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 8 target regions, but only 6 that actually will generate offloading
78 // code and have mapped arguments, and only 4 have all-constant map sizes.
79 
80 
81 
82 // Check target registration is registered as a Ctor.
83 
84 
85 template<typename tx, typename ty>
86 struct TT{
87   tx X;
88   ty Y;
89 };
90 
91 int foo(int n) {
92   int a = 0;
93   short aa = 0;
94   float b[10];
95   float bn[n];
96   double c[5][10];
97   double cn[5][n];
98   TT<long long, char> d;
99 
100   #pragma omp target parallel nowait
101   {
102   }
103 
104   #pragma omp target parallel if(target: 0)
105   {
106     a += 1;
107   }
108 
109 
110   #pragma omp target parallel if(target: 1)
111   {
112     aa += 1;
113 #pragma omp cancel parallel
114   }
115 
116 
117 
118 
119   #pragma omp target parallel if(target: n>10)
120   {
121     a += 1;
122     aa += 1;
123   }
124 
125   // We capture 3 VLA sizes in this target region
126 
127 
128 
129 
130 
131   // The names below are not necessarily consistent with the names used for the
132   // addresses above as some are repeated.
133 
134 
135 
136 
137 
138 
139 
140 
141 
142 
143 
144   #pragma omp target parallel if(target: n>20)
145   {
146     a += 1;
147     b[2] += 1.0;
148     bn[3] += 1.0;
149     c[1][2] += 1.0;
150     cn[1][3] += 1.0;
151     d.X += 1;
152     d.Y += 1;
153   }
154 
155   return a;
156 }
157 
158 // Check that the offloading functions are emitted and that the arguments are
159 // correct and loaded correctly for the target regions in foo().
160 
161 
162 
163 // Create stack storage and store argument in there.
164 
165 // Create stack storage and store argument in there.
166 
167 // Create stack storage and store argument in there.
168 
169 // Create local storage for each capture.
170 
171 
172 
173 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
174 
175 template<typename tx>
176 tx ftemplate(int n) {
177   tx a = 0;
178   short aa = 0;
179   tx b[10];
180 
181   #pragma omp target parallel if(target: n>40)
182   {
183     a += 1;
184     aa += 1;
185     b[2] += 1;
186   }
187 
188   return a;
189 }
190 
191 static
192 int fstatic(int n) {
193   int a = 0;
194   short aa = 0;
195   char aaa = 0;
196   int b[10];
197 
198   #pragma omp target parallel if(target: n>50)
199   {
200     a += 1;
201     aa += 1;
202     aaa += 1;
203     b[2] += 1;
204   }
205 
206   return a;
207 }
208 
209 struct S1 {
210   double a;
211 
212   int r1(int n){
213     int b = n+1;
214     short int c[2][n];
215 
216     #pragma omp target parallel if(target: n>60)
217     {
218       this->a = (double)b + 1.5;
219       c[1][1] = ++a;
220     }
221 
222     return c[1][1] + (int)b;
223   }
224 };
225 
226 int bar(int n){
227   int a = 0;
228 
229   a += foo(n);
230 
231   S1 S;
232   a += S.r1(n);
233 
234   a += fstatic(n);
235 
236   a += ftemplate<int>(n);
237 
238   return a;
239 }
240 
241 
242 
243 // We capture 2 VLA sizes in this target region
244 
245 
246 // The names below are not necessarily consistent with the names used for the
247 // addresses above as some are repeated.
248 
249 
250 
251 
252 
253 
254 
255 
256 
257 
258 
259 
260 
261 
262 
263 
264 
265 
266 
267 // Check that the offloading functions are emitted and that the arguments are
268 // correct and loaded correctly for the target regions of the callees of bar().
269 
270 // Create local storage for each capture.
271 // Store captures in the context.
272 
273 
274 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
275 
276 
277 // Create local storage for each capture.
278 // Store captures in the context.
279 
280 
281 
282 
283 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
284 
285 // Create local storage for each capture.
286 // Store captures in the context.
287 
288 
289 
290 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
291 
292 
293 #endif
294 // CHECK1-LABEL: define {{[^@]+}}@_Z3fooi
295 // CHECK1-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
296 // CHECK1-NEXT:  entry:
297 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
298 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
299 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
300 // CHECK1-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
301 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
302 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
303 // CHECK1-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
304 // CHECK1-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
305 // CHECK1-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
306 // CHECK1-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
307 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
308 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
309 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
310 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
311 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
312 // CHECK1-NEXT:    [[A_CASTED3:%.*]] = alloca i64, align 8
313 // CHECK1-NEXT:    [[AA_CASTED5:%.*]] = alloca i64, align 8
314 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
315 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
316 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
317 // CHECK1-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
318 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [9 x i8*], align 8
319 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [9 x i8*], align 8
320 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [9 x i8*], align 8
321 // CHECK1-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 8
322 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
323 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
324 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
325 // CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
326 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
327 // CHECK1-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
328 // CHECK1-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
329 // CHECK1-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
330 // CHECK1-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
331 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
332 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
333 // CHECK1-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
334 // CHECK1-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
335 // CHECK1-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
336 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
337 // CHECK1-NEXT:    [[TMP7:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
338 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.kmp_task_t_with_privates*
339 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP8]], i32 0, i32 0
340 // CHECK1-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP7]])
341 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A]], align 4
342 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
343 // CHECK1-NEXT:    store i32 [[TMP11]], i32* [[CONV]], align 4
344 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED]], align 8
345 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i64 [[TMP12]]) #[[ATTR3:[0-9]+]]
346 // CHECK1-NEXT:    [[TMP13:%.*]] = load i16, i16* [[AA]], align 2
347 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
348 // CHECK1-NEXT:    store i16 [[TMP13]], i16* [[CONV2]], align 2
349 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[AA_CASTED]], align 8
350 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
351 // CHECK1-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
352 // CHECK1-NEXT:    store i64 [[TMP14]], i64* [[TMP16]], align 8
353 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
354 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
355 // CHECK1-NEXT:    store i64 [[TMP14]], i64* [[TMP18]], align 8
356 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
357 // CHECK1-NEXT:    store i8* null, i8** [[TMP19]], align 8
358 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
359 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
360 // CHECK1-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
361 // CHECK1-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
362 // CHECK1-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
363 // CHECK1:       omp_offload.failed:
364 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i64 [[TMP14]]) #[[ATTR3]]
365 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
366 // CHECK1:       omp_offload.cont:
367 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
368 // CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[A_CASTED3]] to i32*
369 // CHECK1-NEXT:    store i32 [[TMP24]], i32* [[CONV4]], align 4
370 // CHECK1-NEXT:    [[TMP25:%.*]] = load i64, i64* [[A_CASTED3]], align 8
371 // CHECK1-NEXT:    [[TMP26:%.*]] = load i16, i16* [[AA]], align 2
372 // CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED5]] to i16*
373 // CHECK1-NEXT:    store i16 [[TMP26]], i16* [[CONV6]], align 2
374 // CHECK1-NEXT:    [[TMP27:%.*]] = load i64, i64* [[AA_CASTED5]], align 8
375 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[N_ADDR]], align 4
376 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP28]], 10
377 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
378 // CHECK1:       omp_if.then:
379 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
380 // CHECK1-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i64*
381 // CHECK1-NEXT:    store i64 [[TMP25]], i64* [[TMP30]], align 8
382 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
383 // CHECK1-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
384 // CHECK1-NEXT:    store i64 [[TMP25]], i64* [[TMP32]], align 8
385 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
386 // CHECK1-NEXT:    store i8* null, i8** [[TMP33]], align 8
387 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
388 // CHECK1-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i64*
389 // CHECK1-NEXT:    store i64 [[TMP27]], i64* [[TMP35]], align 8
390 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
391 // CHECK1-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
392 // CHECK1-NEXT:    store i64 [[TMP27]], i64* [[TMP37]], align 8
393 // CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
394 // CHECK1-NEXT:    store i8* null, i8** [[TMP38]], align 8
395 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
396 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
397 // CHECK1-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
398 // CHECK1-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
399 // CHECK1-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED10:%.*]], label [[OMP_OFFLOAD_CONT11:%.*]]
400 // CHECK1:       omp_offload.failed10:
401 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR3]]
402 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT11]]
403 // CHECK1:       omp_offload.cont11:
404 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
405 // CHECK1:       omp_if.else:
406 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR3]]
407 // CHECK1-NEXT:    br label [[OMP_IF_END]]
408 // CHECK1:       omp_if.end:
409 // CHECK1-NEXT:    [[TMP43:%.*]] = load i32, i32* [[A]], align 4
410 // CHECK1-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
411 // CHECK1-NEXT:    store i32 [[TMP43]], i32* [[CONV13]], align 4
412 // CHECK1-NEXT:    [[TMP44:%.*]] = load i64, i64* [[A_CASTED12]], align 8
413 // CHECK1-NEXT:    [[TMP45:%.*]] = load i32, i32* [[N_ADDR]], align 4
414 // CHECK1-NEXT:    [[CMP14:%.*]] = icmp sgt i32 [[TMP45]], 20
415 // CHECK1-NEXT:    br i1 [[CMP14]], label [[OMP_IF_THEN15:%.*]], label [[OMP_IF_ELSE21:%.*]]
416 // CHECK1:       omp_if.then15:
417 // CHECK1-NEXT:    [[TMP46:%.*]] = mul nuw i64 [[TMP2]], 4
418 // CHECK1-NEXT:    [[TMP47:%.*]] = mul nuw i64 5, [[TMP5]]
419 // CHECK1-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], 8
420 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
421 // CHECK1-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i64*
422 // CHECK1-NEXT:    store i64 [[TMP44]], i64* [[TMP50]], align 8
423 // CHECK1-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
424 // CHECK1-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i64*
425 // CHECK1-NEXT:    store i64 [[TMP44]], i64* [[TMP52]], align 8
426 // CHECK1-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
427 // CHECK1-NEXT:    store i64 4, i64* [[TMP53]], align 8
428 // CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
429 // CHECK1-NEXT:    store i8* null, i8** [[TMP54]], align 8
430 // CHECK1-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
431 // CHECK1-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
432 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 8
433 // CHECK1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
434 // CHECK1-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
435 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 8
436 // CHECK1-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
437 // CHECK1-NEXT:    store i64 40, i64* [[TMP59]], align 8
438 // CHECK1-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
439 // CHECK1-NEXT:    store i8* null, i8** [[TMP60]], align 8
440 // CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 2
441 // CHECK1-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
442 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP62]], align 8
443 // CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 2
444 // CHECK1-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
445 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP64]], align 8
446 // CHECK1-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
447 // CHECK1-NEXT:    store i64 8, i64* [[TMP65]], align 8
448 // CHECK1-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 2
449 // CHECK1-NEXT:    store i8* null, i8** [[TMP66]], align 8
450 // CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 3
451 // CHECK1-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
452 // CHECK1-NEXT:    store float* [[VLA]], float** [[TMP68]], align 8
453 // CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 3
454 // CHECK1-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
455 // CHECK1-NEXT:    store float* [[VLA]], float** [[TMP70]], align 8
456 // CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
457 // CHECK1-NEXT:    store i64 [[TMP46]], i64* [[TMP71]], align 8
458 // CHECK1-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 3
459 // CHECK1-NEXT:    store i8* null, i8** [[TMP72]], align 8
460 // CHECK1-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 4
461 // CHECK1-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
462 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 8
463 // CHECK1-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 4
464 // CHECK1-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
465 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 8
466 // CHECK1-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
467 // CHECK1-NEXT:    store i64 400, i64* [[TMP77]], align 8
468 // CHECK1-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 4
469 // CHECK1-NEXT:    store i8* null, i8** [[TMP78]], align 8
470 // CHECK1-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 5
471 // CHECK1-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i64*
472 // CHECK1-NEXT:    store i64 5, i64* [[TMP80]], align 8
473 // CHECK1-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 5
474 // CHECK1-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
475 // CHECK1-NEXT:    store i64 5, i64* [[TMP82]], align 8
476 // CHECK1-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
477 // CHECK1-NEXT:    store i64 8, i64* [[TMP83]], align 8
478 // CHECK1-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 5
479 // CHECK1-NEXT:    store i8* null, i8** [[TMP84]], align 8
480 // CHECK1-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 6
481 // CHECK1-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
482 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP86]], align 8
483 // CHECK1-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 6
484 // CHECK1-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
485 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP88]], align 8
486 // CHECK1-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
487 // CHECK1-NEXT:    store i64 8, i64* [[TMP89]], align 8
488 // CHECK1-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 6
489 // CHECK1-NEXT:    store i8* null, i8** [[TMP90]], align 8
490 // CHECK1-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 7
491 // CHECK1-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
492 // CHECK1-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 8
493 // CHECK1-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 7
494 // CHECK1-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
495 // CHECK1-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 8
496 // CHECK1-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
497 // CHECK1-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 8
498 // CHECK1-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 7
499 // CHECK1-NEXT:    store i8* null, i8** [[TMP96]], align 8
500 // CHECK1-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 8
501 // CHECK1-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
502 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 8
503 // CHECK1-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 8
504 // CHECK1-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
505 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 8
506 // CHECK1-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
507 // CHECK1-NEXT:    store i64 16, i64* [[TMP101]], align 8
508 // CHECK1-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 8
509 // CHECK1-NEXT:    store i8* null, i8** [[TMP102]], align 8
510 // CHECK1-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
511 // CHECK1-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
512 // CHECK1-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
513 // CHECK1-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
514 // CHECK1-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
515 // CHECK1-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
516 // CHECK1:       omp_offload.failed19:
517 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
518 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
519 // CHECK1:       omp_offload.cont20:
520 // CHECK1-NEXT:    br label [[OMP_IF_END22:%.*]]
521 // CHECK1:       omp_if.else21:
522 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
523 // CHECK1-NEXT:    br label [[OMP_IF_END22]]
524 // CHECK1:       omp_if.end22:
525 // CHECK1-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
526 // CHECK1-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
527 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
528 // CHECK1-NEXT:    ret i32 [[TMP108]]
529 //
530 //
531 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
532 // CHECK1-SAME: () #[[ATTR2:[0-9]+]] {
533 // CHECK1-NEXT:  entry:
534 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
535 // 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_]])
536 // CHECK1-NEXT:    ret void
537 //
538 //
539 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
540 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
541 // CHECK1-NEXT:  entry:
542 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
543 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
544 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
545 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
546 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
547 // CHECK1-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
548 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
549 // CHECK1-NEXT:    ret void
550 //
551 //
552 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
553 // CHECK1-SAME: (i32 signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
554 // CHECK1-NEXT:  entry:
555 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
556 // CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
557 // CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
558 // CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
559 // CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
