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 0, i32 20, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1080 // CHECK3-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
1081 // CHECK3-NEXT:    br i1 [[TMP10]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1082 // CHECK3:       omp_offload.failed:
1083 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
1084 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1085 // CHECK3:       omp_offload.cont:
1086 // CHECK3-NEXT:    store i16 1, i16* [[B]], align 2
1087 // CHECK3-NEXT:    [[TMP11:%.*]] = load i16, i16* [[B]], align 2
1088 // CHECK3-NEXT:    store i16 [[TMP11]], i16* [[DOTCAPTURE_EXPR_]], align 2
1089 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[A]], align 4
1090 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[A_CASTED]], align 4
1091 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[A_CASTED]], align 4
1092 // CHECK3-NEXT:    [[TMP14:%.*]] = load i16, i16* [[B]], align 2
1093 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
1094 // CHECK3-NEXT:    store i16 [[TMP14]], i16* [[CONV]], align 2
1095 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[B_CASTED]], align 4
1096 // CHECK3-NEXT:    [[TMP16:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
1097 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i16*
1098 // CHECK3-NEXT:    store i16 [[TMP16]], i16* [[CONV1]], align 2
1099 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
1100 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1101 // CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
1102 // CHECK3-NEXT:    store i32 [[TMP13]], i32* [[TMP19]], align 4
1103 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1104 // CHECK3-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
1105 // CHECK3-NEXT:    store i32 [[TMP13]], i32* [[TMP21]], align 4
1106 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1107 // CHECK3-NEXT:    store i8* null, i8** [[TMP22]], align 4
1108 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1109 // CHECK3-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
1110 // CHECK3-NEXT:    store i32 [[TMP15]], i32* [[TMP24]], align 4
1111 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1112 // CHECK3-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
1113 // CHECK3-NEXT:    store i32 [[TMP15]], i32* [[TMP26]], align 4
1114 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1115 // CHECK3-NEXT:    store i8* null, i8** [[TMP27]], align 4
1116 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1117 // CHECK3-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
1118 // CHECK3-NEXT:    store i32 [[TMP17]], i32* [[TMP29]], align 4
1119 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1120 // CHECK3-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i32*
1121 // CHECK3-NEXT:    store i32 [[TMP17]], i32* [[TMP31]], align 4
1122 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1123 // CHECK3-NEXT:    store i8* null, i8** [[TMP32]], align 4
1124 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1125 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1126 // CHECK3-NEXT:    [[TMP35:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
1127 // CHECK3-NEXT:    [[TMP36:%.*]] = sext i16 [[TMP35]] to i32
1128 // CHECK3-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1129 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 0
1130 // CHECK3-NEXT:    store i32 1, i32* [[TMP37]], align 4
1131 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 1
1132 // CHECK3-NEXT:    store i32 3, i32* [[TMP38]], align 4
1133 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 2
1134 // CHECK3-NEXT:    store i8** [[TMP33]], i8*** [[TMP39]], align 4
1135 // CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 3
1136 // CHECK3-NEXT:    store i8** [[TMP34]], i8*** [[TMP40]], align 4
1137 // CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 4
1138 // CHECK3-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP41]], align 4
1139 // CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 5
1140 // CHECK3-NEXT:    store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP42]], align 4
1141 // CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 6
1142 // CHECK3-NEXT:    store i8** null, i8*** [[TMP43]], align 4
1143 // CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 7
1144 // CHECK3-NEXT:    store i8** null, i8*** [[TMP44]], align 4
1145 // CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]], i32 0, i32 8
1146 // CHECK3-NEXT:    store i64 0, i64* [[TMP45]], align 8
1147 // CHECK3-NEXT:    [[TMP46:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB1]], i64 -1, i32 [[TMP36]], i32 1024, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS2]])
1148 // CHECK3-NEXT:    [[TMP47:%.*]] = icmp ne i32 [[TMP46]], 0
1149 // CHECK3-NEXT:    br i1 [[TMP47]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
1150 // CHECK3:       omp_offload.failed3:
1151 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i32 [[TMP13]], i32 [[TMP15]], i32 [[TMP17]]) #[[ATTR2]]
1152 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
1153 // CHECK3:       omp_offload.cont4:
1154 // CHECK3-NEXT:    [[TMP48:%.*]] = load i32, i32* [[A]], align 4
1155 // CHECK3-NEXT:    ret i32 [[TMP48]]
1156 //
1157 //
1158 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
1159 // CHECK3-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
1160 // CHECK3-NEXT:  entry:
1161 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1162 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1163 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1164 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1165 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1166 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1167 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1168 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1169 // CHECK3-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1170 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1171 // CHECK3-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 [[TMP2]])
1172 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_ADDR]], align 4
1173 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[B_CASTED]], align 4
1174 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_CASTED]], align 4
1175 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP4]])
1176 // CHECK3-NEXT:    ret void
1177 //
1178 //
1179 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
1180 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]]) #[[ATTR1]] {
1181 // CHECK3-NEXT:  entry:
1182 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1183 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1184 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1185 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1186 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1187 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1188 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1189 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1190 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1191 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
1192 // CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to double
1193 // CHECK3-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
1194 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1195 // CHECK3-NEXT:    store double [[ADD]], double* [[A]], align 4
1196 // CHECK3-NEXT:    ret void
1197 //
1198 //
1199 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
1200 // CHECK3-SAME: (%struct.S1* noundef [[THIS:%.*]]) #[[ATTR3:[0-9]+]] {
1201 // CHECK3-NEXT:  entry:
1202 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1203 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1204 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1205 // CHECK3-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1206 // CHECK3-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 1024)
1207 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
1208 // CHECK3-NEXT:    ret void
1209 //
1210 //
1211 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
1212 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]]) #[[ATTR1]] {
1213 // CHECK3-NEXT:  entry:
