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 -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 -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 -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=CHECK2
6 // RUN: %clang_cc1 -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 -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 -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=CHECK4
9 
10 // RUN: %clang_cc1 -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 -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 -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 -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 -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 -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 -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 -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 -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 -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=CHECK10
22 // RUN: %clang_cc1 -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 -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 -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 -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=CHECK12
26 
27 // RUN: %clang_cc1 -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 -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 -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 -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 -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 -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 -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 -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 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK17
38 // RUN: %clang_cc1 -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 -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=CHECK18
40 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK19
41 // RUN: %clang_cc1 -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 -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=CHECK20
43 
44 // RUN: %clang_cc1 -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 -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 -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 -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 -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 -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 -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 -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=CHECK25
54 // RUN: %clang_cc1 -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 -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=CHECK26
56 // RUN: %clang_cc1 -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 -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=CHECK27
58 // RUN: %clang_cc1 -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 -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=CHECK28
60 
61 // RUN: %clang_cc1 -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 -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 -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 -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 -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 -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 -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 -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>
85 tx ftemplate(int n) {
86   tx a = 0;
87 
88   #pragma omp target parallel num_threads(tx(20))
89   {
90   }
91 
92   short b = 1;
93   #pragma omp target parallel num_threads(b)
94   {
95     a += b;
96   }
97 
98   return a;
99 }
100 
101 static
102 int fstatic(int n) {
103 
104   #pragma omp target parallel num_threads(n)
105   {
106   }
107 
108   #pragma omp target parallel num_threads(32+n)
109   {
110   }
111 
112   return n+1;
113 }
114 
115 struct S1 {
116   double a;
117 
118   int r1(int n){
119     int b = 1;
120 
121     #pragma omp target parallel num_threads(n-b)
122     {
123       this->a = (double)b + 1.5;
124     }
125 
126     #pragma omp target parallel num_threads(1024)
127     {
128       this->a = 2.5;
129     }
130 
131     return (int)a;
132   }
133 };
134 
135 int bar(int n){
136   int a = 0;
137 
138   S1 S;
139   a += S.r1(n);
140 
141   a += fstatic(n);
142 
143   a += ftemplate<int>(n);
144 
145   return a;
146 }
147 
148 
149 
150 
151 
152 
153 
154 
155 
156 
157 
158 
159 
160 
161 
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 
177 
178 
179 
180 
181 
182 
183 
184 
185 
186 
187 
188 
189 
190 
191 #endif
192 // CHECK1-LABEL: define {{[^@]+}}@_Z3bari
193 // CHECK1-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
194 // CHECK1-NEXT:  entry:
195 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
196 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
197 // CHECK1-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
198 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
199 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
200 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
201 // CHECK1-NEXT:    [[CALL:%.*]] = call signext i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], i32 signext [[TMP0]])
202 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
203 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
204 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
205 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
206 // CHECK1-NEXT:    [[CALL1:%.*]] = call signext i32 @_ZL7fstatici(i32 signext [[TMP2]])
207 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
208 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
209 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
210 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
211 // CHECK1-NEXT:    [[CALL3:%.*]] = call signext i32 @_Z9ftemplateIiET_i(i32 signext [[TMP4]])
212 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
213 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
214 // CHECK1-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
215 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
216 // CHECK1-NEXT:    ret i32 [[TMP6]]
217 //
218 //
219 // CHECK1-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
220 // CHECK1-SAME: (%struct.S1* nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
221 // CHECK1-NEXT:  entry:
222 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
223 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
224 // CHECK1-NEXT:    [[B:%.*]] = alloca i32, align 4
225 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
226 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
227 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
228 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
229 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
230 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
231 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
232 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
233 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
234 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
235 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
236 // CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
237 // CHECK1-NEXT:    store i32 1, i32* [[B]], align 4
238 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
239 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
240 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
241 // CHECK1-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
242 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
243 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
244 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
245 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
246 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
247 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
248 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV2]], align 4
249 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
250 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
251 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
252 // CHECK1-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
253 // CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 8
254 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
255 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
256 // CHECK1-NEXT:    store double* [[A]], double** [[TMP9]], align 8
257 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
258 // CHECK1-NEXT:    store i8* null, i8** [[TMP10]], align 8
259 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
260 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
261 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP12]], align 8
262 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
263 // CHECK1-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
264 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP14]], align 8
265 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
266 // CHECK1-NEXT:    store i8* null, i8** [[TMP15]], align 8
267 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
268 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
269 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP17]], align 8
270 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
271 // CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
272 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP19]], align 8
273 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
274 // CHECK1-NEXT:    store i8* null, i8** [[TMP20]], align 8
275 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
276 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
277 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
278 // CHECK1-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
279 // CHECK1-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
280 // CHECK1-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
281 // CHECK1:       omp_offload.failed:
282 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR2:[0-9]+]]
283 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
284 // CHECK1:       omp_offload.cont:
285 // CHECK1-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
286 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
287 // CHECK1-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
288 // CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 8
289 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
290 // CHECK1-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
291 // CHECK1-NEXT:    store double* [[A3]], double** [[TMP29]], align 8
292 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
293 // CHECK1-NEXT:    store i8* null, i8** [[TMP30]], align 8
294 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
295 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
296 // CHECK1-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
297 // CHECK1-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
298 // CHECK1-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
299 // CHECK1:       omp_offload.failed7:
300 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR2]]
301 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
302 // CHECK1:       omp_offload.cont8:
303 // CHECK1-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
304 // CHECK1-NEXT:    [[TMP35:%.*]] = load double, double* [[A9]], align 8
305 // CHECK1-NEXT:    [[CONV10:%.*]] = fptosi double [[TMP35]] to i32
306 // CHECK1-NEXT:    ret i32 [[CONV10]]
307 //
308 //
309 // CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici
310 // CHECK1-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
311 // CHECK1-NEXT:  entry:
312 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
313 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
314 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
315 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
316 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
317 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
318 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
319 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8
320 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
321 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
322 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
323 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
324 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
325 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
326 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
327 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
328 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV]], align 4
329 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
330 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
331 // CHECK1-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i64*
332 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP4]], align 8
333 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
334 // CHECK1-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
335 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP6]], align 8
336 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
337 // CHECK1-NEXT:    store i8* null, i8** [[TMP7]], align 8
338 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
339 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
340 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
341 // CHECK1-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
342 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
343 // CHECK1-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
344 // CHECK1:       omp_offload.failed:
345 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i64 [[TMP2]]) #[[ATTR2]]
346 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
347 // CHECK1:       omp_offload.cont:
348 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
349 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
350 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
351 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
352 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED2]] to i32*
353 // CHECK1-NEXT:    store i32 [[TMP14]], i32* [[CONV3]], align 4
354 // CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED2]], align 8
355 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
356 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
357 // CHECK1-NEXT:    store i64 [[TMP15]], i64* [[TMP17]], align 8
358 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
359 // CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
360 // CHECK1-NEXT:    store i64 [[TMP15]], i64* [[TMP19]], align 8
361 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
362 // CHECK1-NEXT:    store i8* null, i8** [[TMP20]], align 8
363 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
364 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
365 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
366 // CHECK1-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
367 // CHECK1-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
368 // CHECK1-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
369 // CHECK1:       omp_offload.failed7:
370 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i64 [[TMP15]]) #[[ATTR2]]
371 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
372 // CHECK1:       omp_offload.cont8:
373 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
374 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP26]], 1
375 // CHECK1-NEXT:    ret i32 [[ADD9]]
376 //
377 //
378 // CHECK1-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
379 // CHECK1-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] comdat {
380 // CHECK1-NEXT:  entry:
381 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
382 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
383 // CHECK1-NEXT:    [[B:%.*]] = alloca i16, align 2
384 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
385 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
386 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
387 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
388 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
389 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
390 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
391 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
392 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
393 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
394 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
395 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
396 // CHECK1:       omp_offload.failed:
397 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
398 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
399 // CHECK1:       omp_offload.cont:
400 // CHECK1-NEXT:    store i16 1, i16* [[B]], align 2
401 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
402 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
403 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
404 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
405 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
406 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
407 // CHECK1-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
408 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_CASTED]] to i16*
409 // CHECK1-NEXT:    store i16 [[TMP5]], i16* [[CONV1]], align 2
410 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
411 // CHECK1-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
412 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i16*
413 // CHECK1-NEXT:    store i16 [[TMP7]], i16* [[CONV2]], align 2
414 // CHECK1-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
415 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
416 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
417 // CHECK1-NEXT:    store i64 [[TMP4]], i64* [[TMP10]], align 8
418 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
419 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
420 // CHECK1-NEXT:    store i64 [[TMP4]], i64* [[TMP12]], align 8
421 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
422 // CHECK1-NEXT:    store i8* null, i8** [[TMP13]], align 8
423 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
424 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
425 // CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP15]], align 8
426 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
427 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
428 // CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
429 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
430 // CHECK1-NEXT:    store i8* null, i8** [[TMP18]], align 8
431 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
432 // CHECK1-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
433 // CHECK1-NEXT:    store i64 [[TMP8]], i64* [[TMP20]], align 8
434 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
435 // CHECK1-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
436 // CHECK1-NEXT:    store i64 [[TMP8]], i64* [[TMP22]], align 8
437 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
438 // CHECK1-NEXT:    store i8* null, i8** [[TMP23]], align 8
439 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
440 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
441 // CHECK1-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
442 // CHECK1-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
443 // CHECK1-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
444 // CHECK1-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
445 // CHECK1-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
446 // CHECK1:       omp_offload.failed3:
447 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR2]]
448 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
449 // CHECK1:       omp_offload.cont4:
450 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
451 // CHECK1-NEXT:    ret i32 [[TMP30]]
452 //
453 //
454 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
455 // CHECK1-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
456 // CHECK1-NEXT:  entry:
457 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
458 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
459 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
460 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
461 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
462 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
463 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
464 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
465 // CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
466 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
467 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
468 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
469 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
470 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
471 // CHECK1-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 8
472 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
473 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 8
474 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 8
475 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
476 // CHECK1-NEXT:    ret void
477 //
478 //
479 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
480 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
481 // CHECK1-NEXT:  entry:
482 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
483 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
484 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
485 // CHECK1-NEXT:    [[B:%.*]] = alloca i32, align 4
486 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
487 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
488 // CHECK1-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
489 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
490 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
491 // CHECK1-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
492 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
493 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
494 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
495 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
496 // CHECK1-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
497 // CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
498 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
499 // CHECK1-NEXT:    store double [[ADD]], double* [[A]], align 8
500 // CHECK1-NEXT:    ret void
501 //
502 //
503 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
504 // CHECK1-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
505 // CHECK1-NEXT:  entry:
506 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
507 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
508 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
509 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
510 // CHECK1-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
511 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
512 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
513 // CHECK1-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 8
514 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
515 // CHECK1-NEXT:    ret void
516 //
517 //
518 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
519 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
520 // CHECK1-NEXT:  entry:
521 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
522 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
523 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
524 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
525 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
526 // CHECK1-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
527 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
528 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
529 // CHECK1-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
530 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
531 // CHECK1-NEXT:    store double 2.500000e+00, double* [[A]], align 8
532 // CHECK1-NEXT:    ret void
533 //
534 //
535 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
536 // CHECK1-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
537 // CHECK1-NEXT:  entry:
538 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
539 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 1
540 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
541 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
542 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
543 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
544 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
545 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
546 // CHECK1-NEXT:    ret void
547 //
548 //
549 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
550 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
551 // CHECK1-NEXT:  entry:
552 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
553 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
554 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
555 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
556 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
557 // CHECK1-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
558 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
559 // CHECK1-NEXT:    ret void
560 //
561 //
562 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
563 // CHECK1-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
564 // CHECK1-NEXT:  entry:
565 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
566 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 1
567 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
568 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
569 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
570 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
571 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
572 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
573 // CHECK1-NEXT:    ret void
574 //
575 //
576 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
577 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
578 // CHECK1-NEXT:  entry:
579 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
580 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
581 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
582 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
583 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
584 // CHECK1-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
585 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
586 // CHECK1-NEXT:    ret void
587 //
588 //
589 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
590 // CHECK1-SAME: () #[[ATTR1]] {
591 // CHECK1-NEXT:  entry:
592 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
593 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
594 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
595 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
596 // CHECK1-NEXT:    ret void
597 //
598 //
599 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..10
600 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
601 // CHECK1-NEXT:  entry:
602 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
603 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
604 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
605 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
606 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
607 // CHECK1-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
608 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
609 // CHECK1-NEXT:    ret void
610 //
611 //
612 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
613 // CHECK1-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
614 // CHECK1-NEXT:  entry:
615 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
616 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
617 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
618 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
619 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
620 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
621 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
622 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
623 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
624 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
625 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
626 // CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
627 // CHECK1-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
628 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
629 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
630 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
631 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
632 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
633 // CHECK1-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV1]], align 8
634 // CHECK1-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
635 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
636 // CHECK1-NEXT:    ret void
637 //
638 //
639 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
640 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
641 // CHECK1-NEXT:  entry:
642 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
643 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
644 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
645 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
646 // CHECK1-NEXT:    [[B:%.*]] = alloca i16, align 2
647 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
648 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
649 // CHECK1-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
650 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
651 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
652 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
653 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
654 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
655 // CHECK1-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
656 // CHECK1-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
657 // CHECK1-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
658 // CHECK1-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
659 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
660 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
661 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
662 // CHECK1-NEXT:    ret void
663 //
664 //
665 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
666 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
667 // CHECK1-NEXT:  entry:
668 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
669 // CHECK1-NEXT:    ret void
670 //
671 //
672 // CHECK2-LABEL: define {{[^@]+}}@_Z3bari
673 // CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
674 // CHECK2-NEXT:  entry:
675 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
676 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
677 // CHECK2-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
678 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
679 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
680 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
681 // CHECK2-NEXT:    [[CALL:%.*]] = call signext i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], i32 signext [[TMP0]])
682 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
683 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
684 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
685 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
686 // CHECK2-NEXT:    [[CALL1:%.*]] = call signext i32 @_ZL7fstatici(i32 signext [[TMP2]])
687 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
688 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
689 // CHECK2-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
690 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
691 // CHECK2-NEXT:    [[CALL3:%.*]] = call signext i32 @_Z9ftemplateIiET_i(i32 signext [[TMP4]])
692 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
693 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
694 // CHECK2-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
695 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
696 // CHECK2-NEXT:    ret i32 [[TMP6]]
697 //
698 //
699 // CHECK2-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
700 // CHECK2-SAME: (%struct.S1* nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
701 // CHECK2-NEXT:  entry:
702 // CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
703 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
704 // CHECK2-NEXT:    [[B:%.*]] = alloca i32, align 4
705 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
706 // CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
707 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
708 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
709 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
710 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
711 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
712 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
713 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
714 // CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
715 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
716 // CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
717 // CHECK2-NEXT:    store i32 1, i32* [[B]], align 4
718 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
719 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
720 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
721 // CHECK2-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
722 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
723 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
724 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
725 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
726 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
727 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
728 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[CONV2]], align 4
729 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
730 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
731 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
732 // CHECK2-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
733 // CHECK2-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 8
734 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
735 // CHECK2-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
736 // CHECK2-NEXT:    store double* [[A]], double** [[TMP9]], align 8
737 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
738 // CHECK2-NEXT:    store i8* null, i8** [[TMP10]], align 8
739 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
740 // CHECK2-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
741 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP12]], align 8
742 // CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
743 // CHECK2-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
744 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP14]], align 8
745 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
746 // CHECK2-NEXT:    store i8* null, i8** [[TMP15]], align 8
747 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
748 // CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
749 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP17]], align 8
750 // CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
751 // CHECK2-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
752 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP19]], align 8
753 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
754 // CHECK2-NEXT:    store i8* null, i8** [[TMP20]], align 8
755 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
756 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
757 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
758 // CHECK2-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
759 // CHECK2-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
760 // CHECK2-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
761 // CHECK2:       omp_offload.failed:
762 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR2:[0-9]+]]
763 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
764 // CHECK2:       omp_offload.cont:
765 // CHECK2-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
766 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
767 // CHECK2-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
768 // CHECK2-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 8
769 // CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
770 // CHECK2-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
771 // CHECK2-NEXT:    store double* [[A3]], double** [[TMP29]], align 8
772 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
773 // CHECK2-NEXT:    store i8* null, i8** [[TMP30]], align 8
774 // CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
775 // CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
776 // CHECK2-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
777 // CHECK2-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
778 // CHECK2-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
779 // CHECK2:       omp_offload.failed7:
780 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR2]]
781 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
782 // CHECK2:       omp_offload.cont8:
783 // CHECK2-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
784 // CHECK2-NEXT:    [[TMP35:%.*]] = load double, double* [[A9]], align 8
785 // CHECK2-NEXT:    [[CONV10:%.*]] = fptosi double [[TMP35]] to i32
786 // CHECK2-NEXT:    ret i32 [[CONV10]]
787 //
788 //
789 // CHECK2-LABEL: define {{[^@]+}}@_ZL7fstatici
790 // CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
791 // CHECK2-NEXT:  entry:
792 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
793 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
794 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
795 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
796 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
797 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
798 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
799 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8
800 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
801 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
802 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
803 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
804 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
805 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
806 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
807 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
808 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[CONV]], align 4
809 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
810 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
811 // CHECK2-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i64*
812 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP4]], align 8
813 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
814 // CHECK2-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
815 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP6]], align 8
816 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
817 // CHECK2-NEXT:    store i8* null, i8** [[TMP7]], align 8
818 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
819 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
820 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
821 // CHECK2-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
822 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
823 // CHECK2-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
824 // CHECK2:       omp_offload.failed:
825 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i64 [[TMP2]]) #[[ATTR2]]
826 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
827 // CHECK2:       omp_offload.cont:
828 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
829 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
830 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
831 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
832 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED2]] to i32*
833 // CHECK2-NEXT:    store i32 [[TMP14]], i32* [[CONV3]], align 4