560 // CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 8
561 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
562 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
563 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
564 // CHECK1-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
565 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
566 // CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
567 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
568 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
569 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
570 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
571 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
572 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
573 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
574 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
575 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
576 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
577 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !21
578 // CHECK1-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !21
579 // CHECK1-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
580 // CHECK1-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
581 // CHECK1-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !21
582 // CHECK1-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
583 // CHECK1-NEXT:    [[TMP10:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
584 // CHECK1-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0) #[[ATTR3]]
585 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
586 // CHECK1-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
587 // CHECK1:       omp_offload.failed.i:
588 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR3]]
589 // CHECK1-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
590 // CHECK1:       .omp_outlined..1.exit:
591 // CHECK1-NEXT:    ret i32 0
592 //
593 //
594 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
595 // CHECK1-SAME: (i64 [[A:%.*]]) #[[ATTR2]] {
596 // CHECK1-NEXT:  entry:
597 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
598 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
599 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
600 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
601 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
602 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
603 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
604 // 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..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
605 // CHECK1-NEXT:    ret void
606 //
607 //
608 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
609 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
610 // CHECK1-NEXT:  entry:
611 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
612 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
613 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
614 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
615 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
616 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
617 // CHECK1-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
618 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
619 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
620 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
621 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
622 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
623 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], 1
624 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
625 // CHECK1-NEXT:    ret void
626 //
627 //
628 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
629 // CHECK1-SAME: (i64 [[AA:%.*]]) #[[ATTR2]] {
630 // CHECK1-NEXT:  entry:
631 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
632 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 2
633 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
634 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
635 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
636 // CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
637 // CHECK1-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
638 // 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..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
639 // CHECK1-NEXT:    ret void
640 //
641 //
642 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
643 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
644 // CHECK1-NEXT:  entry:
645 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
646 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
647 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
648 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
649 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
650 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
651 // CHECK1-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
652 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
653 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
654 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
655 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
656 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
657 // CHECK1-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
658 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
659 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
660 // CHECK1-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
661 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
662 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
663 // CHECK1-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
664 // CHECK1-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
665 // CHECK1-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
666 // CHECK1:       .cancel.exit:
667 // CHECK1-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
668 // CHECK1-NEXT:    br label [[DOTCANCEL_CONTINUE]]
669 // CHECK1:       .cancel.continue:
670 // CHECK1-NEXT:    ret void
671 //
672 //
673 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
674 // CHECK1-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR2]] {
675 // CHECK1-NEXT:  entry:
676 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
677 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
678 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 4
679 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
680 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
681 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
682 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
683 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
684 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
685 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
686 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
687 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
688 // CHECK1-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
689 // 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..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
690 // CHECK1-NEXT:    ret void
691 //
692 //
693 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
694 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
695 // CHECK1-NEXT:  entry:
696 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
697 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
698 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
699 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
700 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
701 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
702 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
703 // CHECK1-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
704 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
705 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
706 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
707 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
708 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
709 // CHECK1-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
710 // CHECK1-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
711 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
712 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
713 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
714 // CHECK1-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
715 // CHECK1-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
716 // CHECK1-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
717 // CHECK1-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
718 // CHECK1-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
719 // CHECK1-NEXT:    ret void
720 //
721 //
722 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
723 // CHECK1-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR2]] {
724 // CHECK1-NEXT:  entry:
725 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
726 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
727 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
728 // CHECK1-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
729 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
730 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
731 // CHECK1-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
732 // CHECK1-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
733 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
734 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
735 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
736 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
737 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
738 // CHECK1-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
739 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
740 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
741 // CHECK1-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
742 // CHECK1-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
743 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
744 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
745 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
746 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
747 // CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
748 // CHECK1-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
749 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
750 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
751 // CHECK1-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
752 // CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
753 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
754 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
755 // CHECK1-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 8
756 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
757 // CHECK1-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 8
758 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
759 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP11]], align 8
760 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
761 // CHECK1-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 8
762 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
763 // CHECK1-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 8
764 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
765 // CHECK1-NEXT:    store i64 [[TMP4]], i64* [[TMP14]], align 8
766 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
767 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP15]], align 8
768 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
769 // CHECK1-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 8
770 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
771 // CHECK1-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 8
772 // 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..7 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
773 // CHECK1-NEXT:    ret void
774 //
775 //
776 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
777 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
778 // CHECK1-NEXT:  entry:
779 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
780 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
781 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
782 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
783 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
784 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
785 // CHECK1-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
786 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
787 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
788 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
789 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
790 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
791 // CHECK1-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 8
792 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
793 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
794 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
795 // CHECK1-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 8
796 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
797 // CHECK1-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 8
798 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 5
799 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[TMP11]], align 8
800 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 6
801 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[TMP13]], align 8
802 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 7
803 // CHECK1-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 8
804 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 8
805 // CHECK1-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 8
806 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
807 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
808 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
809 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i64 0, i64 2
810 // CHECK1-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
811 // CHECK1-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
812 // CHECK1-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
813 // CHECK1-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
814 // CHECK1-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
815 // CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 3
816 // CHECK1-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
817 // CHECK1-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
818 // CHECK1-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
819 // CHECK1-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
820 // CHECK1-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
821 // CHECK1-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i64 0, i64 1
822 // CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i64 0, i64 2
823 // CHECK1-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
824 // CHECK1-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
825 // CHECK1-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
826 // CHECK1-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP14]]
827 // CHECK1-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i64 [[TMP23]]
828 // CHECK1-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i64 3
829 // CHECK1-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
830 // CHECK1-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
831 // CHECK1-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
832 // CHECK1-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
833 // CHECK1-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8
834 // CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
835 // CHECK1-NEXT:    store i64 [[ADD13]], i64* [[X]], align 8
836 // CHECK1-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
837 // CHECK1-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8
838 // CHECK1-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
839 // CHECK1-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
840 // CHECK1-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
841 // CHECK1-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 8
842 // CHECK1-NEXT:    ret void
843 //
844 //
845 // CHECK1-LABEL: define {{[^@]+}}@_Z3bari
846 // CHECK1-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
847 // CHECK1-NEXT:  entry:
848 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
849 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
850 // CHECK1-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
851 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
852 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
853 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
854 // CHECK1-NEXT:    [[CALL:%.*]] = call signext i32 @_Z3fooi(i32 signext [[TMP0]])
855 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
856 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
857 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
858 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
859 // CHECK1-NEXT:    [[CALL1:%.*]] = call signext i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], i32 signext [[TMP2]])
860 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
861 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
862 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
863 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
864 // CHECK1-NEXT:    [[CALL3:%.*]] = call signext i32 @_ZL7fstatici(i32 signext [[TMP4]])
865 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
866 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
867 // CHECK1-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
868 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
869 // CHECK1-NEXT:    [[CALL5:%.*]] = call signext i32 @_Z9ftemplateIiET_i(i32 signext [[TMP6]])
870 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
871 // CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
872 // CHECK1-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
873 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
874 // CHECK1-NEXT:    ret i32 [[TMP8]]
875 //
876 //
877 // CHECK1-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
878 // CHECK1-SAME: (%struct.S1* nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
879 // CHECK1-NEXT:  entry:
880 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
881 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
882 // CHECK1-NEXT:    [[B:%.*]] = alloca i32, align 4
883 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
884 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
885 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
886 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
887 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
888 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
889 // CHECK1-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
890 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
891 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
892 // CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
893 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
894 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
895 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
896 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
897 // CHECK1-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
898 // CHECK1-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
899 // CHECK1-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
900 // CHECK1-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
901 // CHECK1-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
902 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
903 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
904 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
905 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
906 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
907 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
908 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
909 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
910 // CHECK1:       omp_if.then:
911 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
912 // CHECK1-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