1214 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1215 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1216 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1217 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1218 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1219 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1220 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1221 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1222 // CHECK3-NEXT:    store double 2.500000e+00, double* [[A]], align 4
1223 // CHECK3-NEXT:    ret void
1224 //
1225 //
1226 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
1227 // CHECK3-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1]] {
1228 // CHECK3-NEXT:  entry:
1229 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1230 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
1231 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1232 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1233 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
1234 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1235 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
1236 // CHECK3-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
1237 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
1238 // CHECK3-NEXT:    ret void
1239 //
1240 //
1241 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
1242 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1243 // CHECK3-NEXT:  entry:
1244 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1245 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1246 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1247 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1248 // CHECK3-NEXT:    ret void
1249 //
1250 //
1251 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
1252 // CHECK3-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1253 // CHECK3-NEXT:  entry:
1254 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1255 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1256 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1257 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1258 // CHECK3-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 [[TMP1]])
1259 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..7 to void (i32*, i32*, ...)*))
1260 // CHECK3-NEXT:    ret void
1261 //
1262 //
1263 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
1264 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1265 // CHECK3-NEXT:  entry:
1266 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1267 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1268 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1269 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1270 // CHECK3-NEXT:    ret void
1271 //
1272 //
1273 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
1274 // CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
1275 // CHECK3-NEXT:  entry:
1276 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1277 // CHECK3-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 20)
1278 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*))
1279 // CHECK3-NEXT:    ret void
1280 //
1281 //
1282 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..10
1283 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1284 // CHECK3-NEXT:  entry:
1285 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1286 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1287 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1288 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1289 // CHECK3-NEXT:    ret void
1290 //
1291 //
1292 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
1293 // CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
1294 // CHECK3-NEXT:  entry:
1295 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1296 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1297 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1298 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
1299 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1300 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1301 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
1302 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1303 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1304 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1305 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
1306 // CHECK3-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 2
1307 // CHECK3-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
1308 // CHECK3-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 1024)
1309 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
1310 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
1311 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
1312 // CHECK3-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
1313 // CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[B_CASTED]] to i16*
1314 // CHECK3-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
1315 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
1316 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]])
1317 // CHECK3-NEXT:    ret void
1318 //
1319 //
1320 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
1321 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR1]] {
1322 // CHECK3-NEXT:  entry:
1323 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1324 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1325 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1326 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1327 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1328 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1329 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
1330 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1331 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1332 // CHECK3-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
1333 // CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
1334 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
1335 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
1336 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
1337 // CHECK3-NEXT:    ret void
1338 //
1339 //
1340 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1341 // CHECK3-SAME: () #[[ATTR5:[0-9]+]] {
1342 // CHECK3-NEXT:  entry:
1343 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
1344 // CHECK3-NEXT:    ret void
1345 //
1346 //
1347 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
1348 // CHECK9-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
1349 // CHECK9-NEXT:  entry:
1350 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1351 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
1352 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
1353 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1354 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
1355 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1356 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
1357 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1358 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV3]], align 4
1359 // CHECK9-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
1360 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
1361 // CHECK9-NEXT:    ret void
1362 //
1363 //
1364 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
1365 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
1366 // CHECK9-NEXT:  entry:
1367 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1368 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1369 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1370 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1371 // CHECK9-NEXT:    ret void
1372 //
1373 //
1374 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
1375 // CHECK9-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1376 // CHECK9-NEXT:  entry:
1377 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1378 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1379 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1380 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1381 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1382 // CHECK9-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 [[TMP1]])
1383 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
1384 // CHECK9-NEXT:    ret void
1385 //
1386 //
1387 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
1388 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
1389 // CHECK9-NEXT:  entry:
1390 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1391 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1392 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1393 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1394 // CHECK9-NEXT:    ret void
1395 //
1396 //
1397 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
1398 // CHECK9-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1399 // CHECK9-NEXT:  entry:
1400 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1401 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1402 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1403 // CHECK9-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
1404 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1405 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1406 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1407 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1408 // CHECK9-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1409 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1410 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1411 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 4
1412 // CHECK9-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 [[TMP2]])
1413 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
1414 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[B_CASTED]] to i32*
1415 // CHECK9-NEXT:    store i32 [[TMP3]], i32* [[CONV2]], align 4
1416 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[B_CASTED]], align 8
1417 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP4]])
1418 // CHECK9-NEXT:    ret void
1419 //
1420 //
1421 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
1422 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
1423 // CHECK9-NEXT:  entry:
1424 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1425 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1426 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1427 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1428 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1429 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1430 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1431 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1432 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1433 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1434 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1435 // CHECK9-NEXT:    [[CONV1:%.*]] = sitofp i32 [[TMP1]] to double
1436 // CHECK9-NEXT:    [[ADD:%.*]] = fadd double [[CONV1]], 1.500000e+00
1437 // CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1438 // CHECK9-NEXT:    store double [[ADD]], double* [[A]], align 8
1439 // CHECK9-NEXT:    ret void
1440 //
1441 //
1442 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
1443 // CHECK9-SAME: (%struct.S1* noundef [[THIS:%.*]]) #[[ATTR2:[0-9]+]] {
1444 // CHECK9-NEXT:  entry:
1445 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1446 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1447 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1448 // CHECK9-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1449 // CHECK9-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 1024)
1450 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
1451 // CHECK9-NEXT:    ret void
1452 //
1453 //
1454 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
1455 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]]) #[[ATTR0]] {
1456 // CHECK9-NEXT:  entry:
1457 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1458 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1459 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1460 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1461 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1462 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1463 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1464 // CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1465 // CHECK9-NEXT:    store double 2.500000e+00, double* [[A]], align 8
1466 // CHECK9-NEXT:    ret void
1467 //
1468 //
1469 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
1470 // CHECK9-SAME: () #[[ATTR3:[0-9]+]] {
1471 // CHECK9-NEXT:  entry:
1472 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1473 // CHECK9-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 20)
1474 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
1475 // CHECK9-NEXT:    ret void
1476 //
1477 //
1478 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
1479 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
1480 // CHECK9-NEXT:  entry:
1481 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1482 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1483 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1484 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1485 // CHECK9-NEXT:    ret void
1486 //
1487 //
1488 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
1489 // CHECK9-SAME: (i64 noundef [[A:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
1490 // CHECK9-NEXT:  entry:
1491 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1492 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1493 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1494 // CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1495 // CHECK9-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
1496 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1497 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1498 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1499 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1500 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1501 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
1502 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
1503 // CHECK9-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 2
1504 // CHECK9-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
1505 // CHECK9-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 1024)
1506 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
1507 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1508 // CHECK9-NEXT:    store i32 [[TMP3]], i32* [[CONV3]], align 4
1509 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
1510 // CHECK9-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV1]], align 2
1511 // CHECK9-NEXT:    [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i16*
1512 // CHECK9-NEXT:    store i16 [[TMP5]], i16* [[CONV4]], align 2
1513 // CHECK9-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
1514 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]])
1515 // CHECK9-NEXT:    ret void
1516 //
1517 //
1518 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
1519 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
1520 // CHECK9-NEXT:  entry:
1521 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1522 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1523 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1524 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1525 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1526 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1527 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1528 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1529 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1530 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
1531 // CHECK9-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV1]], align 2
1532 // CHECK9-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP0]] to i32
1533 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1534 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV2]]
1535 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
1536 // CHECK9-NEXT:    ret void
1537 //
1538 //
1539 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
1540 // CHECK11-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
1541 // CHECK11-NEXT:  entry:
1542 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1543 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
1544 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
1545 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1546 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
1547 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1548 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
1549 // CHECK11-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
1550 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
1551 // CHECK11-NEXT:    ret void
1552 //
1553 //
1554 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