834 // CHECK2-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED2]], align 8
835 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
836 // CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
837 // CHECK2-NEXT:    store i64 [[TMP15]], i64* [[TMP17]], align 8
838 // CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
839 // CHECK2-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
840 // CHECK2-NEXT:    store i64 [[TMP15]], i64* [[TMP19]], align 8
841 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
842 // CHECK2-NEXT:    store i8* null, i8** [[TMP20]], align 8
843 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
844 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
845 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
846 // CHECK2-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
847 // CHECK2-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
848 // CHECK2-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
849 // CHECK2:       omp_offload.failed7:
850 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i64 [[TMP15]]) #[[ATTR2]]
851 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
852 // CHECK2:       omp_offload.cont8:
853 // CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
854 // CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP26]], 1
855 // CHECK2-NEXT:    ret i32 [[ADD9]]
856 //
857 //
858 // CHECK2-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
859 // CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] comdat {
860 // CHECK2-NEXT:  entry:
861 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
862 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
863 // CHECK2-NEXT:    [[B:%.*]] = alloca i16, align 2
864 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
865 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
866 // CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
867 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
868 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
869 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
870 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
871 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
872 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
873 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
874 // CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
875 // CHECK2-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
876 // CHECK2:       omp_offload.failed:
877 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
878 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
879 // CHECK2:       omp_offload.cont:
880 // CHECK2-NEXT:    store i16 1, i16* [[B]], align 2
881 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
882 // CHECK2-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
883 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
884 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
885 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
886 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
887 // CHECK2-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
888 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_CASTED]] to i16*
889 // CHECK2-NEXT:    store i16 [[TMP5]], i16* [[CONV1]], align 2
890 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
891 // CHECK2-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
892 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i16*
893 // CHECK2-NEXT:    store i16 [[TMP7]], i16* [[CONV2]], align 2
894 // CHECK2-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
895 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
896 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
897 // CHECK2-NEXT:    store i64 [[TMP4]], i64* [[TMP10]], align 8
898 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
899 // CHECK2-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
900 // CHECK2-NEXT:    store i64 [[TMP4]], i64* [[TMP12]], align 8
901 // CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
902 // CHECK2-NEXT:    store i8* null, i8** [[TMP13]], align 8
903 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
904 // CHECK2-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
905 // CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP15]], align 8
906 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
907 // CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
908 // CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
909 // CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
910 // CHECK2-NEXT:    store i8* null, i8** [[TMP18]], align 8
911 // CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
912 // CHECK2-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
913 // CHECK2-NEXT:    store i64 [[TMP8]], i64* [[TMP20]], align 8
914 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
915 // CHECK2-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
916 // CHECK2-NEXT:    store i64 [[TMP8]], i64* [[TMP22]], align 8
917 // CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
918 // CHECK2-NEXT:    store i8* null, i8** [[TMP23]], align 8
919 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
920 // CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
921 // CHECK2-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
922 // CHECK2-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
923 // CHECK2-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
924 // CHECK2-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
925 // CHECK2-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
926 // CHECK2:       omp_offload.failed3:
927 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR2]]
928 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
929 // CHECK2:       omp_offload.cont4:
930 // CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
931 // CHECK2-NEXT:    ret i32 [[TMP30]]
932 //
933 //
934 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
935 // CHECK2-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
936 // CHECK2-NEXT:  entry:
937 // CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
938 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
939 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
940 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
941 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
942 // CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
943 // CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
944 // CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
945 // CHECK2-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
946 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
947 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
948 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
949 // CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
950 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
951 // CHECK2-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 8
952 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
953 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 8
954 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 8
955 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
956 // CHECK2-NEXT:    ret void
957 //
958 //
959 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
960 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
961 // CHECK2-NEXT:  entry:
962 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
963 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
964 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
965 // CHECK2-NEXT:    [[B:%.*]] = alloca i32, align 4
966 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
967 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
968 // CHECK2-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
969 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
970 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
971 // CHECK2-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
972 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
973 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
974 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
975 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
976 // CHECK2-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
977 // CHECK2-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
978 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
979 // CHECK2-NEXT:    store double [[ADD]], double* [[A]], align 8
980 // CHECK2-NEXT:    ret void
981 //
982 //
983 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
984 // CHECK2-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
985 // CHECK2-NEXT:  entry:
986 // CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
987 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
988 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
989 // CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
990 // CHECK2-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
991 // CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
992 // CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
993 // CHECK2-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 8
994 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
995 // CHECK2-NEXT:    ret void
996 //
997 //
998 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
999 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1000 // CHECK2-NEXT:  entry:
1001 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1002 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1003 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
1004 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1005 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1006 // CHECK2-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
1007 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
1008 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
1009 // CHECK2-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
1010 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
1011 // CHECK2-NEXT:    store double 2.500000e+00, double* [[A]], align 8
1012 // CHECK2-NEXT:    ret void
1013 //
1014 //
1015 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
1016 // CHECK2-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1017 // CHECK2-NEXT:  entry:
1018 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1019 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 1
1020 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1021 // CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1022 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1023 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
1024 // CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
1025 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
1026 // CHECK2-NEXT:    ret void
1027 //
1028 //
1029 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
1030 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1031 // CHECK2-NEXT:  entry:
1032 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1033 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1034 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
1035 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1036 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1037 // CHECK2-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
1038 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
1039 // CHECK2-NEXT:    ret void
1040 //
1041 //
1042 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
1043 // CHECK2-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1044 // CHECK2-NEXT:  entry:
1045 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1046 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 1
1047 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1048 // CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1049 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1050 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
1051 // CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
1052 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
1053 // CHECK2-NEXT:    ret void
1054 //
1055 //
1056 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..7
1057 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1058 // CHECK2-NEXT:  entry:
1059 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1060 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1061 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
1062 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1063 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1064 // CHECK2-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
1065 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
1066 // CHECK2-NEXT:    ret void
1067 //
1068 //
1069 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
1070 // CHECK2-SAME: () #[[ATTR1]] {
1071 // CHECK2-NEXT:  entry:
1072 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
1073 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1074 // CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
1075 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
1076 // CHECK2-NEXT:    ret void
1077 //
1078 //
1079 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..10
1080 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1081 // CHECK2-NEXT:  entry:
1082 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1083 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1084 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
1085 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1086 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1087 // CHECK2-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
1088 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
1089 // CHECK2-NEXT:    ret void
1090 //
1091 //
1092 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
1093 // CHECK2-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1094 // CHECK2-NEXT:  entry:
1095 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1096 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1097 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1098 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
1099 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1100 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1101 // CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1102 // CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1103 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1104 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
1105 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
1106 // CHECK2-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
1107 // CHECK2-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
1108 // CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
1109 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1110 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
1111 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
1112 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1113 // CHECK2-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV1]], align 8
1114 // CHECK2-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
1115 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
1116 // CHECK2-NEXT:    ret void
1117 //
1118 //
1119 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..11
1120 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1121 // CHECK2-NEXT:  entry:
1122 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1123 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1124 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
1125 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
1126 // CHECK2-NEXT:    [[B:%.*]] = alloca i16, align 2
1127 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1128 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1129 // CHECK2-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
1130 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
1131 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
1132 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1133 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
1134 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
1135 // CHECK2-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
1136 // CHECK2-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
1137 // CHECK2-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
1138 // CHECK2-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
1139 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
1140 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
1141 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1142 // CHECK2-NEXT:    ret void
1143 //
1144 //
1145 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1146 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] {
1147 // CHECK2-NEXT:  entry:
1148 // CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
1149 // CHECK2-NEXT:    ret void
1150 //
1151 //
1152 // CHECK3-LABEL: define {{[^@]+}}@_Z3bari
1153 // CHECK3-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
1154 // CHECK3-NEXT:  entry:
1155 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1156 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
1157 // CHECK3-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
1158 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1159 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
1160 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1161 // CHECK3-NEXT:    [[CALL:%.*]] = call i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], i32 [[TMP0]])
1162 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
1163 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
1164 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1165 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
1166 // CHECK3-NEXT:    [[CALL1:%.*]] = call i32 @_ZL7fstatici(i32 [[TMP2]])
1167 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
1168 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
1169 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
1170 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
1171 // CHECK3-NEXT:    [[CALL3:%.*]] = call i32 @_Z9ftemplateIiET_i(i32 [[TMP4]])
1172 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
1173 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
1174 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
1175 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
1176 // CHECK3-NEXT:    ret i32 [[TMP6]]
1177 //
1178 //
1179 // CHECK3-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
1180 // CHECK3-SAME: (%struct.S1* nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 [[N:%.*]]) #[[ATTR0]] comdat align 2 {
1181 // CHECK3-NEXT:  entry:
1182 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1183 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1184 // CHECK3-NEXT:    [[B:%.*]] = alloca i32, align 4
1185 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1186 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1187 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
1188 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
1189 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
1190 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
1191 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
1192 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
1193 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
1194 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1195 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1196 // CHECK3-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1197 // CHECK3-NEXT:    store i32 1, i32* [[B]], align 4
1198 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1199 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
1200 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
1201 // CHECK3-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
1202 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
1203 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[B_CASTED]], align 4
1204 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
1205 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1206 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
1207 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
1208 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
1209 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1210 // CHECK3-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
1211 // CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 4
1212 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1213 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
1214 // CHECK3-NEXT:    store double* [[A]], double** [[TMP9]], align 4
1215 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1216 // CHECK3-NEXT:    store i8* null, i8** [[TMP10]], align 4
1217 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1218 // CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
1219 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP12]], align 4
1220 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1221 // CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
1222 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP14]], align 4
1223 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1224 // CHECK3-NEXT:    store i8* null, i8** [[TMP15]], align 4
1225 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1226 // CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
1227 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
1228 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1229 // CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
1230 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
1231 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1232 // CHECK3-NEXT:    store i8* null, i8** [[TMP20]], align 4
1233 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1234 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1235 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1236 // CHECK3-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
1237 // CHECK3-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
1238 // CHECK3-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1239 // CHECK3:       omp_offload.failed:
1240 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR2:[0-9]+]]
1241 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1242 // CHECK3:       omp_offload.cont:
1243 // CHECK3-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
1244 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
1245 // CHECK3-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
1246 // CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 4
1247 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
1248 // CHECK3-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
1249 // CHECK3-NEXT:    store double* [[A2]], double** [[TMP29]], align 4
1250 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
1251 // CHECK3-NEXT:    store i8* null, i8** [[TMP30]], align 4
1252 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
1253 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
1254 // CHECK3-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
1255 // CHECK3-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
1256 // CHECK3-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
1257 // CHECK3:       omp_offload.failed6:
1258 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR2]]
1259 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
1260 // CHECK3:       omp_offload.cont7:
1261 // CHECK3-NEXT:    [[A8:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
1262 // CHECK3-NEXT:    [[TMP35:%.*]] = load double, double* [[A8]], align 4
1263 // CHECK3-NEXT:    [[CONV:%.*]] = fptosi double [[TMP35]] to i32
1264 // CHECK3-NEXT:    ret i32 [[CONV]]
1265 //
1266 //
1267 // CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici
1268 // CHECK3-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
1269 // CHECK3-NEXT:  entry:
1270 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1271 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1272 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
1273 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
1274 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
1275 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
1276 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1277 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
1278 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
1279 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
1280 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
1281 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1282 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1283 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
1284 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1285 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
1286 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
1287 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1288 // CHECK3-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
1289 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[TMP4]], align 4
1290 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1291 // CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
1292 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[TMP6]], align 4
1293 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1294 // CHECK3-NEXT:    store i8* null, i8** [[TMP7]], align 4
1295 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1296 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1297 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1298 // CHECK3-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
1299 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1300 // CHECK3-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1301 // CHECK3:       omp_offload.failed:
1302 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i32 [[TMP2]]) #[[ATTR2]]
1303 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1304 // CHECK3:       omp_offload.cont:
1305 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
1306 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
1307 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1308 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1309 // CHECK3-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
1310 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
1311 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
1312 // CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
1313 // CHECK3-NEXT:    store i32 [[TMP15]], i32* [[TMP17]], align 4
1314 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
1315 // CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
1316 // CHECK3-NEXT:    store i32 [[TMP15]], i32* [[TMP19]], align 4
1317 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
1318 // CHECK3-NEXT:    store i8* null, i8** [[TMP20]], align 4
1319 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
1320 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
1321 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1322 // CHECK3-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
1323 // CHECK3-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
1324 // CHECK3-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
1325 // CHECK3:       omp_offload.failed6:
1326 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i32 [[TMP15]]) #[[ATTR2]]
1327 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
1328 // CHECK3:       omp_offload.cont7:
1329 // CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
1330 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP26]], 1
1331 // CHECK3-NEXT:    ret i32 [[ADD8]]
1332 //
1333 //
1334 // CHECK3-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
1335 // CHECK3-SAME: (i32 [[N:%.*]]) #[[ATTR0]] comdat {
1336 // CHECK3-NEXT:  entry:
1337 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1338 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
1339 // CHECK3-NEXT:    [[B:%.*]] = alloca i16, align 2
1340 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
1341 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
1342 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1343 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
1344 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
1345 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
1346 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
1347 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1348 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
1349 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
1350 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
1351 // CHECK3-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1352 // CHECK3:       omp_offload.failed:
1353 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
1354 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1355 // CHECK3:       omp_offload.cont:
1356 // CHECK3-NEXT:    store i16 1, i16* [[B]], align 2
1357 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
1358 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
1359 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
1360 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
1361 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
1362 // CHECK3-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
1363 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
1364 // CHECK3-NEXT:    store i16 [[TMP5]], i16* [[CONV]], align 2
1365 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
1366 // CHECK3-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
1367 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i16*
1368 // CHECK3-NEXT:    store i16 [[TMP7]], i16* [[CONV1]], align 2
1369 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
1370 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1371 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
1372 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[TMP10]], align 4
1373 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1374 // CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
1375 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[TMP12]], align 4
1376 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1377 // CHECK3-NEXT:    store i8* null, i8** [[TMP13]], align 4
1378 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1379 // CHECK3-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
1380 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[TMP15]], align 4
1381 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1382 // CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
1383 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[TMP17]], align 4
1384 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1385 // CHECK3-NEXT:    store i8* null, i8** [[TMP18]], align 4
1386 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1387 // CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
1388 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[TMP20]], align 4
1389 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1390 // CHECK3-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
1391 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[TMP22]], align 4
1392 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1393 // CHECK3-NEXT:    store i8* null, i8** [[TMP23]], align 4
1394 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1395 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1396 // CHECK3-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
1397 // CHECK3-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
1398 // CHECK3-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
1399 // CHECK3-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
1400 // CHECK3-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
1401 // CHECK3:       omp_offload.failed2:
1402 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR2]]
1403 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
1404 // CHECK3:       omp_offload.cont3:
1405 // CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
1406 // CHECK3-NEXT:    ret i32 [[TMP30]]
1407 //
1408 //
1409 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
1410 // CHECK3-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
1411 // CHECK3-NEXT:  entry:
1412 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1413 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1414 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1415 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
1416 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1417 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1418 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1419 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1420 // CHECK3-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1421 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1422 // CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
1423 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1424 // CHECK3-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 4
1425 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1426 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_ADDR]], align 4
1427 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 4
1428 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
1429 // CHECK3-NEXT:    ret void
1430 //
1431 //
1432 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
1433 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1434 // CHECK3-NEXT:  entry:
1435 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1436 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1437 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
1438 // CHECK3-NEXT:    [[B:%.*]] = alloca i32, align 4
1439 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1440 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1441 // CHECK3-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
1442 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
1443 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
1444 // CHECK3-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
1445 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
1446 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1447 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
1448 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
1449 // CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
1450 // CHECK3-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
1451 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
1452 // CHECK3-NEXT:    store double [[ADD]], double* [[A]], align 4
1453 // CHECK3-NEXT:    ret void
1454 //
1455 //
1456 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
1457 // CHECK3-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
1458 // CHECK3-NEXT:  entry:
1459 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1460 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 4
1461 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1462 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1463 // CHECK3-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1464 // CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
1465 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1466 // CHECK3-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 4
1467 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
1468 // CHECK3-NEXT:    ret void
1469 //
1470 //
1471 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
1472 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1473 // CHECK3-NEXT:  entry:
1474 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1475 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1476 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4
1477 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1478 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1479 // CHECK3-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4
1480 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4
1481 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
1482 // CHECK3-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
1483 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
1484 // CHECK3-NEXT:    store double 2.500000e+00, double* [[A]], align 4
1485 // CHECK3-NEXT:    ret void
1486 //
1487 //
1488 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
1489 // CHECK3-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1490 // CHECK3-NEXT:  entry:
1491 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1492 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 1
1493 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1494 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1495 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1496 // CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
1497 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
1498 // CHECK3-NEXT:    ret void
1499 //
1500 //
1501 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
1502 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1503 // CHECK3-NEXT:  entry:
1504 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1505 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1506 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
1507 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1508 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1509 // CHECK3-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
1510 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
1511 // CHECK3-NEXT:    ret void
1512 //
1513 //
1514 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
1515 // CHECK3-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1516 // CHECK3-NEXT:  entry:
1517 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1518 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 1
1519 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1520 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1521 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1522 // CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
1523 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
1524 // CHECK3-NEXT:    ret void
1525 //
1526 //
1527 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
1528 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1529 // CHECK3-NEXT:  entry:
1530 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1531 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1532 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
1533 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1534 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1535 // CHECK3-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
1536 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
1537 // CHECK3-NEXT:    ret void
1538 //
1539 //
1540 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
1541 // CHECK3-SAME: () #[[ATTR1]] {
1542 // CHECK3-NEXT:  entry:
1543 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
1544 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1545 // CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
1546 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
1547 // CHECK3-NEXT:    ret void
1548 //
1549 //
1550 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..10
1551 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1552 // CHECK3-NEXT:  entry:
1553 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1554 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1555 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
1556 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1557 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1558 // CHECK3-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
1559 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
1560 // CHECK3-NEXT:    ret void
1561 //
1562 //
1563 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
1564 // CHECK3-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1565 // CHECK3-NEXT:  entry:
1566 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1567 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1568 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1569 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
1570 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1571 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
1572 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1573 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1574 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