913 // CHECK1-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
914 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
915 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
916 // CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 8
917 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
918 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
919 // CHECK1-NEXT:    store double* [[A]], double** [[TMP13]], align 8
920 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
921 // CHECK1-NEXT:    store i64 8, i64* [[TMP14]], align 8
922 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
923 // CHECK1-NEXT:    store i8* null, i8** [[TMP15]], align 8
924 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
925 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
926 // CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
927 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
928 // CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
929 // CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
930 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
931 // CHECK1-NEXT:    store i64 4, i64* [[TMP20]], align 8
932 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
933 // CHECK1-NEXT:    store i8* null, i8** [[TMP21]], align 8
934 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
935 // CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
936 // CHECK1-NEXT:    store i64 2, i64* [[TMP23]], align 8
937 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
938 // CHECK1-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
939 // CHECK1-NEXT:    store i64 2, i64* [[TMP25]], align 8
940 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
941 // CHECK1-NEXT:    store i64 8, i64* [[TMP26]], align 8
942 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
943 // CHECK1-NEXT:    store i8* null, i8** [[TMP27]], align 8
944 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
945 // CHECK1-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
946 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
947 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
948 // CHECK1-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
949 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP31]], align 8
950 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
951 // CHECK1-NEXT:    store i64 8, i64* [[TMP32]], align 8
952 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
953 // CHECK1-NEXT:    store i8* null, i8** [[TMP33]], align 8
954 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
955 // CHECK1-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
956 // CHECK1-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 8
957 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
958 // CHECK1-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
959 // CHECK1-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 8
960 // CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
961 // CHECK1-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 8
962 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
963 // CHECK1-NEXT:    store i8* null, i8** [[TMP39]], align 8
964 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
965 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
966 // CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
967 // CHECK1-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
968 // CHECK1-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
969 // CHECK1-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
970 // CHECK1:       omp_offload.failed:
971 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
972 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
973 // CHECK1:       omp_offload.cont:
974 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
975 // CHECK1:       omp_if.else:
976 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
977 // CHECK1-NEXT:    br label [[OMP_IF_END]]
978 // CHECK1:       omp_if.end:
979 // CHECK1-NEXT:    [[TMP45:%.*]] = mul nsw i64 1, [[TMP2]]
980 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP45]]
981 // CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
982 // CHECK1-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
983 // CHECK1-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP46]] to i32
984 // CHECK1-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
985 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP47]]
986 // CHECK1-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
987 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
988 // CHECK1-NEXT:    ret i32 [[ADD4]]
989 //
990 //
991 // CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici
992 // CHECK1-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
993 // CHECK1-NEXT:  entry:
994 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
995 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
996 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
997 // CHECK1-NEXT:    [[AAA:%.*]] = alloca i8, align 1
998 // CHECK1-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
999 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1000 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1001 // CHECK1-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
1002 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
1003 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
1004 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
1005 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1006 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
1007 // CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
1008 // CHECK1-NEXT:    store i8 0, i8* [[AAA]], align 1
1009 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
1010 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1011 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
1012 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
1013 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
1014 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1015 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
1016 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1017 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
1018 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
1019 // CHECK1-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
1020 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
1021 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
1022 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
1023 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1024 // CHECK1:       omp_if.then:
1025 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1026 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
1027 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
1028 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1029 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
1030 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
1031 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1032 // CHECK1-NEXT:    store i8* null, i8** [[TMP11]], align 8
1033 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1034 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1035 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
1036 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1037 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
1038 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
1039 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1040 // CHECK1-NEXT:    store i8* null, i8** [[TMP16]], align 8
1041 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1042 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
1043 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
1044 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1045 // CHECK1-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
1046 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
1047 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1048 // CHECK1-NEXT:    store i8* null, i8** [[TMP21]], align 8
1049 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1050 // CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
1051 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
1052 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1053 // CHECK1-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
1054 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
1055 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
1056 // CHECK1-NEXT:    store i8* null, i8** [[TMP26]], align 8
1057 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1058 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1059 // CHECK1-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
1060 // CHECK1-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
1061 // CHECK1-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1062 // CHECK1:       omp_offload.failed:
1063 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
1064 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1065 // CHECK1:       omp_offload.cont:
1066 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1067 // CHECK1:       omp_if.else:
1068 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
1069 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1070 // CHECK1:       omp_if.end:
1071 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
1072 // CHECK1-NEXT:    ret i32 [[TMP31]]
1073 //
1074 //
1075 // CHECK1-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
1076 // CHECK1-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] comdat {
1077 // CHECK1-NEXT:  entry:
1078 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1079 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
1080 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
1081 // CHECK1-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
1082 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1083 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1084 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1085 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1086 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1087 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1088 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
1089 // CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
1090 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
1091 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1092 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
1093 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
1094 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
1095 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1096 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
1097 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1098 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
1099 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
1100 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1101 // CHECK1:       omp_if.then:
1102 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1103 // CHECK1-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
1104 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
1105 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1106 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
1107 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
1108 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1109 // CHECK1-NEXT:    store i8* null, i8** [[TMP9]], align 8
1110 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1111 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
1112 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
1113 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1114 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1115 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
1116 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1117 // CHECK1-NEXT:    store i8* null, i8** [[TMP14]], align 8
1118 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1119 // CHECK1-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
1120 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
1121 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1122 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
1123 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
1124 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1125 // CHECK1-NEXT:    store i8* null, i8** [[TMP19]], align 8
1126 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1127 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1128 // CHECK1-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
1129 // CHECK1-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
1130 // CHECK1-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1131 // CHECK1:       omp_offload.failed:
1132 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
1133 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1134 // CHECK1:       omp_offload.cont:
1135 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1136 // CHECK1:       omp_if.else:
1137 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
1138 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1139 // CHECK1:       omp_if.end:
1140 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
1141 // CHECK1-NEXT:    ret i32 [[TMP24]]
1142 //
1143 //
1144 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
1145 // CHECK1-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
1146 // CHECK1-NEXT:  entry:
1147 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1148 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1149 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1150 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
1151 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
1152 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
1153 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1154 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1155 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1156 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
1157 // CHECK1-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
1158 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1159 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1160 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1161 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
1162 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
1163 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1164 // CHECK1-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 8
1165 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1166 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[CONV]], align 8
1167 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
1168 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
1169 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP7]], align 8
1170 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
1171 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP8]], align 8
1172 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
1173 // CHECK1-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 8
1174 // 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.5*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
1175 // CHECK1-NEXT:    ret void
1176 //
1177 //
1178 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..9
1179 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
1180 // CHECK1-NEXT:  entry:
1181 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1182 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1183 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
1184 // CHECK1-NEXT:    [[B:%.*]] = alloca i32, align 4
1185 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1186 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1187 // CHECK1-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
1188 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
1189 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
1190 // CHECK1-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
1191 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
1192 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
1193 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
1194 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
1195 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
1196 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
1197 // CHECK1-NEXT:    [[TMP8:%.*]] = load i64, i64* [[TMP7]], align 8
1198 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
1199 // CHECK1-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 8
1200 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
1201 // CHECK1-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
1202 // CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
1203 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
1204 // CHECK1-NEXT:    store double [[ADD]], double* [[A]], align 8
1205 // CHECK1-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
1206 // CHECK1-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 8
1207 // CHECK1-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
1208 // CHECK1-NEXT:    store double [[INC]], double* [[A1]], align 8
1209 // CHECK1-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
1210 // CHECK1-NEXT:    [[TMP13:%.*]] = mul nsw i64 1, [[TMP8]]
1211 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i64 [[TMP13]]
1212 // CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
1213 // CHECK1-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
1214 // CHECK1-NEXT:    ret void
1215 //
1216 //
1217 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
1218 // CHECK1-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
1219 // CHECK1-NEXT:  entry:
1220 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1221 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1222 // CHECK1-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
1223 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1224 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 8
1225 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1226 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1227 // CHECK1-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
1228 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1229 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1230 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1231 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
1232 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1233 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1234 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
1235 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
1236 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1237 // CHECK1-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
1238 // CHECK1-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
1239 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
1240 // CHECK1-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV2]], align 8
1241 // CHECK1-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
1242 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
1243 // CHECK1-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 8
1244 // 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.6*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
1245 // CHECK1-NEXT:    ret void
1246 //
1247 //
1248 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
1249 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
1250 // CHECK1-NEXT:  entry:
1251 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1252 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1253 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 8
1254 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
1255 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
1256 // CHECK1-NEXT:    [[AAA:%.*]] = alloca i8, align 1
1257 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1258 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1259 // CHECK1-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 8
1260 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 8
1261 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
1262 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
1263 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
1264 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
1265 // CHECK1-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