1555 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
1556 // CHECK11-NEXT:  entry:
1557 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1558 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1559 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1560 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1561 // CHECK11-NEXT:    ret void
1562 //
1563 //
1564 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
1565 // CHECK11-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1566 // CHECK11-NEXT:  entry:
1567 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1568 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1569 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1570 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1571 // CHECK11-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 [[TMP1]])
1572 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
1573 // CHECK11-NEXT:    ret void
1574 //
1575 //
1576 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
1577 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
1578 // CHECK11-NEXT:  entry:
1579 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1580 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1581 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1582 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1583 // CHECK11-NEXT:    ret void
1584 //
1585 //
1586 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
1587 // CHECK11-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
1588 // CHECK11-NEXT:  entry:
1589 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1590 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1591 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1592 // CHECK11-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1593 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1594 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1595 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1596 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1597 // CHECK11-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1598 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1599 // CHECK11-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 [[TMP2]])
1600 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_ADDR]], align 4
1601 // CHECK11-NEXT:    store i32 [[TMP3]], i32* [[B_CASTED]], align 4
1602 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_CASTED]], align 4
1603 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP4]])
1604 // CHECK11-NEXT:    ret void
1605 //
1606 //
1607 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
1608 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]]) #[[ATTR0]] {
1609 // CHECK11-NEXT:  entry:
1610 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1611 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1612 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1613 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1614 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1615 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1616 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1617 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1618 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1619 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B_ADDR]], align 4
1620 // CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to double
1621 // CHECK11-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
1622 // CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1623 // CHECK11-NEXT:    store double [[ADD]], double* [[A]], align 4
1624 // CHECK11-NEXT:    ret void
1625 //
1626 //
1627 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
1628 // CHECK11-SAME: (%struct.S1* noundef [[THIS:%.*]]) #[[ATTR2:[0-9]+]] {
1629 // CHECK11-NEXT:  entry:
1630 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1631 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1632 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1633 // CHECK11-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1634 // CHECK11-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 1024)
1635 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]])
1636 // CHECK11-NEXT:    ret void
1637 //
1638 //
1639 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
1640 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]]) #[[ATTR0]] {
1641 // CHECK11-NEXT:  entry:
1642 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1643 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1644 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1645 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1646 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1647 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1648 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1649 // CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1650 // CHECK11-NEXT:    store double 2.500000e+00, double* [[A]], align 4
1651 // CHECK11-NEXT:    ret void
1652 //
1653 //
1654 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
1655 // CHECK11-SAME: () #[[ATTR3:[0-9]+]] {
1656 // CHECK11-NEXT:  entry:
1657 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1658 // CHECK11-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 0, i32 20)
1659 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
1660 // CHECK11-NEXT:    ret void
1661 //
1662 //
1663 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
1664 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
1665 // CHECK11-NEXT:  entry:
1666 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1667 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1668 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1669 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1670 // CHECK11-NEXT:    ret void
1671 //
1672 //
1673 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
1674 // CHECK11-SAME: (i32 noundef [[A:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
1675 // CHECK11-NEXT:  entry:
1676 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1677 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1678 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1679 // CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
1680 // CHECK11-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1681 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1682 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
1683 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1684 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1685 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1686 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
1687 // CHECK11-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 2
1688 // CHECK11-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
1689 // CHECK11-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 1024)
1690 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
1691 // CHECK11-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
1692 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
1693 // CHECK11-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
1694 // CHECK11-NEXT:    [[CONV2:%.*]] = bitcast i32* [[B_CASTED]] to i16*
1695 // CHECK11-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
1696 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
1697 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]])
1698 // CHECK11-NEXT:    ret void
1699 //
1700 //
1701 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
1702 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR0]] {
1703 // CHECK11-NEXT:  entry:
1704 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1705 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1706 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1707 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1708 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1709 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1710 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
1711 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1712 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1713 // CHECK11-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
1714 // CHECK11-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
1715 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
1716 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV1]]
1717 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
1718 // CHECK11-NEXT:    ret void
1719 //
1720