1575 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
1576 // CHECK3-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
1577 // CHECK3-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
1578 // CHECK3-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
1579 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1580 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_ADDR]], align 4
1581 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
1582 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1583 // CHECK3-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV]], align 4
1584 // CHECK3-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
1585 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
1586 // CHECK3-NEXT:    ret void
1587 //
1588 //
1589 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
1590 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1591 // CHECK3-NEXT:  entry:
1592 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1593 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1594 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
1595 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
1596 // CHECK3-NEXT:    [[B:%.*]] = alloca i16, align 2
1597 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1598 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1599 // CHECK3-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
1600 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
1601 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
1602 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1603 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
1604 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
1605 // CHECK3-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
1606 // CHECK3-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
1607 // CHECK3-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
1608 // CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
1609 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
1610 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
1611 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1612 // CHECK3-NEXT:    ret void
1613 //
1614 //
1615 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1616 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
1617 // CHECK3-NEXT:  entry:
1618 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
1619 // CHECK3-NEXT:    ret void
1620 //
1621 //
1622 // CHECK4-LABEL: define {{[^@]+}}@_Z3bari
1623 // CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
1624 // CHECK4-NEXT:  entry:
1625 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1626 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
1627 // CHECK4-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
1628 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1629 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
1630 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1631 // CHECK4-NEXT:    [[CALL:%.*]] = call i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], i32 [[TMP0]])
1632 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
1633 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
1634 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1635 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
1636 // CHECK4-NEXT:    [[CALL1:%.*]] = call i32 @_ZL7fstatici(i32 [[TMP2]])
1637 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
1638 // CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
1639 // CHECK4-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
1640 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
1641 // CHECK4-NEXT:    [[CALL3:%.*]] = call i32 @_Z9ftemplateIiET_i(i32 [[TMP4]])
1642 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
1643 // CHECK4-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
1644 // CHECK4-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
1645 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
1646 // CHECK4-NEXT:    ret i32 [[TMP6]]
1647 //
1648 //
1649 // CHECK4-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
1650 // CHECK4-SAME: (%struct.S1* nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 [[N:%.*]]) #[[ATTR0]] comdat align 2 {
1651 // CHECK4-NEXT:  entry:
1652 // CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1653 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1654 // CHECK4-NEXT:    [[B:%.*]] = alloca i32, align 4
1655 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1656 // CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1657 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
1658 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
1659 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
1660 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
1661 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
1662 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
1663 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
1664 // CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1665 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1666 // CHECK4-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1667 // CHECK4-NEXT:    store i32 1, i32* [[B]], align 4
1668 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1669 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
1670 // CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
1671 // CHECK4-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
1672 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
1673 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[B_CASTED]], align 4
1674 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
1675 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1676 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
1677 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
1678 // CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
1679 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1680 // CHECK4-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
1681 // CHECK4-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 4
1682 // CHECK4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1683 // CHECK4-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
1684 // CHECK4-NEXT:    store double* [[A]], double** [[TMP9]], align 4
1685 // CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1686 // CHECK4-NEXT:    store i8* null, i8** [[TMP10]], align 4
1687 // CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1688 // CHECK4-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
1689 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP12]], align 4
1690 // CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1691 // CHECK4-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
1692 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP14]], align 4
1693 // CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1694 // CHECK4-NEXT:    store i8* null, i8** [[TMP15]], align 4
1695 // CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1696 // CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
1697 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
1698 // CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1699 // CHECK4-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
1700 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
1701 // CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1702 // CHECK4-NEXT:    store i8* null, i8** [[TMP20]], align 4
1703 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1704 // CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1705 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1706 // CHECK4-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
1707 // CHECK4-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
1708 // CHECK4-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1709 // CHECK4:       omp_offload.failed:
1710 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR2:[0-9]+]]
1711 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1712 // CHECK4:       omp_offload.cont:
1713 // CHECK4-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
1714 // CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
1715 // CHECK4-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
1716 // CHECK4-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 4
1717 // CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
1718 // CHECK4-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
1719 // CHECK4-NEXT:    store double* [[A2]], double** [[TMP29]], align 4
1720 // CHECK4-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
1721 // CHECK4-NEXT:    store i8* null, i8** [[TMP30]], align 4
1722 // CHECK4-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
1723 // CHECK4-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
1724 // CHECK4-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
1725 // CHECK4-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
1726 // CHECK4-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
1727 // CHECK4:       omp_offload.failed6:
1728 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR2]]
1729 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
1730 // CHECK4:       omp_offload.cont7:
1731 // CHECK4-NEXT:    [[A8:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
1732 // CHECK4-NEXT:    [[TMP35:%.*]] = load double, double* [[A8]], align 4
1733 // CHECK4-NEXT:    [[CONV:%.*]] = fptosi double [[TMP35]] to i32
1734 // CHECK4-NEXT:    ret i32 [[CONV]]
1735 //
1736 //
1737 // CHECK4-LABEL: define {{[^@]+}}@_ZL7fstatici
1738 // CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
1739 // CHECK4-NEXT:  entry:
1740 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1741 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1742 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
1743 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
1744 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
1745 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
1746 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1747 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
1748 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
1749 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
1750 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
1751 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1752 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1753 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
1754 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1755 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
1756 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
1757 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1758 // CHECK4-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
1759 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[TMP4]], align 4
1760 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1761 // CHECK4-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
1762 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[TMP6]], align 4
1763 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1764 // CHECK4-NEXT:    store i8* null, i8** [[TMP7]], align 4
1765 // CHECK4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1766 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1767 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1768 // CHECK4-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
1769 // CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1770 // CHECK4-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1771 // CHECK4:       omp_offload.failed:
1772 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i32 [[TMP2]]) #[[ATTR2]]
1773 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1774 // CHECK4:       omp_offload.cont:
1775 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
1776 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
1777 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1778 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1779 // CHECK4-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
1780 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
1781 // CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
1782 // CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
1783 // CHECK4-NEXT:    store i32 [[TMP15]], i32* [[TMP17]], align 4
1784 // CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
1785 // CHECK4-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
1786 // CHECK4-NEXT:    store i32 [[TMP15]], i32* [[TMP19]], align 4
1787 // CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
1788 // CHECK4-NEXT:    store i8* null, i8** [[TMP20]], align 4
1789 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
1790 // CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
1791 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1792 // CHECK4-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
1793 // CHECK4-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
1794 // CHECK4-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
1795 // CHECK4:       omp_offload.failed6:
1796 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i32 [[TMP15]]) #[[ATTR2]]
1797 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
1798 // CHECK4:       omp_offload.cont7:
1799 // CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
1800 // CHECK4-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP26]], 1
1801 // CHECK4-NEXT:    ret i32 [[ADD8]]
1802 //
1803 //
1804 // CHECK4-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
1805 // CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0]] comdat {
1806 // CHECK4-NEXT:  entry:
1807 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1808 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
1809 // CHECK4-NEXT:    [[B:%.*]] = alloca i16, align 2
1810 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
1811 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
1812 // CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
1813 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
1814 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
1815 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
1816 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
1817 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1818 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
1819 // CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
1820 // CHECK4-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
1821 // CHECK4-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1822 // CHECK4:       omp_offload.failed:
1823 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
1824 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1825 // CHECK4:       omp_offload.cont:
1826 // CHECK4-NEXT:    store i16 1, i16* [[B]], align 2
1827 // CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
1828 // CHECK4-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
1829 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
1830 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
1831 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
1832 // CHECK4-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
1833 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
1834 // CHECK4-NEXT:    store i16 [[TMP5]], i16* [[CONV]], align 2
1835 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
1836 // CHECK4-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
1837 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i16*
1838 // CHECK4-NEXT:    store i16 [[TMP7]], i16* [[CONV1]], align 2
1839 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
1840 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1841 // CHECK4-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
1842 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[TMP10]], align 4
1843 // CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1844 // CHECK4-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
1845 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[TMP12]], align 4
1846 // CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1847 // CHECK4-NEXT:    store i8* null, i8** [[TMP13]], align 4
1848 // CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1849 // CHECK4-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
1850 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[TMP15]], align 4
1851 // CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1852 // CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
1853 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[TMP17]], align 4
1854 // CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1855 // CHECK4-NEXT:    store i8* null, i8** [[TMP18]], align 4
1856 // CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1857 // CHECK4-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
1858 // CHECK4-NEXT:    store i32 [[TMP8]], i32* [[TMP20]], align 4
1859 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1860 // CHECK4-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
1861 // CHECK4-NEXT:    store i32 [[TMP8]], i32* [[TMP22]], align 4
1862 // CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1863 // CHECK4-NEXT:    store i8* null, i8** [[TMP23]], align 4
1864 // CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1865 // CHECK4-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1866 // CHECK4-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
1867 // CHECK4-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
1868 // CHECK4-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
1869 // CHECK4-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
1870 // CHECK4-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
1871 // CHECK4:       omp_offload.failed2:
1872 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR2]]
1873 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
1874 // CHECK4:       omp_offload.cont3:
1875 // CHECK4-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
1876 // CHECK4-NEXT:    ret i32 [[TMP30]]
1877 //
1878 //
1879 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
1880 // CHECK4-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
1881 // CHECK4-NEXT:  entry:
1882 // CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1883 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
1884 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1885 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
1886 // CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1887 // CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1888 // CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
1889 // CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1890 // CHECK4-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1891 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1892 // CHECK4-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
1893 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1894 // CHECK4-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 4
1895 // CHECK4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1896 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_ADDR]], align 4
1897 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 4
1898 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
1899 // CHECK4-NEXT:    ret void
1900 //
1901 //
1902 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
1903 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1904 // CHECK4-NEXT:  entry:
1905 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1906 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1907 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
1908 // CHECK4-NEXT:    [[B:%.*]] = alloca i32, align 4
1909 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1910 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1911 // CHECK4-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
1912 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
1913 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
1914 // CHECK4-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
1915 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
1916 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1917 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
1918 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
1919 // CHECK4-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
1920 // CHECK4-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
1921 // CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
1922 // CHECK4-NEXT:    store double [[ADD]], double* [[A]], align 4
1923 // CHECK4-NEXT:    ret void
1924 //
1925 //
1926 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
1927 // CHECK4-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
1928 // CHECK4-NEXT:  entry:
1929 // CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
1930 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 4
1931 // CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1932 // CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
1933 // CHECK4-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
1934 // CHECK4-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
1935 // CHECK4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1936 // CHECK4-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 4
1937 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
1938 // CHECK4-NEXT:    ret void
1939 //
1940 //
1941 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
1942 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1943 // CHECK4-NEXT:  entry:
1944 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1945 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1946 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4
1947 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1948 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1949 // CHECK4-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4
1950 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4
1951 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
1952 // CHECK4-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
1953 // CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
1954 // CHECK4-NEXT:    store double 2.500000e+00, double* [[A]], align 4
1955 // CHECK4-NEXT:    ret void
1956 //
1957 //
1958 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
1959 // CHECK4-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1960 // CHECK4-NEXT:  entry:
1961 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1962 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 1
1963 // CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1964 // CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1965 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1966 // CHECK4-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
1967 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
1968 // CHECK4-NEXT:    ret void
1969 //
1970 //
1971 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..4
1972 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1973 // CHECK4-NEXT:  entry:
1974 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1975 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1976 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
1977 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1978 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1979 // CHECK4-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
1980 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
1981 // CHECK4-NEXT:    ret void
1982 //
1983 //
1984 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
1985 // CHECK4-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
1986 // CHECK4-NEXT:  entry:
1987 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
1988 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 1
1989 // CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1990 // CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1991 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
1992 // CHECK4-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
1993 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
1994 // CHECK4-NEXT:    ret void
1995 //
1996 //
1997 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..7
1998 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1999 // CHECK4-NEXT:  entry:
2000 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2001 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2002 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
2003 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2004 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2005 // CHECK4-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
2006 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
2007 // CHECK4-NEXT:    ret void
2008 //
2009 //
2010 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
2011 // CHECK4-SAME: () #[[ATTR1]] {
2012 // CHECK4-NEXT:  entry:
2013 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
2014 // CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2015 // CHECK4-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
2016 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
2017 // CHECK4-NEXT:    ret void
2018 //
2019 //
2020 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..10
2021 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
2022 // CHECK4-NEXT:  entry:
2023 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2024 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2025 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
2026 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2027 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2028 // CHECK4-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
2029 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
2030 // CHECK4-NEXT:    ret void
2031 //
2032 //
2033 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
2034 // CHECK4-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
2035 // CHECK4-NEXT:  entry:
2036 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2037 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
2038 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2039 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
2040 // CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2041 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2042 // CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
2043 // CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2044 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
2045 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
2046 // CHECK4-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
2047 // CHECK4-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
2048 // CHECK4-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
2049 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2050 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_ADDR]], align 4
2051 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
2052 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2053 // CHECK4-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV]], align 4
2054 // CHECK4-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
2055 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
2056 // CHECK4-NEXT:    ret void
2057 //
2058 //
2059 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..11
2060 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
2061 // CHECK4-NEXT:  entry:
2062 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2063 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2064 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
2065 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
2066 // CHECK4-NEXT:    [[B:%.*]] = alloca i16, align 2
2067 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2068 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2069 // CHECK4-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
2070 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
2071 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
2072 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2073 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
2074 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
2075 // CHECK4-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
2076 // CHECK4-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
2077 // CHECK4-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
2078 // CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
2079 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
2080 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
2081 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2082 // CHECK4-NEXT:    ret void
2083 //
2084 //
2085 // CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2086 // CHECK4-SAME: () #[[ATTR3:[0-9]+]] {
2087 // CHECK4-NEXT:  entry:
2088 // CHECK4-NEXT:    call void @__tgt_register_requires(i64 1)
2089 // CHECK4-NEXT:    ret void
2090 //
2091 //
2092 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
2093 // CHECK9-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
2094 // CHECK9-NEXT:  entry:
2095 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2096 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
2097 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
2098 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2099 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2100 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
2101 // CHECK9-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
2102 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
2103 // CHECK9-NEXT:    ret void
2104 //
2105 //
2106 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
2107 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2108 // CHECK9-NEXT:  entry:
2109 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2110 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2111 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
2112 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2113 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2114 // CHECK9-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
2115 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
2116 // CHECK9-NEXT:    ret void
2117 //
2118 //
2119 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
2120 // CHECK9-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
2121 // CHECK9-NEXT:  entry:
2122 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2123 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
2124 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2125 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2126 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2127 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
2128 // CHECK9-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
2129 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
2130 // CHECK9-NEXT:    ret void
2131 //
2132 //
2133 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
2134 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2135 // CHECK9-NEXT:  entry:
2136 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2137 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2138 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
2139 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2140 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2141 // CHECK9-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
2142 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
2143 // CHECK9-NEXT:    ret void
2144 //
2145 //
2146 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
2147 // CHECK9-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
2148 // CHECK9-NEXT:  entry:
2149 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
2150 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
2151 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2152 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 8
2153 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2154 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
2155 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
2156 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2157 // CHECK9-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
2158 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
2159 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2160 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
2161 // CHECK9-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
2162 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2163 // CHECK9-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 8
2164 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2165 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 8
2166 // CHECK9-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 8
2167 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
2168 // CHECK9-NEXT:    ret void
2169 //
2170 //
2171 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
2172 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2173 // CHECK9-NEXT:  entry:
2174 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2175 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2176 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
2177 // CHECK9-NEXT:    [[B:%.*]] = alloca i32, align 4
2178 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2179 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2180 // CHECK9-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
2181 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
2182 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
2183 // CHECK9-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
2184 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
2185 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
2186 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
2187 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
2188 // CHECK9-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
2189 // CHECK9-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
2190 // CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
2191 // CHECK9-NEXT:    store double [[ADD]], double* [[A]], align 8
2192 // CHECK9-NEXT:    ret void
2193 //
2194 //
2195 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
2196 // CHECK9-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
2197 // CHECK9-NEXT:  entry:
2198 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
2199 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
2200 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2201 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
2202 // CHECK9-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
2203 // CHECK9-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
2204 // CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2205 // CHECK9-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 8
2206 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
2207 // CHECK9-NEXT:    ret void
2208 //
2209 //
2210 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
2211 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2212 // CHECK9-NEXT:  entry:
2213 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2214 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2215 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
2216 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2217 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2218 // CHECK9-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
2219 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
2220 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
2221 // CHECK9-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
2222 // CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
2223 // CHECK9-NEXT:    store double 2.500000e+00, double* [[A]], align 8
2224 // CHECK9-NEXT:    ret void
2225 //
2226 //
2227 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
2228 // CHECK9-SAME: () #[[ATTR0]] {
2229 // CHECK9-NEXT:  entry:
2230 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
2231 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2232 // CHECK9-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
2233 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
2234 // CHECK9-NEXT:    ret void
2235 //
2236 //
2237 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
2238 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2239 // CHECK9-NEXT:  entry:
2240 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2241 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2242 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
2243 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2244 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2245 // CHECK9-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
2246 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
2247 // CHECK9-NEXT:    ret void
2248 //
2249 //
2250 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
2251 // CHECK9-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
2252 // CHECK9-NEXT:  entry:
2253 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2254 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
2255 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2256 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
2257 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2258 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2259 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
2260 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2261 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2262 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
2263 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
2264 // CHECK9-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
2265 // CHECK9-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
2266 // CHECK9-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
2267 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2268 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
2269 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
2270 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2271 // CHECK9-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV1]], align 8
2272 // CHECK9-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
2273 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
2274 // CHECK9-NEXT:    ret void
2275 //
2276 //
2277 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
2278 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2279 // CHECK9-NEXT:  entry:
2280 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2281 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2282 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
2283 // CHECK9-NEXT:    [[A:%.*]] = alloca i32, align 4
2284 // CHECK9-NEXT:    [[B:%.*]] = alloca i16, align 2
2285 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2286 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2287 // CHECK9-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
2288 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
2289 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
2290 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2291 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
2292 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
2293 // CHECK9-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
2294 // CHECK9-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
2295 // CHECK9-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
2296 // CHECK9-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
2297 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
2298 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
2299 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2300 // CHECK9-NEXT:    ret void
2301 //
2302 //
2303 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
2304 // CHECK10-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
2305 // CHECK10-NEXT:  entry:
2306 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2307 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
2308 // CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
2309 // CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2310 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2311 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
2312 // CHECK10-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
2313 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
2314 // CHECK10-NEXT:    ret void
2315 //
2316 //
2317 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined.