1266 // CHECK1-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
1267 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
1268 // CHECK1-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
1269 // CHECK1-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
1270 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
1271 // CHECK1-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 8
1272 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
1273 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
1274 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1275 // CHECK1-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
1276 // CHECK1-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
1277 // CHECK1-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
1278 // CHECK1-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
1279 // CHECK1-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
1280 // CHECK1-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
1281 // CHECK1-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
1282 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
1283 // CHECK1-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
1284 // CHECK1-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
1285 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i64 0, i64 2
1286 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1287 // CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
1288 // CHECK1-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
1289 // CHECK1-NEXT:    ret void
1290 //
1291 //
1292 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
1293 // CHECK1-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
1294 // CHECK1-NEXT:  entry:
1295 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1296 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1297 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1298 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 8
1299 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1300 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1301 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1302 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1303 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1304 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1305 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1306 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
1307 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
1308 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1309 // CHECK1-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
1310 // CHECK1-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
1311 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
1312 // CHECK1-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 8
1313 // 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.7*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
1314 // CHECK1-NEXT:    ret void
1315 //
1316 //
1317 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..14
1318 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
1319 // CHECK1-NEXT:  entry:
1320 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1321 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1322 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 8
1323 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
1324 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
1325 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1326 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1327 // CHECK1-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 8
1328 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 8
1329 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
1330 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
1331 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
1332 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
1333 // CHECK1-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
1334 // CHECK1-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
1335 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
1336 // CHECK1-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 8
1337 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
1338 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
1339 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1340 // CHECK1-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
1341 // CHECK1-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
1342 // CHECK1-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
1343 // CHECK1-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
1344 // CHECK1-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
1345 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i64 0, i64 2
1346 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1347 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
1348 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
1349 // CHECK1-NEXT:    ret void
1350 //
1351 //
1352 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1353 // CHECK1-SAME: () #[[ATTR6:[0-9]+]] {
1354 // CHECK1-NEXT:  entry:
1355 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
1356 // CHECK1-NEXT:    ret void
1357 //
1358 //
1359 // CHECK2-LABEL: define {{[^@]+}}@_Z3fooi
1360 // CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
1361 // CHECK2-NEXT:  entry:
1362 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1363 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
1364 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
1365 // CHECK2-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
1366 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
1367 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1368 // CHECK2-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
1369 // CHECK2-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
1370 // CHECK2-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
1371 // CHECK2-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
1372 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1373 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1374 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
1375 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
1376 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
1377 // CHECK2-NEXT:    [[A_CASTED3:%.*]] = alloca i64, align 8
1378 // CHECK2-NEXT:    [[AA_CASTED5:%.*]] = alloca i64, align 8
1379 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
1380 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
1381 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
1382 // CHECK2-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
1383 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [9 x i8*], align 8
1384 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [9 x i8*], align 8
1385 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [9 x i8*], align 8
1386 // CHECK2-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 8
1387 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
1388 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1389 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
1390 // CHECK2-NEXT:    store i16 0, i16* [[AA]], align 2
1391 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
1392 // CHECK2-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
1393 // CHECK2-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
1394 // CHECK2-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
1395 // CHECK2-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
1396 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
1397 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
1398 // CHECK2-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
1399 // CHECK2-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
1400 // CHECK2-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
1401 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
1402 // CHECK2-NEXT:    [[TMP7:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
1403 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.kmp_task_t_with_privates*
1404 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP8]], i32 0, i32 0
1405 // CHECK2-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP7]])
1406 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A]], align 4
1407 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1408 // CHECK2-NEXT:    store i32 [[TMP11]], i32* [[CONV]], align 4
1409 // CHECK2-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED]], align 8
1410 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i64 [[TMP12]]) #[[ATTR3:[0-9]+]]
1411 // CHECK2-NEXT:    [[TMP13:%.*]] = load i16, i16* [[AA]], align 2
1412 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1413 // CHECK2-NEXT:    store i16 [[TMP13]], i16* [[CONV2]], align 2
1414 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1415 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1416 // CHECK2-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
1417 // CHECK2-NEXT:    store i64 [[TMP14]], i64* [[TMP16]], align 8
1418 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1419 // CHECK2-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
1420 // CHECK2-NEXT:    store i64 [[TMP14]], i64* [[TMP18]], align 8
1421 // CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1422 // CHECK2-NEXT:    store i8* null, i8** [[TMP19]], align 8
1423 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1424 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1425 // CHECK2-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
1426 // CHECK2-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
1427 // CHECK2-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1428 // CHECK2:       omp_offload.failed:
1429 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i64 [[TMP14]]) #[[ATTR3]]
1430 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1431 // CHECK2:       omp_offload.cont:
1432 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
1433 // CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[A_CASTED3]] to i32*
1434 // CHECK2-NEXT:    store i32 [[TMP24]], i32* [[CONV4]], align 4
1435 // CHECK2-NEXT:    [[TMP25:%.*]] = load i64, i64* [[A_CASTED3]], align 8
1436 // CHECK2-NEXT:    [[TMP26:%.*]] = load i16, i16* [[AA]], align 2
1437 // CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED5]] to i16*
1438 // CHECK2-NEXT:    store i16 [[TMP26]], i16* [[CONV6]], align 2
1439 // CHECK2-NEXT:    [[TMP27:%.*]] = load i64, i64* [[AA_CASTED5]], align 8
1440 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[N_ADDR]], align 4
1441 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP28]], 10
1442 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1443 // CHECK2:       omp_if.then:
1444 // CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
1445 // CHECK2-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i64*
1446 // CHECK2-NEXT:    store i64 [[TMP25]], i64* [[TMP30]], align 8
1447 // CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
1448 // CHECK2-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
1449 // CHECK2-NEXT:    store i64 [[TMP25]], i64* [[TMP32]], align 8
1450 // CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
1451 // CHECK2-NEXT:    store i8* null, i8** [[TMP33]], align 8
1452 // CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
1453 // CHECK2-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i64*
1454 // CHECK2-NEXT:    store i64 [[TMP27]], i64* [[TMP35]], align 8
1455 // CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
1456 // CHECK2-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
1457 // CHECK2-NEXT:    store i64 [[TMP27]], i64* [[TMP37]], align 8
1458 // CHECK2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
1459 // CHECK2-NEXT:    store i8* null, i8** [[TMP38]], align 8
1460 // CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
1461 // CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
1462 // CHECK2-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
1463 // CHECK2-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
1464 // CHECK2-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED10:%.*]], label [[OMP_OFFLOAD_CONT11:%.*]]
1465 // CHECK2:       omp_offload.failed10:
1466 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR3]]
1467 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT11]]
1468 // CHECK2:       omp_offload.cont11:
1469 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
1470 // CHECK2:       omp_if.else:
1471 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR3]]
1472 // CHECK2-NEXT:    br label [[OMP_IF_END]]
1473 // CHECK2:       omp_if.end:
1474 // CHECK2-NEXT:    [[TMP43:%.*]] = load i32, i32* [[A]], align 4
1475 // CHECK2-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
1476 // CHECK2-NEXT:    store i32 [[TMP43]], i32* [[CONV13]], align 4
1477 // CHECK2-NEXT:    [[TMP44:%.*]] = load i64, i64* [[A_CASTED12]], align 8
1478 // CHECK2-NEXT:    [[TMP45:%.*]] = load i32, i32* [[N_ADDR]], align 4
1479 // CHECK2-NEXT:    [[CMP14:%.*]] = icmp sgt i32 [[TMP45]], 20
1480 // CHECK2-NEXT:    br i1 [[CMP14]], label [[OMP_IF_THEN15:%.*]], label [[OMP_IF_ELSE21:%.*]]
1481 // CHECK2:       omp_if.then15:
1482 // CHECK2-NEXT:    [[TMP46:%.*]] = mul nuw i64 [[TMP2]], 4
1483 // CHECK2-NEXT:    [[TMP47:%.*]] = mul nuw i64 5, [[TMP5]]
1484 // CHECK2-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], 8
1485 // CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
1486 // CHECK2-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i64*
1487 // CHECK2-NEXT:    store i64 [[TMP44]], i64* [[TMP50]], align 8
1488 // CHECK2-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
1489 // CHECK2-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i64*
1490 // CHECK2-NEXT:    store i64 [[TMP44]], i64* [[TMP52]], align 8
1491 // CHECK2-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1492 // CHECK2-NEXT:    store i64 4, i64* [[TMP53]], align 8
1493 // CHECK2-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
1494 // CHECK2-NEXT:    store i8* null, i8** [[TMP54]], align 8
1495 // CHECK2-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
1496 // CHECK2-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
1497 // CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 8
1498 // CHECK2-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
1499 // CHECK2-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
1500 // CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 8
1501 // CHECK2-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
1502 // CHECK2-NEXT:    store i64 40, i64* [[TMP59]], align 8
1503 // CHECK2-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
1504 // CHECK2-NEXT:    store i8* null, i8** [[TMP60]], align 8
1505 // CHECK2-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 2
1506 // CHECK2-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
1507 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP62]], align 8
1508 // CHECK2-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 2
1509 // CHECK2-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
1510 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP64]], align 8
1511 // CHECK2-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1512 // CHECK2-NEXT:    store i64 8, i64* [[TMP65]], align 8
1513 // CHECK2-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 2
1514 // CHECK2-NEXT:    store i8* null, i8** [[TMP66]], align 8
1515 // CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 3
1516 // CHECK2-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
1517 // CHECK2-NEXT:    store float* [[VLA]], float** [[TMP68]], align 8
1518 // CHECK2-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 3
1519 // CHECK2-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
1520 // CHECK2-NEXT:    store float* [[VLA]], float** [[TMP70]], align 8
1521 // CHECK2-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
1522 // CHECK2-NEXT:    store i64 [[TMP46]], i64* [[TMP71]], align 8
1523 // CHECK2-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 3
1524 // CHECK2-NEXT:    store i8* null, i8** [[TMP72]], align 8
1525 // CHECK2-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 4
1526 // CHECK2-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
1527 // CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 8
1528 // CHECK2-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 4
1529 // CHECK2-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
1530 // CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 8
1531 // CHECK2-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
1532 // CHECK2-NEXT:    store i64 400, i64* [[TMP77]], align 8
1533 // CHECK2-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 4
1534 // CHECK2-NEXT:    store i8* null, i8** [[TMP78]], align 8
1535 // CHECK2-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 5
1536 // CHECK2-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i64*
1537 // CHECK2-NEXT:    store i64 5, i64* [[TMP80]], align 8
1538 // CHECK2-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 5
1539 // CHECK2-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
1540 // CHECK2-NEXT:    store i64 5, i64* [[TMP82]], align 8
1541 // CHECK2-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
1542 // CHECK2-NEXT:    store i64 8, i64* [[TMP83]], align 8
1543 // CHECK2-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 5
1544 // CHECK2-NEXT:    store i8* null, i8** [[TMP84]], align 8
1545 // CHECK2-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 6
1546 // CHECK2-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
1547 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP86]], align 8
1548 // CHECK2-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 6
1549 // CHECK2-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
1550 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP88]], align 8
1551 // CHECK2-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
1552 // CHECK2-NEXT:    store i64 8, i64* [[TMP89]], align 8
1553 // CHECK2-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 6
1554 // CHECK2-NEXT:    store i8* null, i8** [[TMP90]], align 8
1555 // CHECK2-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 7
1556 // CHECK2-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
1557 // CHECK2-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 8
1558 // CHECK2-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 7
1559 // CHECK2-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
1560 // CHECK2-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 8
1561 // CHECK2-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
1562 // CHECK2-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 8
1563 // CHECK2-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 7
1564 // CHECK2-NEXT:    store i8* null, i8** [[TMP96]], align 8
1565 // CHECK2-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 8
1566 // CHECK2-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
1567 // CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 8
1568 // CHECK2-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 8
1569 // CHECK2-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
1570 // CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 8
1571 // CHECK2-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
1572 // CHECK2-NEXT:    store i64 16, i64* [[TMP101]], align 8
1573 // CHECK2-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 8
1574 // CHECK2-NEXT:    store i8* null, i8** [[TMP102]], align 8
1575 // CHECK2-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
1576 // CHECK2-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
1577 // CHECK2-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1578 // CHECK2-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
1579 // CHECK2-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
1580 // CHECK2-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
1581 // CHECK2:       omp_offload.failed19:
1582 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
1583 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
1584 // CHECK2:       omp_offload.cont20:
1585 // CHECK2-NEXT:    br label [[OMP_IF_END22:%.*]]
1586 // CHECK2:       omp_if.else21:
1587 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
1588 // CHECK2-NEXT:    br label [[OMP_IF_END22]]
1589 // CHECK2:       omp_if.end22:
1590 // CHECK2-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
1591 // CHECK2-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1592 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
1593 // CHECK2-NEXT:    ret i32 [[TMP108]]
1594 //
1595 //
1596 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
1597 // CHECK2-SAME: () #[[ATTR2:[0-9]+]] {
1598 // CHECK2-NEXT:  entry:
1599 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
1600 // 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_]])
1601 // CHECK2-NEXT:    ret void
1602 //
1603 //
1604 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
1605 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
1606 // CHECK2-NEXT:  entry:
1607 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1608 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1609 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
1610 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1611 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1612 // CHECK2-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
1613 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
1614 // CHECK2-NEXT:    ret void
1615 //
1616 //
1617 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
1618 // CHECK2-SAME: (i32 signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
1619 // CHECK2-NEXT:  entry:
1620 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
1621 // CHECK2-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
1622 // CHECK2-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
1623 // CHECK2-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
1624 // CHECK2-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
1625 // CHECK2-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 8
1626 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
1627 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
1628 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
1629 // CHECK2-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
1630 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
1631 // CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
1632 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
1633 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
1634 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
1635 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1636 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
1637 // CHECK2-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
1638 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
1639 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
1640 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
1641 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
1642 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !21
1643 // CHECK2-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !21
1644 // CHECK2-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
1645 // CHECK2-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
1646 // CHECK2-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !21
1647 // CHECK2-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
1648 // CHECK2-NEXT:    [[TMP10:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
1649 // CHECK2-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0) #[[ATTR3]]
1650 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1651 // CHECK2-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
1652 // CHECK2:       omp_offload.failed.i:
1653 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR3]]
1654 // CHECK2-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
1655 // CHECK2:       .omp_outlined..1.exit:
1656 // CHECK2-NEXT:    ret i32 0
1657 //
1658 //
1659 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
1660 // CHECK2-SAME: (i64 [[A:%.*]]) #[[ATTR2]] {
1661 // CHECK2-NEXT:  entry:
1662 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1663 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
1664 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1665 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1666 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1667 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
1668 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
1669 // 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..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
1670 // CHECK2-NEXT:    ret void
1671 //
1672 //
1673 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
1674 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
1675 // CHECK2-NEXT:  entry:
1676 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1677 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1678 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
1679 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
1680 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1681 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1682 // CHECK2-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
1683 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
1684 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
1685 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1686 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
1687 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
1688 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], 1
1689 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1690 // CHECK2-NEXT:    ret void
1691 //
1692 //
1693 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
1694 // CHECK2-SAME: (i64 [[AA:%.*]]) #[[ATTR2]] {
1695 // CHECK2-NEXT:  entry:
1696 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1697 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 2
1698 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1699 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1700 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1701 // CHECK2-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
1702 // CHECK2-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
1703 // 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..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
1704 // CHECK2-NEXT:    ret void
1705 //
1706 //
1707 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
1708 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
1709 // CHECK2-NEXT:  entry:
1710 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1711 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1712 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
1713 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
1714 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1715 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1716 // CHECK2-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
1717 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
1718 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
1719 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
1720 // CHECK2-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
1721 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
1722 // CHECK2-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
1723 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
1724 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
1725 // CHECK2-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
1726 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1727 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
1728 // CHECK2-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
1729 // CHECK2-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
1730 // CHECK2-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
1731 // CHECK2:       .cancel.exit:
1732 // CHECK2-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
1733 // CHECK2-NEXT:    br label [[DOTCANCEL_CONTINUE]]
1734 // CHECK2:       .cancel.continue:
1735 // CHECK2-NEXT:    ret void
1736 //
1737 //
1738 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
1739 // CHECK2-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR2]] {
1740 // CHECK2-NEXT:  entry:
1741 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1742 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1743 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 4
1744 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1745 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1746 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1747 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1748 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1749 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
1750 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
1751 // CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1752 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
1753 // CHECK2-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
1754 // 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..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
1755 // CHECK2-NEXT:    ret void
1756 //
1757 //
1758 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
1759 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
1760 // CHECK2-NEXT:  entry:
1761 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1762 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1763 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
1764 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
1765 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
1766 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1767 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1768 // CHECK2-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
1769 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
1770 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
1771 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1772 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
1773 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
1774 // CHECK2-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
1775 // CHECK2-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
1776 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
1777 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
1778 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1779 // CHECK2-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
1780 // CHECK2-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
1781 // CHECK2-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
1782 // CHECK2-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
1783 // CHECK2-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
1784 // CHECK2-NEXT:    ret void
1785 //
1786 //
1787 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
1788 // CHECK2-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR2]] {
1789 // CHECK2-NEXT:  entry:
1790 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1791 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
1792 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1793 // CHECK2-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
1794 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
1795 // CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
1796 // CHECK2-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
1797 // CHECK2-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
1798 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
1799 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
1800 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1801 // CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
1802 // CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1803 // CHECK2-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
1804 // CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
1805 // CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
1806 // CHECK2-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
1807 // CHECK2-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
1808 // CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
1809 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1810 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
1811 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1812 // CHECK2-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
1813 // CHECK2-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
1814 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
1815 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
1816 // CHECK2-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
1817 // CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
1818 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1819 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
1820 // CHECK2-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 8
1821 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1822 // CHECK2-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 8
1823 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
1824 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP11]], align 8
1825 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
1826 // CHECK2-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 8
1827 // CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
1828 // CHECK2-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 8
1829 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
1830 // CHECK2-NEXT:    store i64 [[TMP4]], i64* [[TMP14]], align 8
1831 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
1832 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP15]], align 8
1833 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
1834 // CHECK2-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 8
1835 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
1836 // CHECK2-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 8
1837 // 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..7 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
1838 // CHECK2-NEXT:    ret void
1839 //
1840 //
1841 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..7
1842 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
1843 // CHECK2-NEXT:  entry:
1844 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1845 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1846 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
1847 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
1848 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1849 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1850 // CHECK2-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
1851 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
1852 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
1853 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
1854 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
1855 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
1856 // CHECK2-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 8
1857 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
1858 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
1859 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
1860 // CHECK2-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 8
1861 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
1862 // CHECK2-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 8
1863 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 5
1864 // CHECK2-NEXT:    [[TMP12:%.*]] = load i64, i64* [[TMP11]], align 8
1865 // CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 6
1866 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[TMP13]], align 8
1867 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 7
1868 // CHECK2-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 8
1869 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 8
1870 // CHECK2-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 8
1871 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
1872 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
1873 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1874 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i64 0, i64 2
1875 // CHECK2-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
1876 // CHECK2-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
1877 // CHECK2-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
1878 // CHECK2-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
1879 // CHECK2-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
1880 // CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 3
1881 // CHECK2-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
1882 // CHECK2-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
1883 // CHECK2-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
1884 // CHECK2-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
1885 // CHECK2-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
1886 // CHECK2-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i64 0, i64 1
1887 // CHECK2-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i64 0, i64 2
1888 // CHECK2-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
1889 // CHECK2-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
1890 // CHECK2-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
1891 // CHECK2-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP14]]
1892 // CHECK2-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i64 [[TMP23]]
1893 // CHECK2-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i64 3
1894 // CHECK2-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
1895 // CHECK2-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
1896 // CHECK2-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
1897 // CHECK2-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
1898 // CHECK2-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8
1899 // CHECK2-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
1900 // CHECK2-NEXT:    store i64 [[ADD13]], i64* [[X]], align 8
1901 // CHECK2-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
1902 // CHECK2-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8
1903 // CHECK2-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
1904 // CHECK2-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
1905 // CHECK2-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
1906 // CHECK2-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 8
1907 // CHECK2-NEXT:    ret void
1908 //
1909 //
1910 // CHECK2-LABEL: define {{[^@]+}}@_Z3bari
1911 // CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
1912 // CHECK2-NEXT:  entry:
1913 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1914 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
1915 // CHECK2-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
1916 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1917 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
1918 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1919 // CHECK2-NEXT:    [[CALL:%.*]] = call signext i32 @_Z3fooi(i32 signext [[TMP0]])
1920 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
1921 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
1922 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1923 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
1924 // CHECK2-NEXT:    [[CALL1:%.*]] = call signext i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], i32 signext [[TMP2]])
1925 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
1926 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
1927 // CHECK2-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
1928 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
1929 // CHECK2-NEXT:    [[CALL3:%.*]] = call signext i32 @_ZL7fstatici(i32 signext [[TMP4]])
1930 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
1931 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
1932 // CHECK2-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
1933 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
1934 // CHECK2-NEXT:    [[CALL5:%.*]] = call signext i32 @_Z9ftemplateIiET_i(i32 signext [[TMP6]])
1935 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
1936 // CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
1937 // CHECK2-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
1938 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
1939 // CHECK2-NEXT:    ret i32 [[TMP8]]
1940 //
1941 //
1942 // CHECK2-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