2318 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2319 // CHECK10-NEXT:  entry:
2320 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2321 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2322 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
2323 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2324 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2325 // CHECK10-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
2326 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
2327 // CHECK10-NEXT:    ret void
2328 //
2329 //
2330 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
2331 // CHECK10-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
2332 // CHECK10-NEXT:  entry:
2333 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2334 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
2335 // CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2336 // CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2337 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2338 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
2339 // CHECK10-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
2340 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
2341 // CHECK10-NEXT:    ret void
2342 //
2343 //
2344 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
2345 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2346 // CHECK10-NEXT:  entry:
2347 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2348 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2349 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
2350 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2351 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2352 // CHECK10-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
2353 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
2354 // CHECK10-NEXT:    ret void
2355 //
2356 //
2357 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
2358 // CHECK10-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
2359 // CHECK10-NEXT:  entry:
2360 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
2361 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
2362 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2363 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 8
2364 // CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2365 // CHECK10-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
2366 // CHECK10-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
2367 // CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2368 // CHECK10-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
2369 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
2370 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2371 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
2372 // CHECK10-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
2373 // CHECK10-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2374 // CHECK10-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 8
2375 // CHECK10-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2376 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 8
2377 // CHECK10-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 8
2378 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
2379 // CHECK10-NEXT:    ret void
2380 //
2381 //
2382 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
2383 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2384 // CHECK10-NEXT:  entry:
2385 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2386 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2387 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
2388 // CHECK10-NEXT:    [[B:%.*]] = alloca i32, align 4
2389 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2390 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2391 // CHECK10-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
2392 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
2393 // CHECK10-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
2394 // CHECK10-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
2395 // CHECK10-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
2396 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
2397 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
2398 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
2399 // CHECK10-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
2400 // CHECK10-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
2401 // CHECK10-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
2402 // CHECK10-NEXT:    store double [[ADD]], double* [[A]], align 8
2403 // CHECK10-NEXT:    ret void
2404 //
2405 //
2406 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
2407 // CHECK10-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
2408 // CHECK10-NEXT:  entry:
2409 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
2410 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
2411 // CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2412 // CHECK10-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
2413 // CHECK10-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
2414 // CHECK10-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
2415 // CHECK10-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2416 // CHECK10-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 8
2417 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
2418 // CHECK10-NEXT:    ret void
2419 //
2420 //
2421 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
2422 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2423 // CHECK10-NEXT:  entry:
2424 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2425 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2426 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
2427 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2428 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2429 // CHECK10-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
2430 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
2431 // CHECK10-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
2432 // CHECK10-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
2433 // CHECK10-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
2434 // CHECK10-NEXT:    store double 2.500000e+00, double* [[A]], align 8
2435 // CHECK10-NEXT:    ret void
2436 //
2437 //
2438 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
2439 // CHECK10-SAME: () #[[ATTR0]] {
2440 // CHECK10-NEXT:  entry:
2441 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
2442 // CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2443 // CHECK10-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
2444 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
2445 // CHECK10-NEXT:    ret void
2446 //
2447 //
2448 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..4
2449 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2450 // CHECK10-NEXT:  entry:
2451 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2452 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2453 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
2454 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2455 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2456 // CHECK10-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
2457 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
2458 // CHECK10-NEXT:    ret void
2459 //
2460 //
2461 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
2462 // CHECK10-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
2463 // CHECK10-NEXT:  entry:
2464 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2465 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
2466 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2467 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
2468 // CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2469 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2470 // CHECK10-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
2471 // CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2472 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2473 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
2474 // CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
2475 // CHECK10-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
2476 // CHECK10-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
2477 // CHECK10-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
2478 // CHECK10-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2479 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
2480 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
2481 // CHECK10-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2482 // CHECK10-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV1]], align 8
2483 // CHECK10-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
2484 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
2485 // CHECK10-NEXT:    ret void
2486 //
2487 //
2488 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..5
2489 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2490 // CHECK10-NEXT:  entry:
2491 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2492 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2493 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
2494 // CHECK10-NEXT:    [[A:%.*]] = alloca i32, align 4
2495 // CHECK10-NEXT:    [[B:%.*]] = alloca i16, align 2
2496 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2497 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2498 // CHECK10-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
2499 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
2500 // CHECK10-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
2501 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2502 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
2503 // CHECK10-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
2504 // CHECK10-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
2505 // CHECK10-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
2506 // CHECK10-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
2507 // CHECK10-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
2508 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
2509 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
2510 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2511 // CHECK10-NEXT:    ret void
2512 //
2513 //
2514 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
2515 // CHECK11-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
2516 // CHECK11-NEXT:  entry:
2517 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2518 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
2519 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
2520 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2521 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2522 // CHECK11-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
2523 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
2524 // CHECK11-NEXT:    ret void
2525 //
2526 //
2527 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
2528 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2529 // CHECK11-NEXT:  entry:
2530 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2531 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2532 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
2533 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2534 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2535 // CHECK11-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
2536 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
2537 // CHECK11-NEXT:    ret void
2538 //
2539 //
2540 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
2541 // CHECK11-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
2542 // CHECK11-NEXT:  entry:
2543 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2544 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
2545 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2546 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2547 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2548 // CHECK11-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
2549 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
2550 // CHECK11-NEXT:    ret void
2551 //
2552 //
2553 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
2554 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2555 // CHECK11-NEXT:  entry:
2556 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2557 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2558 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4
2559 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2560 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2561 // CHECK11-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4
2562 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4
2563 // CHECK11-NEXT:    ret void
2564 //
2565 //
2566 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
2567 // CHECK11-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
2568 // CHECK11-NEXT:  entry:
2569 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
2570 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
2571 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2572 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
2573 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2574 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
2575 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
2576 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2577 // CHECK11-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
2578 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2579 // CHECK11-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
2580 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2581 // CHECK11-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 4
2582 // CHECK11-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2583 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_ADDR]], align 4
2584 // CHECK11-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 4
2585 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
2586 // CHECK11-NEXT:    ret void
2587 //
2588 //
2589 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
2590 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2591 // CHECK11-NEXT:  entry:
2592 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2593 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2594 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
2595 // CHECK11-NEXT:    [[B:%.*]] = alloca i32, align 4
2596 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2597 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2598 // CHECK11-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
2599 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
2600 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
2601 // CHECK11-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
2602 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
2603 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2604 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
2605 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
2606 // CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
2607 // CHECK11-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
2608 // CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
2609 // CHECK11-NEXT:    store double [[ADD]], double* [[A]], align 4
2610 // CHECK11-NEXT:    ret void
2611 //
2612 //
2613 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
2614 // CHECK11-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
2615 // CHECK11-NEXT:  entry:
2616 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
2617 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 4
2618 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2619 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
2620 // CHECK11-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
2621 // CHECK11-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
2622 // CHECK11-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2623 // CHECK11-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 4
2624 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
2625 // CHECK11-NEXT:    ret void
2626 //
2627 //
2628 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
2629 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2630 // CHECK11-NEXT:  entry:
2631 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2632 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2633 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
2634 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2635 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2636 // CHECK11-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
2637 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
2638 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
2639 // CHECK11-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
2640 // CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
2641 // CHECK11-NEXT:    store double 2.500000e+00, double* [[A]], align 4
2642 // CHECK11-NEXT:    ret void
2643 //
2644 //
2645 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
2646 // CHECK11-SAME: () #[[ATTR0]] {
2647 // CHECK11-NEXT:  entry:
2648 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
2649 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2650 // CHECK11-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
2651 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
2652 // CHECK11-NEXT:    ret void
2653 //
2654 //
2655 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
2656 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2657 // CHECK11-NEXT:  entry:
2658 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2659 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2660 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
2661 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2662 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2663 // CHECK11-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
2664 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
2665 // CHECK11-NEXT:    ret void
2666 //
2667 //
2668 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
2669 // CHECK11-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
2670 // CHECK11-NEXT:  entry:
2671 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2672 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
2673 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2674 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
2675 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2676 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2677 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
2678 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2679 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
2680 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
2681 // CHECK11-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
2682 // CHECK11-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
2683 // CHECK11-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
2684 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2685 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_ADDR]], align 4
2686 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
2687 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2688 // CHECK11-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV]], align 4
2689 // CHECK11-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
2690 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
2691 // CHECK11-NEXT:    ret void
2692 //
2693 //
2694 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
2695 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2696 // CHECK11-NEXT:  entry:
2697 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2698 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2699 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
2700 // CHECK11-NEXT:    [[A:%.*]] = alloca i32, align 4
2701 // CHECK11-NEXT:    [[B:%.*]] = alloca i16, align 2
2702 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2703 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2704 // CHECK11-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
2705 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
2706 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
2707 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2708 // CHECK11-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
2709 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
2710 // CHECK11-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
2711 // CHECK11-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
2712 // CHECK11-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
2713 // CHECK11-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
2714 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
2715 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
2716 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2717 // CHECK11-NEXT:    ret void
2718 //
2719 //
2720 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
2721 // CHECK12-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
2722 // CHECK12-NEXT:  entry:
2723 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2724 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
2725 // CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
2726 // CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2727 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2728 // CHECK12-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
2729 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
2730 // CHECK12-NEXT:    ret void
2731 //
2732 //
2733 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined.
2734 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2735 // CHECK12-NEXT:  entry:
2736 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2737 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2738 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
2739 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2740 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2741 // CHECK12-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
2742 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
2743 // CHECK12-NEXT:    ret void
2744 //
2745 //
2746 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
2747 // CHECK12-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
2748 // CHECK12-NEXT:  entry:
2749 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2750 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
2751 // CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2752 // CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2753 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2754 // CHECK12-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
2755 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
2756 // CHECK12-NEXT:    ret void
2757 //
2758 //
2759 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..1
2760 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2761 // CHECK12-NEXT:  entry:
2762 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2763 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2764 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4
2765 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2766 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2767 // CHECK12-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4
2768 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4
2769 // CHECK12-NEXT:    ret void
2770 //
2771 //
2772 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
2773 // CHECK12-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
2774 // CHECK12-NEXT:  entry:
2775 // CHECK12-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
2776 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
2777 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2778 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
2779 // CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2780 // CHECK12-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
2781 // CHECK12-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
2782 // CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2783 // CHECK12-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
2784 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2785 // CHECK12-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
2786 // CHECK12-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2787 // CHECK12-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 4
2788 // CHECK12-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2789 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_ADDR]], align 4
2790 // CHECK12-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 4
2791 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
2792 // CHECK12-NEXT:    ret void
2793 //
2794 //
2795 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..2
2796 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2797 // CHECK12-NEXT:  entry:
2798 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2799 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2800 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
2801 // CHECK12-NEXT:    [[B:%.*]] = alloca i32, align 4
2802 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2803 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2804 // CHECK12-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
2805 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
2806 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
2807 // CHECK12-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
2808 // CHECK12-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
2809 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2810 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
2811 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
2812 // CHECK12-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
2813 // CHECK12-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
2814 // CHECK12-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
2815 // CHECK12-NEXT:    store double [[ADD]], double* [[A]], align 4
2816 // CHECK12-NEXT:    ret void
2817 //
2818 //
2819 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
2820 // CHECK12-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
2821 // CHECK12-NEXT:  entry:
2822 // CHECK12-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
2823 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 4
2824 // CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2825 // CHECK12-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
2826 // CHECK12-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
2827 // CHECK12-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
2828 // CHECK12-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2829 // CHECK12-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 4
2830 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
2831 // CHECK12-NEXT:    ret void
2832 //
2833 //
2834 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..3
2835 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2836 // CHECK12-NEXT:  entry:
2837 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2838 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2839 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
2840 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2841 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2842 // CHECK12-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
2843 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
2844 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
2845 // CHECK12-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
2846 // CHECK12-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
2847 // CHECK12-NEXT:    store double 2.500000e+00, double* [[A]], align 4
2848 // CHECK12-NEXT:    ret void
2849 //
2850 //
2851 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
2852 // CHECK12-SAME: () #[[ATTR0]] {
2853 // CHECK12-NEXT:  entry:
2854 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
2855 // CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2856 // CHECK12-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
2857 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
2858 // CHECK12-NEXT:    ret void
2859 //
2860 //
2861 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..4
2862 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2863 // CHECK12-NEXT:  entry:
2864 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2865 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2866 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
2867 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2868 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2869 // CHECK12-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
2870 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
2871 // CHECK12-NEXT:    ret void
2872 //
2873 //
2874 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
2875 // CHECK12-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
2876 // CHECK12-NEXT:  entry:
2877 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2878 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
2879 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2880 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
2881 // CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2882 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2883 // CHECK12-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
2884 // CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2885 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
2886 // CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
2887 // CHECK12-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
2888 // CHECK12-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
2889 // CHECK12-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
2890 // CHECK12-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2891 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_ADDR]], align 4
2892 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
2893 // CHECK12-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2894 // CHECK12-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV]], align 4
2895 // CHECK12-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
2896 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
2897 // CHECK12-NEXT:    ret void
2898 //
2899 //
2900 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..5
2901 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
2902 // CHECK12-NEXT:  entry:
2903 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2904 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2905 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
2906 // CHECK12-NEXT:    [[A:%.*]] = alloca i32, align 4
2907 // CHECK12-NEXT:    [[B:%.*]] = alloca i16, align 2
2908 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2909 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2910 // CHECK12-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
2911 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
2912 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
2913 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2914 // CHECK12-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
2915 // CHECK12-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
2916 // CHECK12-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
2917 // CHECK12-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
2918 // CHECK12-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
2919 // CHECK12-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
2920 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
2921 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
2922 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2923 // CHECK12-NEXT:    ret void
2924 //
2925 //
2926 // CHECK17-LABEL: define {{[^@]+}}@_Z3bari
2927 // CHECK17-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
2928 // CHECK17-NEXT:  entry:
2929 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2930 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
2931 // CHECK17-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
2932 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2933 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
2934 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
2935 // CHECK17-NEXT:    [[CALL:%.*]] = call signext i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], i32 signext [[TMP0]])
2936 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
2937 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
2938 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2939 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
2940 // CHECK17-NEXT:    [[CALL1:%.*]] = call signext i32 @_ZL7fstatici(i32 signext [[TMP2]])
2941 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
2942 // CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
2943 // CHECK17-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
2944 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
2945 // CHECK17-NEXT:    [[CALL3:%.*]] = call signext i32 @_Z9ftemplateIiET_i(i32 signext [[TMP4]])
2946 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
2947 // CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
2948 // CHECK17-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
2949 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
2950 // CHECK17-NEXT:    ret i32 [[TMP6]]
2951 //
2952 //
2953 // CHECK17-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
2954 // CHECK17-SAME: (%struct.S1* nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
2955 // CHECK17-NEXT:  entry:
2956 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
2957 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2958 // CHECK17-NEXT:    [[B:%.*]] = alloca i32, align 4
2959 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2960 // CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
2961 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
2962 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
2963 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
2964 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
2965 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
2966 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
2967 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
2968 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
2969 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2970 // CHECK17-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
2971 // CHECK17-NEXT:    store i32 1, i32* [[B]], align 4
2972 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
2973 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
2974 // CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
2975 // CHECK17-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
2976 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
2977 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
2978 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
2979 // CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
2980 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2981 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
2982 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[CONV2]], align 4
2983 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
2984 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
2985 // CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2986 // CHECK17-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
2987 // CHECK17-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 8
2988 // CHECK17-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2989 // CHECK17-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
2990 // CHECK17-NEXT:    store double* [[A]], double** [[TMP9]], align 8
2991 // CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2992 // CHECK17-NEXT:    store i8* null, i8** [[TMP10]], align 8
2993 // CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2994 // CHECK17-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
2995 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP12]], align 8
2996 // CHECK17-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2997 // CHECK17-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
2998 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP14]], align 8
2999 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3000 // CHECK17-NEXT:    store i8* null, i8** [[TMP15]], align 8
3001 // CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3002 // CHECK17-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
3003 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP17]], align 8
3004 // CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3005 // CHECK17-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
3006 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP19]], align 8
3007 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3008 // CHECK17-NEXT:    store i8* null, i8** [[TMP20]], align 8
3009 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3010 // CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3011 // CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3012 // CHECK17-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
3013 // CHECK17-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
3014 // CHECK17-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3015 // CHECK17:       omp_offload.failed:
3016 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR2:[0-9]+]]
3017 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3018 // CHECK17:       omp_offload.cont:
3019 // CHECK17-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
3020 // CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
3021 // CHECK17-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
3022 // CHECK17-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 8
3023 // CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
3024 // CHECK17-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
3025 // CHECK17-NEXT:    store double* [[A3]], double** [[TMP29]], align 8
3026 // CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
3027 // CHECK17-NEXT:    store i8* null, i8** [[TMP30]], align 8
3028 // CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
3029 // CHECK17-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
3030 // CHECK17-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
3031 // CHECK17-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
3032 // CHECK17-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
3033 // CHECK17:       omp_offload.failed7:
3034 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR2]]
3035 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
3036 // CHECK17:       omp_offload.cont8:
3037 // CHECK17-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
3038 // CHECK17-NEXT:    [[TMP35:%.*]] = load double, double* [[A9]], align 8
3039 // CHECK17-NEXT:    [[CONV10:%.*]] = fptosi double [[TMP35]] to i32
3040 // CHECK17-NEXT:    ret i32 [[CONV10]]
3041 //
3042 //
3043 // CHECK17-LABEL: define {{[^@]+}}@_ZL7fstatici
3044 // CHECK17-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
3045 // CHECK17-NEXT:  entry:
3046 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3047 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3048 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
3049 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
3050 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
3051 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
3052 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3053 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8
3054 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
3055 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
3056 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
3057 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3058 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3059 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
3060 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3061 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
3062 // CHECK17-NEXT:    store i32 [[TMP1]], i32* [[CONV]], align 4
3063 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
3064 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3065 // CHECK17-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i64*
3066 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP4]], align 8
3067 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3068 // CHECK17-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
3069 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP6]], align 8
3070 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3071 // CHECK17-NEXT:    store i8* null, i8** [[TMP7]], align 8
3072 // CHECK17-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3073 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3074 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3075 // CHECK17-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
3076 // CHECK17-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3077 // CHECK17-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3078 // CHECK17:       omp_offload.failed:
3079 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i64 [[TMP2]]) #[[ATTR2]]
3080 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3081 // CHECK17:       omp_offload.cont:
3082 // CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
3083 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
3084 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3085 // CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3086 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED2]] to i32*
3087 // CHECK17-NEXT:    store i32 [[TMP14]], i32* [[CONV3]], align 4
3088 // CHECK17-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED2]], align 8
3089 // CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
3090 // CHECK17-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
3091 // CHECK17-NEXT:    store i64 [[TMP15]], i64* [[TMP17]], align 8
3092 // CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
3093 // CHECK17-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
3094 // CHECK17-NEXT:    store i64 [[TMP15]], i64* [[TMP19]], align 8
3095 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
3096 // CHECK17-NEXT:    store i8* null, i8** [[TMP20]], align 8
3097 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
3098 // CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
3099 // CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3100 // CHECK17-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
3101 // CHECK17-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
3102 // CHECK17-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
3103 // CHECK17:       omp_offload.failed7:
3104 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i64 [[TMP15]]) #[[ATTR2]]
3105 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
3106 // CHECK17:       omp_offload.cont8:
3107 // CHECK17-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
3108 // CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP26]], 1
3109 // CHECK17-NEXT:    ret i32 [[ADD9]]
3110 //
3111 //
3112 // CHECK17-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
3113 // CHECK17-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] comdat {
3114 // CHECK17-NEXT:  entry:
3115 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3116 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
3117 // CHECK17-NEXT:    [[B:%.*]] = alloca i16, align 2
3118 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
3119 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
3120 // CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
3121 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
3122 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
3123 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
3124 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
3125 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3126 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
3127 // CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
3128 // CHECK17-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
3129 // CHECK17-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3130 // CHECK17:       omp_offload.failed:
3131 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
3132 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3133 // CHECK17:       omp_offload.cont:
3134 // CHECK17-NEXT:    store i16 1, i16* [[B]], align 2
3135 // CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
3136 // CHECK17-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
3137 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
3138 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
3139 // CHECK17-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
3140 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
3141 // CHECK17-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
3142 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_CASTED]] to i16*
3143 // CHECK17-NEXT:    store i16 [[TMP5]], i16* [[CONV1]], align 2
3144 // CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
3145 // CHECK17-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
3146 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i16*
3147 // CHECK17-NEXT:    store i16 [[TMP7]], i16* [[CONV2]], align 2
3148 // CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
3149 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3150 // CHECK17-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
3151 // CHECK17-NEXT:    store i64 [[TMP4]], i64* [[TMP10]], align 8
3152 // CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3153 // CHECK17-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
3154 // CHECK17-NEXT:    store i64 [[TMP4]], i64* [[TMP12]], align 8
3155 // CHECK17-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3156 // CHECK17-NEXT:    store i8* null, i8** [[TMP13]], align 8
3157 // CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3158 // CHECK17-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
3159 // CHECK17-NEXT:    store i64 [[TMP6]], i64* [[TMP15]], align 8
3160 // CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3161 // CHECK17-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
3162 // CHECK17-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
3163 // CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3164 // CHECK17-NEXT:    store i8* null, i8** [[TMP18]], align 8
3165 // CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3166 // CHECK17-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
3167 // CHECK17-NEXT:    store i64 [[TMP8]], i64* [[TMP20]], align 8
3168 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3169 // CHECK17-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
3170 // CHECK17-NEXT:    store i64 [[TMP8]], i64* [[TMP22]], align 8
3171 // CHECK17-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3172 // CHECK17-NEXT:    store i8* null, i8** [[TMP23]], align 8
3173 // CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3174 // CHECK17-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3175 // CHECK17-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
3176 // CHECK17-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
3177 // CHECK17-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
3178 // CHECK17-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
3179 // CHECK17-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
3180 // CHECK17:       omp_offload.failed3:
3181 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR2]]
3182 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
3183 // CHECK17:       omp_offload.cont4:
3184 // CHECK17-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
3185 // CHECK17-NEXT:    ret i32 [[TMP30]]
3186 //
3187 //
3188 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
3189 // CHECK17-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
3190 // CHECK17-NEXT:  entry:
3191 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
3192 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
3193 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3194 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
3195 // CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3196 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
3197 // CHECK17-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
3198 // CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3199 // CHECK17-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
3200 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
3201 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
3202 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
3203 // CHECK17-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
3204 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3205 // CHECK17-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 8
3206 // CHECK17-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3207 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 8
3208 // CHECK17-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 8
3209 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
3210 // CHECK17-NEXT:    ret void
3211 //
3212 //
3213 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
3214 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3215 // CHECK17-NEXT:  entry:
3216 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3217 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3218 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
3219 // CHECK17-NEXT:    [[B:%.*]] = alloca i32, align 4
3220 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3221 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3222 // CHECK17-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
3223 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
3224 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
3225 // CHECK17-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
3226 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
3227 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
3228 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
3229 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
3230 // CHECK17-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
3231 // CHECK17-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
3232 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
3233 // CHECK17-NEXT:    store double [[ADD]], double* [[A]], align 8
3234 // CHECK17-NEXT:    ret void
3235 //
3236 //
3237 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
3238 // CHECK17-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
3239 // CHECK17-NEXT:  entry:
3240 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
3241 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
3242 // CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3243 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
3244 // CHECK17-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
3245 // CHECK17-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
3246 // CHECK17-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3247 // CHECK17-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 8
3248 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
3249 // CHECK17-NEXT:    ret void
3250 //
3251 //
3252 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..1
3253 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3254 // CHECK17-NEXT:  entry:
3255 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3256 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3257 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
3258 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3259 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3260 // CHECK17-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
3261 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
3262 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
3263 // CHECK17-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
3264 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
3265 // CHECK17-NEXT:    store double 2.500000e+00, double* [[A]], align 8
3266 // CHECK17-NEXT:    ret void
3267 //
3268 //
3269 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
3270 // CHECK17-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
3271 // CHECK17-NEXT:  entry:
3272 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3273 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 1
3274 // CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3275 // CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3276 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
3277 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
3278 // CHECK17-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
3279 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
3280 // CHECK17-NEXT:    ret void
3281 //
3282 //
3283 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..4
3284 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3285 // CHECK17-NEXT:  entry:
3286 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3287 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3288 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
3289 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3290 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3291 // CHECK17-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
3292 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
3293 // CHECK17-NEXT:    ret void
3294 //
3295 //
3296 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
3297 // CHECK17-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
3298 // CHECK17-NEXT:  entry:
3299 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3300 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 1
3301 // CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3302 // CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3303 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
3304 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
3305 // CHECK17-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
3306 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
3307 // CHECK17-NEXT:    ret void
3308 //
3309 //
3310 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..7
3311 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3312 // CHECK17-NEXT:  entry:
3313 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3314 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3315 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
3316 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3317 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3318 // CHECK17-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
3319 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
3320 // CHECK17-NEXT:    ret void
3321 //
3322 //
3323 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
3324 // CHECK17-SAME: () #[[ATTR1]] {
3325 // CHECK17-NEXT:  entry:
3326 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
3327 // CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3328 // CHECK17-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
3329 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
3330 // CHECK17-NEXT:    ret void
3331 //
3332 //
3333 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..10
3334 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3335 // CHECK17-NEXT:  entry:
3336 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3337 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3338 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
3339 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3340 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3341 // CHECK17-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
3342 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
3343 // CHECK17-NEXT:    ret void
3344 //
3345 //
3346 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
3347 // CHECK17-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
3348 // CHECK17-NEXT:  entry:
3349 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
3350 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
3351 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3352 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
3353 // CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3354 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
3355 // CHECK17-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
3356 // CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3357 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
3358 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
3359 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
3360 // CHECK17-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
3361 // CHECK17-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
3362 // CHECK17-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
3363 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3364 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
3365 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
3366 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3367 // CHECK17-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV1]], align 8
3368 // CHECK17-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
3369 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
3370 // CHECK17-NEXT:    ret void
3371 //
3372 //
3373 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..11
3374 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3375 // CHECK17-NEXT:  entry:
3376 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3377 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3378 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
3379 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
3380 // CHECK17-NEXT:    [[B:%.*]] = alloca i16, align 2
3381 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3382 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3383 // CHECK17-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
3384 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
3385 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
3386 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3387 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
3388 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
3389 // CHECK17-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
3390 // CHECK17-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
3391 // CHECK17-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
3392 // CHECK17-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
3393 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
3394 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
3395 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
3396 // CHECK17-NEXT:    ret void
3397 //
3398 //
3399 // CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3400 // CHECK17-SAME: () #[[ATTR3:[0-9]+]] {
3401 // CHECK17-NEXT:  entry:
3402 // CHECK17-NEXT:    call void @__tgt_register_requires(i64 1)
3403 // CHECK17-NEXT:    ret void
3404 //
3405 //
3406 // CHECK18-LABEL: define {{[^@]+}}@_Z3bari
3407 // CHECK18-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
3408 // CHECK18-NEXT:  entry:
3409 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3410 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
3411 // CHECK18-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
3412 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3413 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
3414 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3415 // CHECK18-NEXT:    [[CALL:%.*]] = call signext i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], i32 signext [[TMP0]])
3416 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
3417 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
3418 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
3419 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
3420 // CHECK18-NEXT:    [[CALL1:%.*]] = call signext i32 @_ZL7fstatici(i32 signext [[TMP2]])
3421 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
3422 // CHECK18-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
3423 // CHECK18-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
3424 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
3425 // CHECK18-NEXT:    [[CALL3:%.*]] = call signext i32 @_Z9ftemplateIiET_i(i32 signext [[TMP4]])
3426 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
3427 // CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
3428 // CHECK18-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
3429 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
3430 // CHECK18-NEXT:    ret i32 [[TMP6]]
3431 //
3432 //
3433 // CHECK18-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
3434 // CHECK18-SAME: (%struct.S1* nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
3435 // CHECK18-NEXT:  entry:
3436 // CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
3437 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3438 // CHECK18-NEXT:    [[B:%.*]] = alloca i32, align 4
3439 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3440 // CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
3441 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
3442 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
3443 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
3444 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
3445 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
3446 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
3447 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
3448 // CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
3449 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3450 // CHECK18-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
3451 // CHECK18-NEXT:    store i32 1, i32* [[B]], align 4
3452 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3453 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
3454 // CHECK18-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
3455 // CHECK18-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
3456 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
3457 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
3458 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[CONV]], align 4
3459 // CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[B_CASTED]], align 8
3460 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3461 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
3462 // CHECK18-NEXT:    store i32 [[TMP4]], i32* [[CONV2]], align 4
3463 // CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
3464 // CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
3465 // CHECK18-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3466 // CHECK18-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
3467 // CHECK18-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 8
3468 // CHECK18-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3469 // CHECK18-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
3470 // CHECK18-NEXT:    store double* [[A]], double** [[TMP9]], align 8
3471 // CHECK18-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3472 // CHECK18-NEXT:    store i8* null, i8** [[TMP10]], align 8
3473 // CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3474 // CHECK18-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
3475 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP12]], align 8
3476 // CHECK18-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3477 // CHECK18-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
3478 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP14]], align 8
3479 // CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3480 // CHECK18-NEXT:    store i8* null, i8** [[TMP15]], align 8
3481 // CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3482 // CHECK18-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
3483 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP17]], align 8
3484 // CHECK18-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3485 // CHECK18-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
3486 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP19]], align 8
3487 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3488 // CHECK18-NEXT:    store i8* null, i8** [[TMP20]], align 8
3489 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3490 // CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3491 // CHECK18-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3492 // CHECK18-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
3493 // CHECK18-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
3494 // CHECK18-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3495 // CHECK18:       omp_offload.failed:
3496 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR2:[0-9]+]]
3497 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3498 // CHECK18:       omp_offload.cont:
3499 // CHECK18-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
3500 // CHECK18-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
3501 // CHECK18-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
3502 // CHECK18-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 8
3503 // CHECK18-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
3504 // CHECK18-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
3505 // CHECK18-NEXT:    store double* [[A3]], double** [[TMP29]], align 8
3506 // CHECK18-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
3507 // CHECK18-NEXT:    store i8* null, i8** [[TMP30]], align 8
3508 // CHECK18-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
3509 // CHECK18-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
3510 // CHECK18-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
3511 // CHECK18-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
3512 // CHECK18-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
3513 // CHECK18:       omp_offload.failed7:
3514 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR2]]
3515 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
3516 // CHECK18:       omp_offload.cont8:
3517 // CHECK18-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
3518 // CHECK18-NEXT:    [[TMP35:%.*]] = load double, double* [[A9]], align 8
3519 // CHECK18-NEXT:    [[CONV10:%.*]] = fptosi double [[TMP35]] to i32
3520 // CHECK18-NEXT:    ret i32 [[CONV10]]
3521 //
3522 //
3523 // CHECK18-LABEL: define {{[^@]+}}@_ZL7fstatici
3524 // CHECK18-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
3525 // CHECK18-NEXT:  entry:
3526 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3527 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3528 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
3529 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
3530 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
3531 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
3532 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3533 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8
3534 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [1 x i8*], align 8
3535 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [1 x i8*], align 8
3536 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [1 x i8*], align 8
3537 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3538 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3539 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
3540 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3541 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
3542 // CHECK18-NEXT:    store i32 [[TMP1]], i32* [[CONV]], align 4
3543 // CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
3544 // CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3545 // CHECK18-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i64*
3546 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP4]], align 8
3547 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3548 // CHECK18-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
3549 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP6]], align 8
3550 // CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3551 // CHECK18-NEXT:    store i8* null, i8** [[TMP7]], align 8
3552 // CHECK18-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3553 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3554 // CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3555 // CHECK18-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
3556 // CHECK18-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3557 // CHECK18-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3558 // CHECK18:       omp_offload.failed:
3559 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i64 [[TMP2]]) #[[ATTR2]]
3560 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3561 // CHECK18:       omp_offload.cont:
3562 // CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
3563 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
3564 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3565 // CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3566 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED2]] to i32*
3567 // CHECK18-NEXT:    store i32 [[TMP14]], i32* [[CONV3]], align 4
3568 // CHECK18-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED2]], align 8
3569 // CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
3570 // CHECK18-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
3571 // CHECK18-NEXT:    store i64 [[TMP15]], i64* [[TMP17]], align 8
3572 // CHECK18-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
3573 // CHECK18-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
3574 // CHECK18-NEXT:    store i64 [[TMP15]], i64* [[TMP19]], align 8
3575 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
3576 // CHECK18-NEXT:    store i8* null, i8** [[TMP20]], align 8
3577 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
3578 // CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
3579 // CHECK18-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3580 // CHECK18-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
3581 // CHECK18-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
3582 // CHECK18-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
3583 // CHECK18:       omp_offload.failed7:
3584 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i64 [[TMP15]]) #[[ATTR2]]
3585 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
3586 // CHECK18:       omp_offload.cont8:
3587 // CHECK18-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
3588 // CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP26]], 1
3589 // CHECK18-NEXT:    ret i32 [[ADD9]]
3590 //
3591 //
3592 // CHECK18-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
3593 // CHECK18-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] comdat {
3594 // CHECK18-NEXT:  entry:
3595 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3596 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
3597 // CHECK18-NEXT:    [[B:%.*]] = alloca i16, align 2
3598 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
3599 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
3600 // CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
3601 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
3602 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
3603 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
3604 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
3605 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3606 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
3607 // CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
3608 // CHECK18-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
3609 // CHECK18-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3610 // CHECK18:       omp_offload.failed:
3611 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
3612 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3613 // CHECK18:       omp_offload.cont:
3614 // CHECK18-NEXT:    store i16 1, i16* [[B]], align 2
3615 // CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
3616 // CHECK18-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
3617 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
3618 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
3619 // CHECK18-NEXT:    store i32 [[TMP3]], i32* [[CONV]], align 4
3620 // CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
3621 // CHECK18-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
3622 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_CASTED]] to i16*
3623 // CHECK18-NEXT:    store i16 [[TMP5]], i16* [[CONV1]], align 2
3624 // CHECK18-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
3625 // CHECK18-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
3626 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i16*
3627 // CHECK18-NEXT:    store i16 [[TMP7]], i16* [[CONV2]], align 2
3628 // CHECK18-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
3629 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3630 // CHECK18-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
3631 // CHECK18-NEXT:    store i64 [[TMP4]], i64* [[TMP10]], align 8
3632 // CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3633 // CHECK18-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
3634 // CHECK18-NEXT:    store i64 [[TMP4]], i64* [[TMP12]], align 8
3635 // CHECK18-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3636 // CHECK18-NEXT:    store i8* null, i8** [[TMP13]], align 8
3637 // CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3638 // CHECK18-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
3639 // CHECK18-NEXT:    store i64 [[TMP6]], i64* [[TMP15]], align 8
3640 // CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3641 // CHECK18-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
3642 // CHECK18-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
3643 // CHECK18-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3644 // CHECK18-NEXT:    store i8* null, i8** [[TMP18]], align 8
3645 // CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3646 // CHECK18-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
3647 // CHECK18-NEXT:    store i64 [[TMP8]], i64* [[TMP20]], align 8
3648 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3649 // CHECK18-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
3650 // CHECK18-NEXT:    store i64 [[TMP8]], i64* [[TMP22]], align 8
3651 // CHECK18-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3652 // CHECK18-NEXT:    store i8* null, i8** [[TMP23]], align 8
3653 // CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3654 // CHECK18-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3655 // CHECK18-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
3656 // CHECK18-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
3657 // CHECK18-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
3658 // CHECK18-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
3659 // CHECK18-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
3660 // CHECK18:       omp_offload.failed3:
3661 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR2]]
3662 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
3663 // CHECK18:       omp_offload.cont4:
3664 // CHECK18-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
3665 // CHECK18-NEXT:    ret i32 [[TMP30]]
3666 //
3667 //
3668 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
3669 // CHECK18-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
3670 // CHECK18-NEXT:  entry:
3671 // CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
3672 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
3673 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3674 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
3675 // CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3676 // CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
3677 // CHECK18-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
3678 // CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3679 // CHECK18-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
3680 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
3681 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
3682 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
3683 // CHECK18-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
3684 // CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3685 // CHECK18-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 8
3686 // CHECK18-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3687 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 8
3688 // CHECK18-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 8
3689 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
3690 // CHECK18-NEXT:    ret void
3691 //
3692 //
3693 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined.