1943 // CHECK2-SAME: (%struct.S1* nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
1944 // CHECK2-NEXT:  entry:
1945 // CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1946 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1947 // CHECK2-NEXT:    [[B:%.*]] = alloca i32, align 4
1948 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
1949 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1950 // CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
1951 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
1952 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
1953 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
1954 // CHECK2-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
1955 // CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1956 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1957 // CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1958 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1959 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
1960 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
1961 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
1962 // CHECK2-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
1963 // CHECK2-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
1964 // CHECK2-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
1965 // CHECK2-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
1966 // CHECK2-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
1967 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
1968 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
1969 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
1970 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
1971 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
1972 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
1973 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
1974 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1975 // CHECK2:       omp_if.then:
1976 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
1977 // CHECK2-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
1978 // CHECK2-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
1979 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1980 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
1981 // CHECK2-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 8
1982 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1983 // CHECK2-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
1984 // CHECK2-NEXT:    store double* [[A]], double** [[TMP13]], align 8
1985 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1986 // CHECK2-NEXT:    store i64 8, i64* [[TMP14]], align 8
1987 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1988 // CHECK2-NEXT:    store i8* null, i8** [[TMP15]], align 8
1989 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1990 // CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
1991 // CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
1992 // CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1993 // CHECK2-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
1994 // CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
1995 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
1996 // CHECK2-NEXT:    store i64 4, i64* [[TMP20]], align 8
1997 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1998 // CHECK2-NEXT:    store i8* null, i8** [[TMP21]], align 8
1999 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2000 // CHECK2-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
2001 // CHECK2-NEXT:    store i64 2, i64* [[TMP23]], align 8
2002 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2003 // CHECK2-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
2004 // CHECK2-NEXT:    store i64 2, i64* [[TMP25]], align 8
2005 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
2006 // CHECK2-NEXT:    store i64 8, i64* [[TMP26]], align 8
2007 // CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
2008 // CHECK2-NEXT:    store i8* null, i8** [[TMP27]], align 8
2009 // CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
2010 // CHECK2-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
2011 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
2012 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
2013 // CHECK2-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
2014 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP31]], align 8
2015 // CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
2016 // CHECK2-NEXT:    store i64 8, i64* [[TMP32]], align 8
2017 // CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
2018 // CHECK2-NEXT:    store i8* null, i8** [[TMP33]], align 8
2019 // CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
2020 // CHECK2-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
2021 // CHECK2-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 8
2022 // CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
2023 // CHECK2-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
2024 // CHECK2-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 8
2025 // CHECK2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
2026 // CHECK2-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 8
2027 // CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
2028 // CHECK2-NEXT:    store i8* null, i8** [[TMP39]], align 8
2029 // CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2030 // CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2031 // CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2032 // CHECK2-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
2033 // CHECK2-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
2034 // CHECK2-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2035 // CHECK2:       omp_offload.failed:
2036 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
2037 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2038 // CHECK2:       omp_offload.cont:
2039 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
2040 // CHECK2:       omp_if.else:
2041 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
2042 // CHECK2-NEXT:    br label [[OMP_IF_END]]
2043 // CHECK2:       omp_if.end:
2044 // CHECK2-NEXT:    [[TMP45:%.*]] = mul nsw i64 1, [[TMP2]]
2045 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP45]]
2046 // CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
2047 // CHECK2-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
2048 // CHECK2-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP46]] to i32
2049 // CHECK2-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
2050 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP47]]
2051 // CHECK2-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
2052 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
2053 // CHECK2-NEXT:    ret i32 [[ADD4]]
2054 //
2055 //
2056 // CHECK2-LABEL: define {{[^@]+}}@_ZL7fstatici
2057 // CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
2058 // CHECK2-NEXT:  entry:
2059 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2060 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
2061 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
2062 // CHECK2-NEXT:    [[AAA:%.*]] = alloca i8, align 1
2063 // CHECK2-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
2064 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2065 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2066 // CHECK2-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
2067 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
2068 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
2069 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
2070 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2071 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
2072 // CHECK2-NEXT:    store i16 0, i16* [[AA]], align 2
2073 // CHECK2-NEXT:    store i8 0, i8* [[AAA]], align 1
2074 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
2075 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2076 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
2077 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
2078 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
2079 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2080 // CHECK2-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
2081 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2082 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
2083 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
2084 // CHECK2-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
2085 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
2086 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
2087 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
2088 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2089 // CHECK2:       omp_if.then:
2090 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2091 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
2092 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
2093 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2094 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
2095 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
2096 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2097 // CHECK2-NEXT:    store i8* null, i8** [[TMP11]], align 8
2098 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2099 // CHECK2-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
2100 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
2101 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2102 // CHECK2-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
2103 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
2104 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
2105 // CHECK2-NEXT:    store i8* null, i8** [[TMP16]], align 8
2106 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2107 // CHECK2-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
2108 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
2109 // CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2110 // CHECK2-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
2111 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
2112 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
2113 // CHECK2-NEXT:    store i8* null, i8** [[TMP21]], align 8
2114 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
2115 // CHECK2-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
2116 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
2117 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
2118 // CHECK2-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
2119 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
2120 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
2121 // CHECK2-NEXT:    store i8* null, i8** [[TMP26]], align 8
2122 // CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2123 // CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2124 // CHECK2-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
2125 // CHECK2-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
2126 // CHECK2-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2127 // CHECK2:       omp_offload.failed:
2128 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
2129 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2130 // CHECK2:       omp_offload.cont:
2131 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
2132 // CHECK2:       omp_if.else:
2133 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
2134 // CHECK2-NEXT:    br label [[OMP_IF_END]]
2135 // CHECK2:       omp_if.end:
2136 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
2137 // CHECK2-NEXT:    ret i32 [[TMP31]]
2138 //
2139 //
2140 // CHECK2-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
2141 // CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] comdat {
2142 // CHECK2-NEXT:  entry:
2143 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2144 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
2145 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
2146 // CHECK2-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
2147 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2148 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2149 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
2150 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
2151 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
2152 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2153 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
2154 // CHECK2-NEXT:    store i16 0, i16* [[AA]], align 2
2155 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
2156 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2157 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
2158 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
2159 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
2160 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2161 // CHECK2-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
2162 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2163 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
2164 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
2165 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2166 // CHECK2:       omp_if.then:
2167 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2168 // CHECK2-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
2169 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
2170 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2171 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
2172 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
2173 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2174 // CHECK2-NEXT:    store i8* null, i8** [[TMP9]], align 8
2175 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2176 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
2177 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
2178 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2179 // CHECK2-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
2180 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
2181 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
2182 // CHECK2-NEXT:    store i8* null, i8** [[TMP14]], align 8
2183 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2184 // CHECK2-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
2185 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
2186 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2187 // CHECK2-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
2188 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
2189 // CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
2190 // CHECK2-NEXT:    store i8* null, i8** [[TMP19]], align 8
2191 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2192 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2193 // CHECK2-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
2194 // CHECK2-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
2195 // CHECK2-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2196 // CHECK2:       omp_offload.failed:
2197 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
2198 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2199 // CHECK2:       omp_offload.cont:
2200 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
2201 // CHECK2:       omp_if.else:
2202 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
2203 // CHECK2-NEXT:    br label [[OMP_IF_END]]
2204 // CHECK2:       omp_if.end:
2205 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
2206 // CHECK2-NEXT:    ret i32 [[TMP24]]
2207 //
2208 //
2209 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
2210 // CHECK2-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
2211 // CHECK2-NEXT:  entry:
2212 // CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
2213 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
2214 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
2215 // CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
2216 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
2217 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
2218 // CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
2219 // CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
2220 // CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2221 // CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
2222 // CHECK2-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
2223 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
2224 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
2225 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2226 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
2227 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
2228 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2229 // CHECK2-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 8
2230 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2231 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[CONV]], align 8
2232 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
2233 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
2234 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP7]], align 8
2235 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
2236 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP8]], align 8
2237 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
2238 // CHECK2-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 8
2239 // 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.5*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
2240 // CHECK2-NEXT:    ret void
2241 //
2242 //
2243 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..9
2244 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
2245 // CHECK2-NEXT:  entry:
2246 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2247 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2248 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
2249 // CHECK2-NEXT:    [[B:%.*]] = alloca i32, align 4
2250 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2251 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2252 // CHECK2-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
2253 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
2254 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
2255 // CHECK2-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
2256 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
2257 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
2258 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
2259 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
2260 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
2261 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
2262 // CHECK2-NEXT:    [[TMP8:%.*]] = load i64, i64* [[TMP7]], align 8
2263 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
2264 // CHECK2-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 8
2265 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
2266 // CHECK2-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
2267 // CHECK2-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
2268 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
2269 // CHECK2-NEXT:    store double [[ADD]], double* [[A]], align 8
2270 // CHECK2-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
2271 // CHECK2-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 8
2272 // CHECK2-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
2273 // CHECK2-NEXT:    store double [[INC]], double* [[A1]], align 8
2274 // CHECK2-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
2275 // CHECK2-NEXT:    [[TMP13:%.*]] = mul nsw i64 1, [[TMP8]]
2276 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i64 [[TMP13]]
2277 // CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
2278 // CHECK2-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
2279 // CHECK2-NEXT:    ret void
2280 //
2281 //
2282 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
2283 // CHECK2-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
2284 // CHECK2-NEXT:  entry:
2285 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2286 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2287 // CHECK2-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
2288 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
2289 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 8