3694 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3695 // CHECK18-NEXT:  entry:
3696 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3697 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3698 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
3699 // CHECK18-NEXT:    [[B:%.*]] = alloca i32, align 4
3700 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3701 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3702 // CHECK18-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
3703 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
3704 // CHECK18-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
3705 // CHECK18-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
3706 // CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
3707 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
3708 // CHECK18-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
3709 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
3710 // CHECK18-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
3711 // CHECK18-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
3712 // CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
3713 // CHECK18-NEXT:    store double [[ADD]], double* [[A]], align 8
3714 // CHECK18-NEXT:    ret void
3715 //
3716 //
3717 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
3718 // CHECK18-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
3719 // CHECK18-NEXT:  entry:
3720 // CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
3721 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
3722 // CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3723 // CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
3724 // CHECK18-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
3725 // CHECK18-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
3726 // CHECK18-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3727 // CHECK18-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 8
3728 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
3729 // CHECK18-NEXT:    ret void
3730 //
3731 //
3732 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..1
3733 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3734 // CHECK18-NEXT:  entry:
3735 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3736 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3737 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
3738 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3739 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3740 // CHECK18-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
3741 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
3742 // CHECK18-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
3743 // CHECK18-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
3744 // CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
3745 // CHECK18-NEXT:    store double 2.500000e+00, double* [[A]], align 8
3746 // CHECK18-NEXT:    ret void
3747 //
3748 //
3749 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
3750 // CHECK18-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
3751 // CHECK18-NEXT:  entry:
3752 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3753 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 1
3754 // CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3755 // CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3756 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
3757 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
3758 // CHECK18-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
3759 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
3760 // CHECK18-NEXT:    ret void
3761 //
3762 //
3763 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..4
3764 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3765 // CHECK18-NEXT:  entry:
3766 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3767 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3768 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
3769 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3770 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3771 // CHECK18-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
3772 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
3773 // CHECK18-NEXT:    ret void
3774 //
3775 //
3776 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
3777 // CHECK18-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
3778 // CHECK18-NEXT:  entry:
3779 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3780 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 1
3781 // CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3782 // CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3783 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
3784 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
3785 // CHECK18-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
3786 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
3787 // CHECK18-NEXT:    ret void
3788 //
3789 //
3790 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..7
3791 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3792 // CHECK18-NEXT:  entry:
3793 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3794 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3795 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
3796 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3797 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3798 // CHECK18-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
3799 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
3800 // CHECK18-NEXT:    ret void
3801 //
3802 //
3803 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
3804 // CHECK18-SAME: () #[[ATTR1]] {
3805 // CHECK18-NEXT:  entry:
3806 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
3807 // CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3808 // CHECK18-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
3809 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
3810 // CHECK18-NEXT:    ret void
3811 //
3812 //
3813 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..10
3814 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3815 // CHECK18-NEXT:  entry:
3816 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3817 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3818 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
3819 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3820 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3821 // CHECK18-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
3822 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
3823 // CHECK18-NEXT:    ret void
3824 //
3825 //
3826 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
3827 // CHECK18-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
3828 // CHECK18-NEXT:  entry:
3829 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
3830 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
3831 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3832 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
3833 // CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3834 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
3835 // CHECK18-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
3836 // CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3837 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
3838 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
3839 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
3840 // CHECK18-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
3841 // CHECK18-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
3842 // CHECK18-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
3843 // CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3844 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
3845 // CHECK18-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
3846 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3847 // CHECK18-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV1]], align 8
3848 // CHECK18-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
3849 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
3850 // CHECK18-NEXT:    ret void
3851 //
3852 //
3853 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..11
3854 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3855 // CHECK18-NEXT:  entry:
3856 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3857 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3858 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
3859 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
3860 // CHECK18-NEXT:    [[B:%.*]] = alloca i16, align 2
3861 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3862 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3863 // CHECK18-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
3864 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
3865 // CHECK18-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
3866 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3867 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
3868 // CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
3869 // CHECK18-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
3870 // CHECK18-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
3871 // CHECK18-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
3872 // CHECK18-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
3873 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
3874 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
3875 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
3876 // CHECK18-NEXT:    ret void
3877 //
3878 //
3879 // CHECK18-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3880 // CHECK18-SAME: () #[[ATTR3:[0-9]+]] {
3881 // CHECK18-NEXT:  entry:
3882 // CHECK18-NEXT:    call void @__tgt_register_requires(i64 1)
3883 // CHECK18-NEXT:    ret void
3884 //
3885 //
3886 // CHECK19-LABEL: define {{[^@]+}}@_Z3bari
3887 // CHECK19-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
3888 // CHECK19-NEXT:  entry:
3889 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3890 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
3891 // CHECK19-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
3892 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3893 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
3894 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3895 // CHECK19-NEXT:    [[CALL:%.*]] = call i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], i32 [[TMP0]])
3896 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
3897 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
3898 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
3899 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
3900 // CHECK19-NEXT:    [[CALL1:%.*]] = call i32 @_ZL7fstatici(i32 [[TMP2]])
3901 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
3902 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
3903 // CHECK19-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
3904 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
3905 // CHECK19-NEXT:    [[CALL3:%.*]] = call i32 @_Z9ftemplateIiET_i(i32 [[TMP4]])
3906 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
3907 // CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
3908 // CHECK19-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
3909 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
3910 // CHECK19-NEXT:    ret i32 [[TMP6]]
3911 //
3912 //
3913 // CHECK19-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
3914 // CHECK19-SAME: (%struct.S1* nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 [[N:%.*]]) #[[ATTR0]] comdat align 2 {
3915 // CHECK19-NEXT:  entry:
3916 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
3917 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3918 // CHECK19-NEXT:    [[B:%.*]] = alloca i32, align 4
3919 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3920 // CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
3921 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
3922 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
3923 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
3924 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
3925 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
3926 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
3927 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
3928 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
3929 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3930 // CHECK19-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
3931 // CHECK19-NEXT:    store i32 1, i32* [[B]], align 4
3932 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3933 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
3934 // CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
3935 // CHECK19-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
3936 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
3937 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[B_CASTED]], align 4
3938 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
3939 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3940 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
3941 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
3942 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
3943 // CHECK19-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3944 // CHECK19-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
3945 // CHECK19-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 4
3946 // CHECK19-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3947 // CHECK19-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
3948 // CHECK19-NEXT:    store double* [[A]], double** [[TMP9]], align 4
3949 // CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3950 // CHECK19-NEXT:    store i8* null, i8** [[TMP10]], align 4
3951 // CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3952 // CHECK19-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
3953 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP12]], align 4
3954 // CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3955 // CHECK19-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
3956 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP14]], align 4
3957 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3958 // CHECK19-NEXT:    store i8* null, i8** [[TMP15]], align 4
3959 // CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3960 // CHECK19-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
3961 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
3962 // CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3963 // CHECK19-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
3964 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
3965 // CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3966 // CHECK19-NEXT:    store i8* null, i8** [[TMP20]], align 4
3967 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3968 // CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3969 // CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3970 // CHECK19-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
3971 // CHECK19-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
3972 // CHECK19-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3973 // CHECK19:       omp_offload.failed:
3974 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR2:[0-9]+]]
3975 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3976 // CHECK19:       omp_offload.cont:
3977 // CHECK19-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
3978 // CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
3979 // CHECK19-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
3980 // CHECK19-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 4
3981 // CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
3982 // CHECK19-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
3983 // CHECK19-NEXT:    store double* [[A2]], double** [[TMP29]], align 4
3984 // CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
3985 // CHECK19-NEXT:    store i8* null, i8** [[TMP30]], align 4
3986 // CHECK19-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
3987 // CHECK19-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
3988 // CHECK19-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
3989 // CHECK19-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
3990 // CHECK19-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
3991 // CHECK19:       omp_offload.failed6:
3992 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR2]]
3993 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
3994 // CHECK19:       omp_offload.cont7:
3995 // CHECK19-NEXT:    [[A8:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
3996 // CHECK19-NEXT:    [[TMP35:%.*]] = load double, double* [[A8]], align 4
3997 // CHECK19-NEXT:    [[CONV:%.*]] = fptosi double [[TMP35]] to i32
3998 // CHECK19-NEXT:    ret i32 [[CONV]]
3999 //
4000 //
4001 // CHECK19-LABEL: define {{[^@]+}}@_ZL7fstatici
4002 // CHECK19-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
4003 // CHECK19-NEXT:  entry:
4004 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4005 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4006 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4007 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
4008 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
4009 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
4010 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4011 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
4012 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
4013 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
4014 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
4015 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4016 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
4017 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
4018 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4019 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4020 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4021 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4022 // CHECK19-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
4023 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[TMP4]], align 4
4024 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4025 // CHECK19-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
4026 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[TMP6]], align 4
4027 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4028 // CHECK19-NEXT:    store i8* null, i8** [[TMP7]], align 4
4029 // CHECK19-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4030 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4031 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4032 // CHECK19-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
4033 // CHECK19-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4034 // CHECK19-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4035 // CHECK19:       omp_offload.failed:
4036 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i32 [[TMP2]]) #[[ATTR2]]
4037 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4038 // CHECK19:       omp_offload.cont:
4039 // CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
4040 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
4041 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4042 // CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4043 // CHECK19-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
4044 // CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
4045 // CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
4046 // CHECK19-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
4047 // CHECK19-NEXT:    store i32 [[TMP15]], i32* [[TMP17]], align 4
4048 // CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
4049 // CHECK19-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
4050 // CHECK19-NEXT:    store i32 [[TMP15]], i32* [[TMP19]], align 4
4051 // CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
4052 // CHECK19-NEXT:    store i8* null, i8** [[TMP20]], align 4
4053 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
4054 // CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
4055 // CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4056 // CHECK19-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
4057 // CHECK19-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
4058 // CHECK19-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
4059 // CHECK19:       omp_offload.failed6:
4060 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i32 [[TMP15]]) #[[ATTR2]]
4061 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
4062 // CHECK19:       omp_offload.cont7:
4063 // CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
4064 // CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP26]], 1
4065 // CHECK19-NEXT:    ret i32 [[ADD8]]
4066 //
4067 //
4068 // CHECK19-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
4069 // CHECK19-SAME: (i32 [[N:%.*]]) #[[ATTR0]] comdat {
4070 // CHECK19-NEXT:  entry:
4071 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4072 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
4073 // CHECK19-NEXT:    [[B:%.*]] = alloca i16, align 2
4074 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
4075 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4076 // CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
4077 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4078 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
4079 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
4080 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
4081 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4082 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
4083 // CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
4084 // CHECK19-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
4085 // CHECK19-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4086 // CHECK19:       omp_offload.failed:
4087 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
4088 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4089 // CHECK19:       omp_offload.cont:
4090 // CHECK19-NEXT:    store i16 1, i16* [[B]], align 2
4091 // CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
4092 // CHECK19-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
4093 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
4094 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
4095 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
4096 // CHECK19-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
4097 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
4098 // CHECK19-NEXT:    store i16 [[TMP5]], i16* [[CONV]], align 2
4099 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
4100 // CHECK19-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
4101 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i16*
4102 // CHECK19-NEXT:    store i16 [[TMP7]], i16* [[CONV1]], align 2
4103 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4104 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4105 // CHECK19-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
4106 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[TMP10]], align 4
4107 // CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4108 // CHECK19-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
4109 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[TMP12]], align 4
4110 // CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4111 // CHECK19-NEXT:    store i8* null, i8** [[TMP13]], align 4
4112 // CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4113 // CHECK19-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
4114 // CHECK19-NEXT:    store i32 [[TMP6]], i32* [[TMP15]], align 4
4115 // CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4116 // CHECK19-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
4117 // CHECK19-NEXT:    store i32 [[TMP6]], i32* [[TMP17]], align 4
4118 // CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4119 // CHECK19-NEXT:    store i8* null, i8** [[TMP18]], align 4
4120 // CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4121 // CHECK19-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
4122 // CHECK19-NEXT:    store i32 [[TMP8]], i32* [[TMP20]], align 4
4123 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4124 // CHECK19-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
4125 // CHECK19-NEXT:    store i32 [[TMP8]], i32* [[TMP22]], align 4
4126 // CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4127 // CHECK19-NEXT:    store i8* null, i8** [[TMP23]], align 4
4128 // CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4129 // CHECK19-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4130 // CHECK19-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
4131 // CHECK19-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
4132 // CHECK19-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
4133 // CHECK19-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
4134 // CHECK19-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
4135 // CHECK19:       omp_offload.failed2:
4136 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR2]]
4137 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
4138 // CHECK19:       omp_offload.cont3:
4139 // CHECK19-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
4140 // CHECK19-NEXT:    ret i32 [[TMP30]]
4141 //
4142 //
4143 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
4144 // CHECK19-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
4145 // CHECK19-NEXT:  entry:
4146 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
4147 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
4148 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4149 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
4150 // CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4151 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
4152 // CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
4153 // CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4154 // CHECK19-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
4155 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4156 // CHECK19-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
4157 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4158 // CHECK19-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 4
4159 // CHECK19-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4160 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_ADDR]], align 4
4161 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 4
4162 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
4163 // CHECK19-NEXT:    ret void
4164 //
4165 //
4166 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined.
4167 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4168 // CHECK19-NEXT:  entry:
4169 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4170 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4171 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
4172 // CHECK19-NEXT:    [[B:%.*]] = alloca i32, align 4
4173 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4174 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4175 // CHECK19-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
4176 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
4177 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
4178 // CHECK19-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
4179 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
4180 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4181 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
4182 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
4183 // CHECK19-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
4184 // CHECK19-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
4185 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
4186 // CHECK19-NEXT:    store double [[ADD]], double* [[A]], align 4
4187 // CHECK19-NEXT:    ret void
4188 //
4189 //
4190 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
4191 // CHECK19-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
4192 // CHECK19-NEXT:  entry:
4193 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
4194 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 4
4195 // CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4196 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
4197 // CHECK19-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
4198 // CHECK19-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
4199 // CHECK19-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4200 // CHECK19-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 4
4201 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
4202 // CHECK19-NEXT:    ret void
4203 //
4204 //
4205 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..1
4206 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4207 // CHECK19-NEXT:  entry:
4208 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4209 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4210 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4
4211 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4212 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4213 // CHECK19-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4
4214 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4
4215 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
4216 // CHECK19-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
4217 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
4218 // CHECK19-NEXT:    store double 2.500000e+00, double* [[A]], align 4
4219 // CHECK19-NEXT:    ret void
4220 //
4221 //
4222 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
4223 // CHECK19-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
4224 // CHECK19-NEXT:  entry:
4225 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4226 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 1
4227 // CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4228 // CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4229 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4230 // CHECK19-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
4231 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
4232 // CHECK19-NEXT:    ret void
4233 //
4234 //
4235 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..4
4236 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4237 // CHECK19-NEXT:  entry:
4238 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4239 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4240 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
4241 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4242 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4243 // CHECK19-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
4244 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
4245 // CHECK19-NEXT:    ret void
4246 //
4247 //
4248 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
4249 // CHECK19-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
4250 // CHECK19-NEXT:  entry:
4251 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4252 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 1
4253 // CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4254 // CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4255 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4256 // CHECK19-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
4257 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
4258 // CHECK19-NEXT:    ret void
4259 //
4260 //
4261 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..7
4262 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4263 // CHECK19-NEXT:  entry:
4264 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4265 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4266 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
4267 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4268 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4269 // CHECK19-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
4270 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
4271 // CHECK19-NEXT:    ret void
4272 //
4273 //
4274 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
4275 // CHECK19-SAME: () #[[ATTR1]] {
4276 // CHECK19-NEXT:  entry:
4277 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
4278 // CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4279 // CHECK19-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
4280 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
4281 // CHECK19-NEXT:    ret void
4282 //
4283 //
4284 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..10
4285 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4286 // CHECK19-NEXT:  entry:
4287 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4288 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4289 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
4290 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4291 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4292 // CHECK19-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
4293 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
4294 // CHECK19-NEXT:    ret void
4295 //
4296 //
4297 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
4298 // CHECK19-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
4299 // CHECK19-NEXT:  entry:
4300 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4301 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
4302 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4303 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
4304 // CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4305 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4306 // CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
4307 // CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4308 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
4309 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
4310 // CHECK19-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
4311 // CHECK19-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
4312 // CHECK19-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
4313 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4314 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_ADDR]], align 4
4315 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
4316 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4317 // CHECK19-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV]], align 4
4318 // CHECK19-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
4319 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
4320 // CHECK19-NEXT:    ret void
4321 //
4322 //
4323 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..11
4324 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4325 // CHECK19-NEXT:  entry:
4326 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4327 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4328 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
4329 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
4330 // CHECK19-NEXT:    [[B:%.*]] = alloca i16, align 2
4331 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4332 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4333 // CHECK19-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
4334 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
4335 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
4336 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4337 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
4338 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
4339 // CHECK19-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
4340 // CHECK19-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
4341 // CHECK19-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
4342 // CHECK19-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
4343 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
4344 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
4345 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
4346 // CHECK19-NEXT:    ret void
4347 //
4348 //
4349 // CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4350 // CHECK19-SAME: () #[[ATTR3:[0-9]+]] {
4351 // CHECK19-NEXT:  entry:
4352 // CHECK19-NEXT:    call void @__tgt_register_requires(i64 1)
4353 // CHECK19-NEXT:    ret void
4354 //
4355 //
4356 // CHECK20-LABEL: define {{[^@]+}}@_Z3bari
4357 // CHECK20-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
4358 // CHECK20-NEXT:  entry:
4359 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4360 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
4361 // CHECK20-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
4362 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4363 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
4364 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
4365 // CHECK20-NEXT:    [[CALL:%.*]] = call i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], i32 [[TMP0]])
4366 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
4367 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
4368 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
4369 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
4370 // CHECK20-NEXT:    [[CALL1:%.*]] = call i32 @_ZL7fstatici(i32 [[TMP2]])
4371 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
4372 // CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
4373 // CHECK20-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
4374 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
4375 // CHECK20-NEXT:    [[CALL3:%.*]] = call i32 @_Z9ftemplateIiET_i(i32 [[TMP4]])
4376 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
4377 // CHECK20-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
4378 // CHECK20-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
4379 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
4380 // CHECK20-NEXT:    ret i32 [[TMP6]]
4381 //
4382 //
4383 // CHECK20-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
4384 // CHECK20-SAME: (%struct.S1* nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 [[N:%.*]]) #[[ATTR0]] comdat align 2 {
4385 // CHECK20-NEXT:  entry:
4386 // CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
4387 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4388 // CHECK20-NEXT:    [[B:%.*]] = alloca i32, align 4
4389 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4390 // CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
4391 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4392 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
4393 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
4394 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
4395 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
4396 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
4397 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
4398 // CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
4399 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4400 // CHECK20-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
4401 // CHECK20-NEXT:    store i32 1, i32* [[B]], align 4
4402 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
4403 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[B]], align 4
4404 // CHECK20-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
4405 // CHECK20-NEXT:    store i32 [[SUB]], i32* [[DOTCAPTURE_EXPR_]], align 4
4406 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[B]], align 4
4407 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[B_CASTED]], align 4
4408 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[B_CASTED]], align 4
4409 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4410 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4411 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4412 // CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
4413 // CHECK20-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4414 // CHECK20-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to %struct.S1**
4415 // CHECK20-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP7]], align 4
4416 // CHECK20-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4417 // CHECK20-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to double**
4418 // CHECK20-NEXT:    store double* [[A]], double** [[TMP9]], align 4
4419 // CHECK20-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4420 // CHECK20-NEXT:    store i8* null, i8** [[TMP10]], align 4
4421 // CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4422 // CHECK20-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
4423 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP12]], align 4
4424 // CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4425 // CHECK20-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
4426 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP14]], align 4
4427 // CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4428 // CHECK20-NEXT:    store i8* null, i8** [[TMP15]], align 4
4429 // CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4430 // CHECK20-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
4431 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
4432 // CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4433 // CHECK20-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
4434 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
4435 // CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4436 // CHECK20-NEXT:    store i8* null, i8** [[TMP20]], align 4
4437 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4438 // CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4439 // CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4440 // CHECK20-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1:[0-9]+]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, i32 3, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
4441 // CHECK20-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
4442 // CHECK20-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4443 // CHECK20:       omp_offload.failed:
4444 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(%struct.S1* [[THIS1]], i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR2:[0-9]+]]
4445 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4446 // CHECK20:       omp_offload.cont:
4447 // CHECK20-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
4448 // CHECK20-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
4449 // CHECK20-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to %struct.S1**
4450 // CHECK20-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP27]], align 4
4451 // CHECK20-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
4452 // CHECK20-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to double**
4453 // CHECK20-NEXT:    store double* [[A2]], double** [[TMP29]], align 4
4454 // CHECK20-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
4455 // CHECK20-NEXT:    store i8* null, i8** [[TMP30]], align 4
4456 // CHECK20-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
4457 // CHECK20-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
4458 // CHECK20-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, i32 1, i8** [[TMP31]], i8** [[TMP32]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 1, i32 1024)
4459 // CHECK20-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
4460 // CHECK20-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
4461 // CHECK20:       omp_offload.failed6:
4462 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(%struct.S1* [[THIS1]]) #[[ATTR2]]
4463 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
4464 // CHECK20:       omp_offload.cont7:
4465 // CHECK20-NEXT:    [[A8:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
4466 // CHECK20-NEXT:    [[TMP35:%.*]] = load double, double* [[A8]], align 4
4467 // CHECK20-NEXT:    [[CONV:%.*]] = fptosi double [[TMP35]] to i32
4468 // CHECK20-NEXT:    ret i32 [[CONV]]
4469 //
4470 //
4471 // CHECK20-LABEL: define {{[^@]+}}@_ZL7fstatici
4472 // CHECK20-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
4473 // CHECK20-NEXT:  entry:
4474 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4475 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4476 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4477 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
4478 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
4479 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
4480 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4481 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
4482 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
4483 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
4484 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
4485 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4486 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
4487 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[DOTCAPTURE_EXPR_]], align 4
4488 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4489 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4490 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4491 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4492 // CHECK20-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
4493 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[TMP4]], align 4
4494 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4495 // CHECK20-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
4496 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[TMP6]], align 4
4497 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4498 // CHECK20-NEXT:    store i8* null, i8** [[TMP7]], align 4
4499 // CHECK20-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4500 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4501 // CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4502 // CHECK20-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP10]])
4503 // CHECK20-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4504 // CHECK20-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4505 // CHECK20:       omp_offload.failed:
4506 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i32 [[TMP2]]) #[[ATTR2]]
4507 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4508 // CHECK20:       omp_offload.cont:
4509 // CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
4510 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 32, [[TMP13]]
4511 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4512 // CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4513 // CHECK20-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
4514 // CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED2]], align 4
4515 // CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
4516 // CHECK20-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
4517 // CHECK20-NEXT:    store i32 [[TMP15]], i32* [[TMP17]], align 4
4518 // CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
4519 // CHECK20-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
4520 // CHECK20-NEXT:    store i32 [[TMP15]], i32* [[TMP19]], align 4
4521 // CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
4522 // CHECK20-NEXT:    store i8* null, i8** [[TMP20]], align 4
4523 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
4524 // CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
4525 // CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4526 // CHECK20-NEXT:    [[TMP24:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, i32 1, i8** [[TMP21]], i8** [[TMP22]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP23]])
4527 // CHECK20-NEXT:    [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
4528 // CHECK20-NEXT:    br i1 [[TMP25]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
4529 // CHECK20:       omp_offload.failed6:
4530 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i32 [[TMP15]]) #[[ATTR2]]
4531 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
4532 // CHECK20:       omp_offload.cont7:
4533 // CHECK20-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
4534 // CHECK20-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP26]], 1
4535 // CHECK20-NEXT:    ret i32 [[ADD8]]
4536 //
4537 //
4538 // CHECK20-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
4539 // CHECK20-SAME: (i32 [[N:%.*]]) #[[ATTR0]] comdat {
4540 // CHECK20-NEXT:  entry:
4541 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4542 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
4543 // CHECK20-NEXT:    [[B:%.*]] = alloca i16, align 2
4544 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
4545 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4546 // CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
4547 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4548 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
4549 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
4550 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
4551 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4552 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
4553 // CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 20)
4554 // CHECK20-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
4555 // CHECK20-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4556 // CHECK20:       omp_offload.failed:
4557 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
4558 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4559 // CHECK20:       omp_offload.cont:
4560 // CHECK20-NEXT:    store i16 1, i16* [[B]], align 2
4561 // CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[B]], align 2
4562 // CHECK20-NEXT:    store i16 [[TMP2]], i16* [[DOTCAPTURE_EXPR_]], align 2
4563 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
4564 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
4565 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
4566 // CHECK20-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
4567 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_CASTED]] to i16*
4568 // CHECK20-NEXT:    store i16 [[TMP5]], i16* [[CONV]], align 2
4569 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
4570 // CHECK20-NEXT:    [[TMP7:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
4571 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i16*
4572 // CHECK20-NEXT:    store i16 [[TMP7]], i16* [[CONV1]], align 2
4573 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4574 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4575 // CHECK20-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
4576 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[TMP10]], align 4
4577 // CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4578 // CHECK20-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
4579 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[TMP12]], align 4
4580 // CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4581 // CHECK20-NEXT:    store i8* null, i8** [[TMP13]], align 4
4582 // CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4583 // CHECK20-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
4584 // CHECK20-NEXT:    store i32 [[TMP6]], i32* [[TMP15]], align 4
4585 // CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4586 // CHECK20-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
4587 // CHECK20-NEXT:    store i32 [[TMP6]], i32* [[TMP17]], align 4
4588 // CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4589 // CHECK20-NEXT:    store i8* null, i8** [[TMP18]], align 4
4590 // CHECK20-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4591 // CHECK20-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
4592 // CHECK20-NEXT:    store i32 [[TMP8]], i32* [[TMP20]], align 4
4593 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4594 // CHECK20-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
4595 // CHECK20-NEXT:    store i32 [[TMP8]], i32* [[TMP22]], align 4
4596 // CHECK20-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4597 // CHECK20-NEXT:    store i8* null, i8** [[TMP23]], align 4
4598 // CHECK20-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4599 // CHECK20-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4600 // CHECK20-NEXT:    [[TMP26:%.*]] = load i16, i16* [[DOTCAPTURE_EXPR_]], align 2
4601 // CHECK20-NEXT:    [[TMP27:%.*]] = zext i16 [[TMP26]] to i32
4602 // CHECK20-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP27]])
4603 // CHECK20-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
4604 // CHECK20-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
4605 // CHECK20:       omp_offload.failed2:
4606 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR2]]
4607 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
4608 // CHECK20:       omp_offload.cont3:
4609 // CHECK20-NEXT:    [[TMP30:%.*]] = load i32, i32* [[A]], align 4
4610 // CHECK20-NEXT:    ret i32 [[TMP30]]
4611 //
4612 //
4613 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
4614 // CHECK20-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
4615 // CHECK20-NEXT:  entry:
4616 // CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
4617 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
4618 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4619 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
4620 // CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4621 // CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
4622 // CHECK20-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
4623 // CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4624 // CHECK20-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
4625 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4626 // CHECK20-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
4627 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4628 // CHECK20-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 4
4629 // CHECK20-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4630 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_ADDR]], align 4
4631 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 4
4632 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
4633 // CHECK20-NEXT:    ret void
4634 //
4635 //
4636 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined.
4637 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4638 // CHECK20-NEXT:  entry:
4639 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4640 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4641 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
4642 // CHECK20-NEXT:    [[B:%.*]] = alloca i32, align 4
4643 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4644 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4645 // CHECK20-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
4646 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
4647 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
4648 // CHECK20-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
4649 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
4650 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4651 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
4652 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
4653 // CHECK20-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
4654 // CHECK20-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
4655 // CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
4656 // CHECK20-NEXT:    store double [[ADD]], double* [[A]], align 4
4657 // CHECK20-NEXT:    ret void
4658 //
4659 //
4660 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
4661 // CHECK20-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR1]] {
4662 // CHECK20-NEXT:  entry:
4663 // CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
4664 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 4
4665 // CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4666 // CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
4667 // CHECK20-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
4668 // CHECK20-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
4669 // CHECK20-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4670 // CHECK20-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 4
4671 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
4672 // CHECK20-NEXT:    ret void
4673 //
4674 //
4675 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..1
4676 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4677 // CHECK20-NEXT:  entry:
4678 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4679 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4680 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4
4681 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4682 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4683 // CHECK20-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4
4684 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4
4685 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
4686 // CHECK20-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
4687 // CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
4688 // CHECK20-NEXT:    store double 2.500000e+00, double* [[A]], align 4
4689 // CHECK20-NEXT:    ret void
4690 //
4691 //
4692 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
4693 // CHECK20-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
4694 // CHECK20-NEXT:  entry:
4695 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4696 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 1
4697 // CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4698 // CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4699 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4700 // CHECK20-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
4701 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
4702 // CHECK20-NEXT:    ret void
4703 //
4704 //
4705 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..4
4706 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4707 // CHECK20-NEXT:  entry:
4708 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4709 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4710 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
4711 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4712 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4713 // CHECK20-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
4714 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
4715 // CHECK20-NEXT:    ret void
4716 //
4717 //
4718 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
4719 // CHECK20-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
4720 // CHECK20-NEXT:  entry:
4721 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4722 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 1
4723 // CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4724 // CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4725 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4726 // CHECK20-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
4727 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
4728 // CHECK20-NEXT:    ret void
4729 //
4730 //
4731 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..7
4732 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4733 // CHECK20-NEXT:  entry:
4734 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4735 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4736 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
4737 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4738 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4739 // CHECK20-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
4740 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
4741 // CHECK20-NEXT:    ret void
4742 //
4743 //
4744 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
4745 // CHECK20-SAME: () #[[ATTR1]] {
4746 // CHECK20-NEXT:  entry:
4747 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
4748 // CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4749 // CHECK20-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
4750 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
4751 // CHECK20-NEXT:    ret void
4752 //
4753 //
4754 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..10
4755 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4756 // CHECK20-NEXT:  entry:
4757 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4758 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4759 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
4760 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4761 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4762 // CHECK20-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
4763 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
4764 // CHECK20-NEXT:    ret void
4765 //
4766 //
4767 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
4768 // CHECK20-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
4769 // CHECK20-NEXT:  entry:
4770 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4771 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
4772 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4773 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
4774 // CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4775 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4776 // CHECK20-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
4777 // CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4778 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
4779 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
4780 // CHECK20-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
4781 // CHECK20-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
4782 // CHECK20-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
4783 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4784 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_ADDR]], align 4
4785 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
4786 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4787 // CHECK20-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV]], align 4
4788 // CHECK20-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
4789 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
4790 // CHECK20-NEXT:    ret void
4791 //
4792 //
4793 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..11
4794 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4795 // CHECK20-NEXT:  entry:
4796 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4797 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4798 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
4799 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
4800 // CHECK20-NEXT:    [[B:%.*]] = alloca i16, align 2
4801 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4802 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4803 // CHECK20-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
4804 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
4805 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
4806 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4807 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
4808 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
4809 // CHECK20-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
4810 // CHECK20-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
4811 // CHECK20-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
4812 // CHECK20-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
4813 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
4814 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
4815 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
4816 // CHECK20-NEXT:    ret void
4817 //
4818 //
4819 // CHECK20-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4820 // CHECK20-SAME: () #[[ATTR3:[0-9]+]] {
4821 // CHECK20-NEXT:  entry:
4822 // CHECK20-NEXT:    call void @__tgt_register_requires(i64 1)
4823 // CHECK20-NEXT:    ret void
4824 //
4825 //
4826 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
4827 // CHECK25-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
4828 // CHECK25-NEXT:  entry:
4829 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
4830 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
4831 // CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
4832 // CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
4833 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
4834 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
4835 // CHECK25-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
4836 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
4837 // CHECK25-NEXT:    ret void
4838 //
4839 //
4840 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined.