2290 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2291 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2292 // CHECK2-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
2293 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
2294 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2295 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2296 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
2297 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
2298 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2299 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
2300 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
2301 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2302 // CHECK2-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
2303 // CHECK2-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
2304 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
2305 // CHECK2-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV2]], align 8
2306 // CHECK2-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
2307 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
2308 // CHECK2-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 8
2309 // 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.6*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
2310 // CHECK2-NEXT:    ret void
2311 //
2312 //
2313 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..11
2314 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
2315 // CHECK2-NEXT:  entry:
2316 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2317 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2318 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 8
2319 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
2320 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
2321 // CHECK2-NEXT:    [[AAA:%.*]] = alloca i8, align 1
2322 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2323 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2324 // CHECK2-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 8
2325 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 8
2326 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
2327 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
2328 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
2329 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
2330 // CHECK2-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
2331 // CHECK2-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
2332 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
2333 // CHECK2-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
2334 // CHECK2-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
2335 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
2336 // CHECK2-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 8
2337 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
2338 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
2339 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2340 // CHECK2-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
2341 // CHECK2-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
2342 // CHECK2-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
2343 // CHECK2-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
2344 // CHECK2-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
2345 // CHECK2-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
2346 // CHECK2-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
2347 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
2348 // CHECK2-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
2349 // CHECK2-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
2350 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i64 0, i64 2
2351 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2352 // CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
2353 // CHECK2-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
2354 // CHECK2-NEXT:    ret void
2355 //
2356 //
2357 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
2358 // CHECK2-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
2359 // CHECK2-NEXT:  entry:
2360 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2361 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2362 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
2363 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 8
2364 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2365 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2366 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
2367 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2368 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2369 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
2370 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2371 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
2372 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
2373 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2374 // CHECK2-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
2375 // CHECK2-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
2376 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
2377 // CHECK2-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 8
2378 // 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.7*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
2379 // CHECK2-NEXT:    ret void
2380 //
2381 //
2382 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..14
2383 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
2384 // CHECK2-NEXT:  entry:
2385 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2386 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2387 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 8
2388 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
2389 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
2390 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2391 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2392 // CHECK2-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 8
2393 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 8
2394 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
2395 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
2396 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
2397 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
2398 // CHECK2-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
2399 // CHECK2-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
2400 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
2401 // CHECK2-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 8
2402 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
2403 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
2404 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2405 // CHECK2-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
2406 // CHECK2-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
2407 // CHECK2-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
2408 // CHECK2-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
2409 // CHECK2-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
2410 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i64 0, i64 2
2411 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2412 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
2413 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
2414 // CHECK2-NEXT:    ret void
2415 //
2416 //
2417 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2418 // CHECK2-SAME: () #[[ATTR6:[0-9]+]] {
2419 // CHECK2-NEXT:  entry:
2420 // CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
2421 // CHECK2-NEXT:    ret void
2422 //
2423 //
2424 // CHECK3-LABEL: define {{[^@]+}}@_Z3fooi
2425 // CHECK3-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
2426 // CHECK3-NEXT:  entry:
2427 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2428 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
2429 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
2430 // CHECK3-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
2431 // CHECK3-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
2432 // CHECK3-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
2433 // CHECK3-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
2434 // CHECK3-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
2435 // CHECK3-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
2436 // CHECK3-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
2437 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
2438 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
2439 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
2440 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
2441 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
2442 // CHECK3-NEXT:    [[A_CASTED2:%.*]] = alloca i32, align 4
2443 // CHECK3-NEXT:    [[AA_CASTED3:%.*]] = alloca i32, align 4
2444 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4
2445 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4
2446 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4
2447 // CHECK3-NEXT:    [[A_CASTED10:%.*]] = alloca i32, align 4
2448 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [9 x i8*], align 4
2449 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [9 x i8*], align 4
2450 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [9 x i8*], align 4
2451 // CHECK3-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 4
2452 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
2453 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2454 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
2455 // CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
2456 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
2457 // CHECK3-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
2458 // CHECK3-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
2459 // CHECK3-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
2460 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
2461 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
2462 // CHECK3-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
2463 // CHECK3-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
2464 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
2465 // CHECK3-NEXT:    [[TMP5:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i32 20, i32 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
2466 // CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to %struct.kmp_task_t_with_privates*
2467 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP6]], i32 0, i32 0
2468 // CHECK3-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP5]])
2469 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
2470 // CHECK3-NEXT:    store i32 [[TMP9]], i32* [[A_CASTED]], align 4
2471 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[A_CASTED]], align 4
2472 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i32 [[TMP10]]) #[[ATTR3:[0-9]+]]
2473 // CHECK3-NEXT:    [[TMP11:%.*]] = load i16, i16* [[AA]], align 2
2474 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
2475 // CHECK3-NEXT:    store i16 [[TMP11]], i16* [[CONV]], align 2
2476 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[AA_CASTED]], align 4
2477 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2478 // CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
2479 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP14]], align 4
2480 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2481 // CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
2482 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP16]], align 4
2483 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2484 // CHECK3-NEXT:    store i8* null, i8** [[TMP17]], align 4
2485 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2486 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2487 // CHECK3-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP18]], i8** [[TMP19]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
2488 // CHECK3-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
2489 // CHECK3-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2490 // CHECK3:       omp_offload.failed:
2491 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i32 [[TMP12]]) #[[ATTR3]]
2492 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2493 // CHECK3:       omp_offload.cont:
2494 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[A]], align 4
2495 // CHECK3-NEXT:    store i32 [[TMP22]], i32* [[A_CASTED2]], align 4
2496 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A_CASTED2]], align 4
2497 // CHECK3-NEXT:    [[TMP24:%.*]] = load i16, i16* [[AA]], align 2
2498 // CHECK3-NEXT:    [[CONV4:%.*]] = bitcast i32* [[AA_CASTED3]] to i16*
2499 // CHECK3-NEXT:    store i16 [[TMP24]], i16* [[CONV4]], align 2
2500 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[AA_CASTED3]], align 4
2501 // CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
2502 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP26]], 10
2503 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2504 // CHECK3:       omp_if.then:
2505 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
2506 // CHECK3-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
2507 // CHECK3-NEXT:    store i32 [[TMP23]], i32* [[TMP28]], align 4
2508 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
2509 // CHECK3-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
2510 // CHECK3-NEXT:    store i32 [[TMP23]], i32* [[TMP30]], align 4
2511 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
2512 // CHECK3-NEXT:    store i8* null, i8** [[TMP31]], align 4
2513 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
2514 // CHECK3-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i32*
2515 // CHECK3-NEXT:    store i32 [[TMP25]], i32* [[TMP33]], align 4
2516 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
2517 // CHECK3-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i32*
2518 // CHECK3-NEXT:    store i32 [[TMP25]], i32* [[TMP35]], align 4
2519 // CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
2520 // CHECK3-NEXT:    store i8* null, i8** [[TMP36]], align 4
2521 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
2522 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
2523 // CHECK3-NEXT:    [[TMP39:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP37]], i8** [[TMP38]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
2524 // CHECK3-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
2525 // CHECK3-NEXT:    br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
2526 // CHECK3:       omp_offload.failed8:
2527 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR3]]
2528 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
2529 // CHECK3:       omp_offload.cont9:
2530 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
2531 // CHECK3:       omp_if.else:
2532 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR3]]
2533 // CHECK3-NEXT:    br label [[OMP_IF_END]]
2534 // CHECK3:       omp_if.end:
2535 // CHECK3-NEXT:    [[TMP41:%.*]] = load i32, i32* [[A]], align 4
2536 // CHECK3-NEXT:    store i32 [[TMP41]], i32* [[A_CASTED10]], align 4
2537 // CHECK3-NEXT:    [[TMP42:%.*]] = load i32, i32* [[A_CASTED10]], align 4
2538 // CHECK3-NEXT:    [[TMP43:%.*]] = load i32, i32* [[N_ADDR]], align 4
2539 // CHECK3-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP43]], 20
2540 // CHECK3-NEXT:    br i1 [[CMP11]], label [[OMP_IF_THEN12:%.*]], label [[OMP_IF_ELSE18:%.*]]
2541 // CHECK3:       omp_if.then12:
2542 // CHECK3-NEXT:    [[TMP44:%.*]] = mul nuw i32 [[TMP1]], 4
2543 // CHECK3-NEXT:    [[TMP45:%.*]] = sext i32 [[TMP44]] to i64
2544 // CHECK3-NEXT:    [[TMP46:%.*]] = mul nuw i32 5, [[TMP3]]
2545 // CHECK3-NEXT:    [[TMP47:%.*]] = mul nuw i32 [[TMP46]], 8
2546 // CHECK3-NEXT:    [[TMP48:%.*]] = sext i32 [[TMP47]] to i64
2547 // CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
2548 // CHECK3-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32*
2549 // CHECK3-NEXT:    store i32 [[TMP42]], i32* [[TMP50]], align 4
2550 // CHECK3-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
2551 // CHECK3-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
2552 // CHECK3-NEXT:    store i32 [[TMP42]], i32* [[TMP52]], align 4
2553 // CHECK3-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2554 // CHECK3-NEXT:    store i64 4, i64* [[TMP53]], align 4
2555 // CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
2556 // CHECK3-NEXT:    store i8* null, i8** [[TMP54]], align 4
2557 // CHECK3-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 1
2558 // CHECK3-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
2559 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 4
2560 // CHECK3-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 1
2561 // CHECK3-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
2562 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 4
2563 // CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
2564 // CHECK3-NEXT:    store i64 40, i64* [[TMP59]], align 4
2565 // CHECK3-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 1
2566 // CHECK3-NEXT:    store i8* null, i8** [[TMP60]], align 4
2567 // CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 2
2568 // CHECK3-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
2569 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP62]], align 4
2570 // CHECK3-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 2
2571 // CHECK3-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i32*
2572 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP64]], align 4
2573 // CHECK3-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
2574 // CHECK3-NEXT:    store i64 4, i64* [[TMP65]], align 4
2575 // CHECK3-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 2
2576 // CHECK3-NEXT:    store i8* null, i8** [[TMP66]], align 4
2577 // CHECK3-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 3
2578 // CHECK3-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
2579 // CHECK3-NEXT:    store float* [[VLA]], float** [[TMP68]], align 4
2580 // CHECK3-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 3
2581 // CHECK3-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
2582 // CHECK3-NEXT:    store float* [[VLA]], float** [[TMP70]], align 4
2583 // CHECK3-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
2584 // CHECK3-NEXT:    store i64 [[TMP45]], i64* [[TMP71]], align 4
2585 // CHECK3-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 3
2586 // CHECK3-NEXT:    store i8* null, i8** [[TMP72]], align 4
2587 // CHECK3-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 4
2588 // CHECK3-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
2589 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 4
2590 // CHECK3-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 4
2591 // CHECK3-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
2592 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 4
2593 // CHECK3-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
2594 // CHECK3-NEXT:    store i64 400, i64* [[TMP77]], align 4
2595 // CHECK3-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 4
2596 // CHECK3-NEXT:    store i8* null, i8** [[TMP78]], align 4
2597 // CHECK3-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 5
2598 // CHECK3-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i32*
2599 // CHECK3-NEXT:    store i32 5, i32* [[TMP80]], align 4
2600 // CHECK3-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 5
2601 // CHECK3-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i32*
2602 // CHECK3-NEXT:    store i32 5, i32* [[TMP82]], align 4
2603 // CHECK3-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
2604 // CHECK3-NEXT:    store i64 4, i64* [[TMP83]], align 4
2605 // CHECK3-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 5
2606 // CHECK3-NEXT:    store i8* null, i8** [[TMP84]], align 4
2607 // CHECK3-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 6
2608 // CHECK3-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i32*
2609 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP86]], align 4
2610 // CHECK3-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 6
2611 // CHECK3-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
2612 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP88]], align 4
2613 // CHECK3-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
2614 // CHECK3-NEXT:    store i64 4, i64* [[TMP89]], align 4
2615 // CHECK3-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 6