4841 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4842 // CHECK25-NEXT:  entry:
4843 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4844 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4845 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
4846 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4847 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4848 // CHECK25-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
4849 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
4850 // CHECK25-NEXT:    ret void
4851 //
4852 //
4853 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
4854 // CHECK25-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
4855 // CHECK25-NEXT:  entry:
4856 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
4857 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
4858 // CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4859 // CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
4860 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
4861 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
4862 // CHECK25-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
4863 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
4864 // CHECK25-NEXT:    ret void
4865 //
4866 //
4867 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..1
4868 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4869 // CHECK25-NEXT:  entry:
4870 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4871 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4872 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
4873 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4874 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4875 // CHECK25-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
4876 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
4877 // CHECK25-NEXT:    ret void
4878 //
4879 //
4880 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
4881 // CHECK25-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
4882 // CHECK25-NEXT:  entry:
4883 // CHECK25-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
4884 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
4885 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
4886 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 8
4887 // CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4888 // CHECK25-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
4889 // CHECK25-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
4890 // CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
4891 // CHECK25-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
4892 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
4893 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
4894 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
4895 // CHECK25-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
4896 // CHECK25-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4897 // CHECK25-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 8
4898 // CHECK25-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4899 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 8
4900 // CHECK25-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 8
4901 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
4902 // CHECK25-NEXT:    ret void
4903 //
4904 //
4905 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..2
4906 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4907 // CHECK25-NEXT:  entry:
4908 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4909 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4910 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
4911 // CHECK25-NEXT:    [[B:%.*]] = alloca i32, align 4
4912 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4913 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4914 // CHECK25-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
4915 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
4916 // CHECK25-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
4917 // CHECK25-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
4918 // CHECK25-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
4919 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
4920 // CHECK25-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
4921 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
4922 // CHECK25-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
4923 // CHECK25-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
4924 // CHECK25-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
4925 // CHECK25-NEXT:    store double [[ADD]], double* [[A]], align 8
4926 // CHECK25-NEXT:    ret void
4927 //
4928 //
4929 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
4930 // CHECK25-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
4931 // CHECK25-NEXT:  entry:
4932 // CHECK25-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
4933 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
4934 // CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4935 // CHECK25-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
4936 // CHECK25-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
4937 // CHECK25-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
4938 // CHECK25-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4939 // CHECK25-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 8
4940 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
4941 // CHECK25-NEXT:    ret void
4942 //
4943 //
4944 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..3
4945 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4946 // CHECK25-NEXT:  entry:
4947 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4948 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4949 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
4950 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4951 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4952 // CHECK25-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
4953 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
4954 // CHECK25-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
4955 // CHECK25-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
4956 // CHECK25-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
4957 // CHECK25-NEXT:    store double 2.500000e+00, double* [[A]], align 8
4958 // CHECK25-NEXT:    ret void
4959 //
4960 //
4961 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
4962 // CHECK25-SAME: () #[[ATTR0]] {
4963 // CHECK25-NEXT:  entry:
4964 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
4965 // CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4966 // CHECK25-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
4967 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
4968 // CHECK25-NEXT:    ret void
4969 //
4970 //
4971 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..4
4972 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4973 // CHECK25-NEXT:  entry:
4974 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4975 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4976 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
4977 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4978 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4979 // CHECK25-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
4980 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
4981 // CHECK25-NEXT:    ret void
4982 //
4983 //
4984 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
4985 // CHECK25-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
4986 // CHECK25-NEXT:  entry:
4987 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4988 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
4989 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
4990 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
4991 // CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4992 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4993 // CHECK25-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
4994 // CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
4995 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4996 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
4997 // CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
4998 // CHECK25-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
4999 // CHECK25-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
5000 // CHECK25-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
5001 // CHECK25-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5002 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
5003 // CHECK25-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
5004 // CHECK25-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5005 // CHECK25-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV1]], align 8
5006 // CHECK25-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
5007 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
5008 // CHECK25-NEXT:    ret void
5009 //
5010 //
5011 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..5
5012 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5013 // CHECK25-NEXT:  entry:
5014 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5015 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5016 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
5017 // CHECK25-NEXT:    [[A:%.*]] = alloca i32, align 4
5018 // CHECK25-NEXT:    [[B:%.*]] = alloca i16, align 2
5019 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5020 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5021 // CHECK25-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
5022 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
5023 // CHECK25-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
5024 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5025 // CHECK25-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5026 // CHECK25-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
5027 // CHECK25-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
5028 // CHECK25-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
5029 // CHECK25-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
5030 // CHECK25-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
5031 // CHECK25-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
5032 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
5033 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5034 // CHECK25-NEXT:    ret void
5035 //
5036 //
5037 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
5038 // CHECK26-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
5039 // CHECK26-NEXT:  entry:
5040 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
5041 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
5042 // CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
5043 // CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
5044 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
5045 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
5046 // CHECK26-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
5047 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
5048 // CHECK26-NEXT:    ret void
5049 //
5050 //
5051 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined.
5052 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5053 // CHECK26-NEXT:  entry:
5054 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5055 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5056 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
5057 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5058 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5059 // CHECK26-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
5060 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
5061 // CHECK26-NEXT:    ret void
5062 //
5063 //
5064 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
5065 // CHECK26-SAME: (i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
5066 // CHECK26-NEXT:  entry:
5067 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
5068 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
5069 // CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5070 // CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
5071 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
5072 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
5073 // CHECK26-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
5074 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
5075 // CHECK26-NEXT:    ret void
5076 //
5077 //
5078 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..1
5079 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5080 // CHECK26-NEXT:  entry:
5081 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5082 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5083 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
5084 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5085 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5086 // CHECK26-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
5087 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
5088 // CHECK26-NEXT:    ret void
5089 //
5090 //
5091 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
5092 // CHECK26-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
5093 // CHECK26-NEXT:  entry:
5094 // CHECK26-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
5095 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
5096 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
5097 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 8
5098 // CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5099 // CHECK26-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
5100 // CHECK26-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
5101 // CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
5102 // CHECK26-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
5103 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
5104 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
5105 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 8
5106 // CHECK26-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
5107 // CHECK26-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5108 // CHECK26-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 8
5109 // CHECK26-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5110 // CHECK26-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 8
5111 // CHECK26-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 8
5112 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
5113 // CHECK26-NEXT:    ret void
5114 //
5115 //
5116 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..2
5117 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5118 // CHECK26-NEXT:  entry:
5119 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5120 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5121 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
5122 // CHECK26-NEXT:    [[B:%.*]] = alloca i32, align 4
5123 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5124 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5125 // CHECK26-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
5126 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
5127 // CHECK26-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
5128 // CHECK26-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
5129 // CHECK26-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
5130 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
5131 // CHECK26-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
5132 // CHECK26-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
5133 // CHECK26-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
5134 // CHECK26-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
5135 // CHECK26-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
5136 // CHECK26-NEXT:    store double [[ADD]], double* [[A]], align 8
5137 // CHECK26-NEXT:    ret void
5138 //
5139 //
5140 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
5141 // CHECK26-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
5142 // CHECK26-NEXT:  entry:
5143 // CHECK26-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
5144 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
5145 // CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5146 // CHECK26-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
5147 // CHECK26-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
5148 // CHECK26-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
5149 // CHECK26-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5150 // CHECK26-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 8
5151 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
5152 // CHECK26-NEXT:    ret void
5153 //
5154 //
5155 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..3
5156 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5157 // CHECK26-NEXT:  entry:
5158 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5159 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5160 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
5161 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5162 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5163 // CHECK26-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
5164 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
5165 // CHECK26-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
5166 // CHECK26-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
5167 // CHECK26-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
5168 // CHECK26-NEXT:    store double 2.500000e+00, double* [[A]], align 8
5169 // CHECK26-NEXT:    ret void
5170 //
5171 //
5172 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
5173 // CHECK26-SAME: () #[[ATTR0]] {
5174 // CHECK26-NEXT:  entry:
5175 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
5176 // CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5177 // CHECK26-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
5178 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
5179 // CHECK26-NEXT:    ret void
5180 //
5181 //
5182 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..4
5183 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5184 // CHECK26-NEXT:  entry:
5185 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5186 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5187 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
5188 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5189 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5190 // CHECK26-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
5191 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
5192 // CHECK26-NEXT:    ret void
5193 //
5194 //
5195 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
5196 // CHECK26-SAME: (i64 [[A:%.*]], i64 [[B:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
5197 // CHECK26-NEXT:  entry:
5198 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
5199 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
5200 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
5201 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
5202 // CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5203 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
5204 // CHECK26-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
5205 // CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
5206 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
5207 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[B_ADDR]] to i16*
5208 // CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i16*
5209 // CHECK26-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV2]], align 8
5210 // CHECK26-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
5211 // CHECK26-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
5212 // CHECK26-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5213 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 8
5214 // CHECK26-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
5215 // CHECK26-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5216 // CHECK26-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV1]], align 8
5217 // CHECK26-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
5218 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
5219 // CHECK26-NEXT:    ret void
5220 //
5221 //
5222 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..5
5223 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5224 // CHECK26-NEXT:  entry:
5225 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5226 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5227 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
5228 // CHECK26-NEXT:    [[A:%.*]] = alloca i32, align 4
5229 // CHECK26-NEXT:    [[B:%.*]] = alloca i16, align 2
5230 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5231 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5232 // CHECK26-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
5233 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
5234 // CHECK26-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
5235 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5236 // CHECK26-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5237 // CHECK26-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
5238 // CHECK26-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
5239 // CHECK26-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
5240 // CHECK26-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
5241 // CHECK26-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
5242 // CHECK26-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
5243 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
5244 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5245 // CHECK26-NEXT:    ret void
5246 //
5247 //
5248 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
5249 // CHECK27-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
5250 // CHECK27-NEXT:  entry:
5251 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
5252 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
5253 // CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
5254 // CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5255 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5256 // CHECK27-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
5257 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
5258 // CHECK27-NEXT:    ret void
5259 //
5260 //
5261 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined.
5262 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5263 // CHECK27-NEXT:  entry:
5264 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5265 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5266 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
5267 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5268 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5269 // CHECK27-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
5270 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
5271 // CHECK27-NEXT:    ret void
5272 //
5273 //
5274 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
5275 // CHECK27-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
5276 // CHECK27-NEXT:  entry:
5277 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
5278 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
5279 // CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5280 // CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5281 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5282 // CHECK27-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
5283 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
5284 // CHECK27-NEXT:    ret void
5285 //
5286 //
5287 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..1
5288 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5289 // CHECK27-NEXT:  entry:
5290 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5291 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5292 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4
5293 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5294 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5295 // CHECK27-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4
5296 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4
5297 // CHECK27-NEXT:    ret void
5298 //
5299 //
5300 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
5301 // CHECK27-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
5302 // CHECK27-NEXT:  entry:
5303 // CHECK27-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
5304 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
5305 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
5306 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
5307 // CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5308 // CHECK27-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
5309 // CHECK27-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
5310 // CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5311 // CHECK27-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
5312 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5313 // CHECK27-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
5314 // CHECK27-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5315 // CHECK27-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 4
5316 // CHECK27-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5317 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_ADDR]], align 4
5318 // CHECK27-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 4
5319 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
5320 // CHECK27-NEXT:    ret void
5321 //
5322 //
5323 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..2
5324 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5325 // CHECK27-NEXT:  entry:
5326 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5327 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5328 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
5329 // CHECK27-NEXT:    [[B:%.*]] = alloca i32, align 4
5330 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5331 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5332 // CHECK27-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
5333 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
5334 // CHECK27-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
5335 // CHECK27-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
5336 // CHECK27-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
5337 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
5338 // CHECK27-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
5339 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
5340 // CHECK27-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
5341 // CHECK27-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
5342 // CHECK27-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
5343 // CHECK27-NEXT:    store double [[ADD]], double* [[A]], align 4
5344 // CHECK27-NEXT:    ret void
5345 //
5346 //
5347 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
5348 // CHECK27-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
5349 // CHECK27-NEXT:  entry:
5350 // CHECK27-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
5351 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 4
5352 // CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5353 // CHECK27-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
5354 // CHECK27-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
5355 // CHECK27-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
5356 // CHECK27-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5357 // CHECK27-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 4
5358 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
5359 // CHECK27-NEXT:    ret void
5360 //
5361 //
5362 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..3
5363 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5364 // CHECK27-NEXT:  entry:
5365 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5366 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5367 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
5368 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5369 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5370 // CHECK27-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
5371 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
5372 // CHECK27-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
5373 // CHECK27-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
5374 // CHECK27-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
5375 // CHECK27-NEXT:    store double 2.500000e+00, double* [[A]], align 4
5376 // CHECK27-NEXT:    ret void
5377 //
5378 //
5379 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
5380 // CHECK27-SAME: () #[[ATTR0]] {
5381 // CHECK27-NEXT:  entry:
5382 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
5383 // CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5384 // CHECK27-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
5385 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
5386 // CHECK27-NEXT:    ret void
5387 //
5388 //
5389 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..4
5390 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5391 // CHECK27-NEXT:  entry:
5392 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5393 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5394 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
5395 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5396 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5397 // CHECK27-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
5398 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
5399 // CHECK27-NEXT:    ret void
5400 //
5401 //
5402 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
5403 // CHECK27-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
5404 // CHECK27-NEXT:  entry:
5405 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5406 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
5407 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
5408 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
5409 // CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5410 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5411 // CHECK27-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
5412 // CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5413 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
5414 // CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
5415 // CHECK27-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
5416 // CHECK27-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
5417 // CHECK27-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
5418 // CHECK27-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5419 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_ADDR]], align 4
5420 // CHECK27-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
5421 // CHECK27-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5422 // CHECK27-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV]], align 4
5423 // CHECK27-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
5424 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
5425 // CHECK27-NEXT:    ret void
5426 //
5427 //
5428 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..5
5429 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5430 // CHECK27-NEXT:  entry:
5431 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5432 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5433 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
5434 // CHECK27-NEXT:    [[A:%.*]] = alloca i32, align 4
5435 // CHECK27-NEXT:    [[B:%.*]] = alloca i16, align 2
5436 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5437 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5438 // CHECK27-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
5439 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
5440 // CHECK27-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
5441 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5442 // CHECK27-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5443 // CHECK27-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
5444 // CHECK27-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
5445 // CHECK27-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
5446 // CHECK27-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
5447 // CHECK27-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
5448 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
5449 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
5450 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5451 // CHECK27-NEXT:    ret void
5452 //
5453 //
5454 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
5455 // CHECK28-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0:[0-9]+]] {
5456 // CHECK28-NEXT:  entry:
5457 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
5458 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
5459 // CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
5460 // CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5461 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5462 // CHECK28-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
5463 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
5464 // CHECK28-NEXT:    ret void
5465 //
5466 //
5467 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined.
5468 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5469 // CHECK28-NEXT:  entry:
5470 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5471 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5472 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
5473 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5474 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5475 // CHECK28-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
5476 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
5477 // CHECK28-NEXT:    ret void
5478 //
5479 //
5480 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
5481 // CHECK28-SAME: (i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
5482 // CHECK28-NEXT:  entry:
5483 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
5484 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
5485 // CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5486 // CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5487 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5488 // CHECK28-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]])
5489 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
5490 // CHECK28-NEXT:    ret void
5491 //
5492 //
5493 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..1
5494 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5495 // CHECK28-NEXT:  entry:
5496 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5497 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5498 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4
5499 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5500 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5501 // CHECK28-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4
5502 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4
5503 // CHECK28-NEXT:    ret void
5504 //
5505 //
5506 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
5507 // CHECK28-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
5508 // CHECK28-NEXT:  entry:
5509 // CHECK28-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
5510 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
5511 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
5512 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
5513 // CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5514 // CHECK28-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
5515 // CHECK28-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
5516 // CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5517 // CHECK28-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
5518 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5519 // CHECK28-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
5520 // CHECK28-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5521 // CHECK28-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP3]], align 4
5522 // CHECK28-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5523 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_ADDR]], align 4
5524 // CHECK28-NEXT:    store i32 [[TMP5]], i32* [[TMP4]], align 4
5525 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
5526 // CHECK28-NEXT:    ret void
5527 //
5528 //
5529 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..2
5530 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5531 // CHECK28-NEXT:  entry:
5532 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5533 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5534 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
5535 // CHECK28-NEXT:    [[B:%.*]] = alloca i32, align 4
5536 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5537 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5538 // CHECK28-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
5539 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
5540 // CHECK28-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
5541 // CHECK28-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
5542 // CHECK28-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
5543 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
5544 // CHECK28-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
5545 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
5546 // CHECK28-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP5]] to double
5547 // CHECK28-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
5548 // CHECK28-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
5549 // CHECK28-NEXT:    store double [[ADD]], double* [[A]], align 4
5550 // CHECK28-NEXT:    ret void
5551 //
5552 //
5553 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
5554 // CHECK28-SAME: (%struct.S1* [[THIS:%.*]]) #[[ATTR0]] {
5555 // CHECK28-NEXT:  entry:
5556 // CHECK28-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
5557 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 4
5558 // CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5559 // CHECK28-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
5560 // CHECK28-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
5561 // CHECK28-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1024)
5562 // CHECK28-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5563 // CHECK28-NEXT:    store %struct.S1* [[TMP1]], %struct.S1** [[TMP2]], align 4
5564 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
5565 // CHECK28-NEXT:    ret void
5566 //
5567 //
5568 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..3
5569 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5570 // CHECK28-NEXT:  entry:
5571 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5572 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5573 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
5574 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5575 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5576 // CHECK28-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
5577 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
5578 // CHECK28-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
5579 // CHECK28-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
5580 // CHECK28-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
5581 // CHECK28-NEXT:    store double 2.500000e+00, double* [[A]], align 4
5582 // CHECK28-NEXT:    ret void
5583 //
5584 //
5585 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
5586 // CHECK28-SAME: () #[[ATTR0]] {
5587 // CHECK28-NEXT:  entry:
5588 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 1
5589 // CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5590 // CHECK28-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 20)
5591 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
5592 // CHECK28-NEXT:    ret void
5593 //
5594 //
5595 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..4
5596 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5597 // CHECK28-NEXT:  entry:
5598 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5599 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5600 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
5601 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5602 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5603 // CHECK28-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
5604 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
5605 // CHECK28-NEXT:    ret void
5606 //
5607 //
5608 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
5609 // CHECK28-SAME: (i32 [[A:%.*]], i32 [[B:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
5610 // CHECK28-NEXT:  entry:
5611 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5612 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
5613 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
5614 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
5615 // CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
5616 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5617 // CHECK28-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
5618 // CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5619 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[B_ADDR]] to i16*
5620 // CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i16*
5621 // CHECK28-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 4
5622 // CHECK28-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
5623 // CHECK28-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]])
5624 // CHECK28-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5625 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_ADDR]], align 4
5626 // CHECK28-NEXT:    store i32 [[TMP4]], i32* [[TMP3]], align 4
5627 // CHECK28-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5628 // CHECK28-NEXT:    [[TMP6:%.*]] = load i16, i16* [[CONV]], align 4
5629 // CHECK28-NEXT:    store i16 [[TMP6]], i16* [[TMP5]], align 4
5630 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
5631 // CHECK28-NEXT:    ret void
5632 //
5633 //
5634 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..5
5635 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5636 // CHECK28-NEXT:  entry:
5637 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5638 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5639 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
5640 // CHECK28-NEXT:    [[A:%.*]] = alloca i32, align 4
5641 // CHECK28-NEXT:    [[B:%.*]] = alloca i16, align 2
5642 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5643 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5644 // CHECK28-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
5645 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
5646 // CHECK28-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
5647 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5648 // CHECK28-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5649 // CHECK28-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
5650 // CHECK28-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
5651 // CHECK28-NEXT:    store i16 [[TMP4]], i16* [[B]], align 2
5652 // CHECK28-NEXT:    [[TMP5:%.*]] = load i16, i16* [[B]], align 2
5653 // CHECK28-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
5654 // CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
5655 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[CONV]]
5656 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5657 // CHECK28-NEXT:    ret void
5658 //
5659