2616 // CHECK3-NEXT:    store i8* null, i8** [[TMP90]], align 4
2617 // CHECK3-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 7
2618 // CHECK3-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
2619 // CHECK3-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 4
2620 // CHECK3-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 7
2621 // CHECK3-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
2622 // CHECK3-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 4
2623 // CHECK3-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
2624 // CHECK3-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 4
2625 // CHECK3-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 7
2626 // CHECK3-NEXT:    store i8* null, i8** [[TMP96]], align 4
2627 // CHECK3-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 8
2628 // CHECK3-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
2629 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 4
2630 // CHECK3-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 8
2631 // CHECK3-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
2632 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 4
2633 // CHECK3-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
2634 // CHECK3-NEXT:    store i64 12, i64* [[TMP101]], align 4
2635 // CHECK3-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 8
2636 // CHECK3-NEXT:    store i8* null, i8** [[TMP102]], align 4
2637 // CHECK3-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
2638 // CHECK3-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
2639 // CHECK3-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2640 // CHECK3-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
2641 // CHECK3-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
2642 // CHECK3-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
2643 // CHECK3:       omp_offload.failed16:
2644 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
2645 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
2646 // CHECK3:       omp_offload.cont17:
2647 // CHECK3-NEXT:    br label [[OMP_IF_END19:%.*]]
2648 // CHECK3:       omp_if.else18:
2649 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
2650 // CHECK3-NEXT:    br label [[OMP_IF_END19]]
2651 // CHECK3:       omp_if.end19:
2652 // CHECK3-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
2653 // CHECK3-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
2654 // CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
2655 // CHECK3-NEXT:    ret i32 [[TMP108]]
2656 //
2657 //
2658 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
2659 // CHECK3-SAME: () #[[ATTR2:[0-9]+]] {
2660 // CHECK3-NEXT:  entry:
2661 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
2662 // 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_]])
2663 // CHECK3-NEXT:    ret void
2664 //
2665 //
2666 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
2667 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
2668 // CHECK3-NEXT:  entry:
2669 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2670 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2671 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
2672 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2673 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2674 // CHECK3-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
2675 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
2676 // CHECK3-NEXT:    ret void
2677 //
2678 //
2679 // CHECK3-LABEL: define {{[^@]+}}@.omp_task_entry.
2680 // CHECK3-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
2681 // CHECK3-NEXT:  entry:
2682 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
2683 // CHECK3-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
2684 // CHECK3-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
2685 // CHECK3-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
2686 // CHECK3-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
2687 // CHECK3-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 4
2688 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
2689 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
2690 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
2691 // CHECK3-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
2692 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
2693 // CHECK3-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
2694 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
2695 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
2696 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
2697 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
2698 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
2699 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
2700 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
2701 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
2702 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
2703 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
2704 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !22
2705 // CHECK3-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !22
2706 // CHECK3-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
2707 // CHECK3-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
2708 // CHECK3-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !22
2709 // CHECK3-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
2710 // CHECK3-NEXT:    [[TMP10:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
2711 // CHECK3-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0) #[[ATTR3]]
2712 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2713 // CHECK3-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
2714 // CHECK3:       omp_offload.failed.i:
2715 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR3]]
2716 // CHECK3-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
2717 // CHECK3:       .omp_outlined..1.exit:
2718 // CHECK3-NEXT:    ret i32 0
2719 //
2720 //
2721 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
2722 // CHECK3-SAME: (i32 [[A:%.*]]) #[[ATTR2]] {
2723 // CHECK3-NEXT:  entry:
2724 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2725 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
2726 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2727 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2728 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
2729 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
2730 // 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..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
2731 // CHECK3-NEXT:    ret void
2732 //
2733 //
2734 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
2735 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
2736 // CHECK3-NEXT:  entry:
2737 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2738 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2739 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
2740 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
2741 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2742 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2743 // CHECK3-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
2744 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
2745 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
2746 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2747 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
2748 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
2749 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], 1
2750 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2751 // CHECK3-NEXT:    ret void
2752 //
2753 //
2754 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
2755 // CHECK3-SAME: (i32 [[AA:%.*]]) #[[ATTR2]] {
2756 // CHECK3-NEXT:  entry:
2757 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
2758 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 2
2759 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
2760 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
2761 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2762 // CHECK3-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
2763 // CHECK3-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
2764 // 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..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
2765 // CHECK3-NEXT:    ret void
2766 //
2767 //
2768 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
2769 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
2770 // CHECK3-NEXT:  entry:
2771 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2772 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2773 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
2774 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
2775 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2776 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2777 // CHECK3-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
2778 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
2779 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
2780 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
2781 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
2782 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
2783 // CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
2784 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
2785 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
2786 // CHECK3-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
2787 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2788 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
2789 // CHECK3-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
2790 // CHECK3-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
2791 // CHECK3-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
2792 // CHECK3:       .cancel.exit:
2793 // CHECK3-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
2794 // CHECK3-NEXT:    br label [[DOTCANCEL_CONTINUE]]
2795 // CHECK3:       .cancel.continue:
2796 // CHECK3-NEXT:    ret void
2797 //
2798 //
2799 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
2800 // CHECK3-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR2]] {
2801 // CHECK3-NEXT:  entry:
2802 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2803 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
2804 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 4
2805 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2806 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
2807 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
2808 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2809 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
2810 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
2811 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2812 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
2813 // CHECK3-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
2814 // 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..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
2815 // CHECK3-NEXT:    ret void
2816 //
2817 //
2818 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
2819 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
2820 // CHECK3-NEXT:  entry:
2821 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2822 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2823 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
2824 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
2825 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
2826 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2827 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2828 // CHECK3-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
2829 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
2830 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
2831 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2832 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
2833 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
2834 // CHECK3-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
2835 // CHECK3-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
2836 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
2837 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
2838 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2839 // CHECK3-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
2840 // CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
2841 // CHECK3-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
2842 // CHECK3-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
2843 // CHECK3-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
2844 // CHECK3-NEXT:    ret void
2845 //
2846 //
2847 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
2848 // CHECK3-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR2]] {
2849 // CHECK3-NEXT:  entry:
2850 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2851 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
2852 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2853 // CHECK3-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
2854 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
2855 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
2856 // CHECK3-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
2857 // CHECK3-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
2858 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
2859 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
2860 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2861 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
2862 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2863 // CHECK3-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
2864 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
2865 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
2866 // CHECK3-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
2867 // CHECK3-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
2868 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
2869 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
2870 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2871 // CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
2872 // CHECK3-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
2873 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
2874 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
2875 // CHECK3-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
2876 // CHECK3-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
2877 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2878 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
2879 // CHECK3-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 4
2880 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2881 // CHECK3-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 4
2882 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
2883 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP11]], align 4
2884 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
2885 // CHECK3-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 4
2886 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
2887 // CHECK3-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 4
2888 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
2889 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[TMP14]], align 4
2890 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
2891 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP15]], align 4
2892 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
2893 // CHECK3-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 4
2894 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
2895 // CHECK3-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 4
2896 // 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..7 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
2897 // CHECK3-NEXT:    ret void
2898 //
2899 //
2900 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
2901 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
2902 // CHECK3-NEXT:  entry:
2903 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2904 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2905 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
2906 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
2907 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2908 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2909 // CHECK3-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
2910 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
2911 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
2912 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2913 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
2914 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
2915 // CHECK3-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 4
2916 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
2917 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
2918 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
2919 // CHECK3-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 4
2920 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
2921 // CHECK3-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 4
2922 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 5
2923 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
2924 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 6
2925 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
2926 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 7
2927 // CHECK3-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 4
2928 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 8
2929 // CHECK3-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 4
2930 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
2931 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
2932 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2933 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i32 0, i32 2
2934 // CHECK3-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
2935 // CHECK3-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
2936 // CHECK3-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
2937 // CHECK3-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
2938 // CHECK3-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
2939 // CHECK3-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i32 3
2940 // CHECK3-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
2941 // CHECK3-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
2942 // CHECK3-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
2943 // CHECK3-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
2944 // CHECK3-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
2945 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i32 0, i32 1
2946 // CHECK3-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i32 0, i32 2
2947 // CHECK3-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
2948 // CHECK3-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
2949 // CHECK3-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
2950 // CHECK3-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP14]]
2951 // CHECK3-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i32 [[TMP23]]
2952 // CHECK3-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i32 3
2953 // CHECK3-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
2954 // CHECK3-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
2955 // CHECK3-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
2956 // CHECK3-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
2957 // CHECK3-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4
2958 // CHECK3-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
2959 // CHECK3-NEXT:    store i64 [[ADD13]], i64* [[X]], align 4
2960 // CHECK3-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
2961 // CHECK3-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4
2962 // CHECK3-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
2963 // CHECK3-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
2964 // CHECK3-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
2965 // CHECK3-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 4
2966 // CHECK3-NEXT:    ret void
2967 //
2968 //
2969 // CHECK3-LABEL: define {{[^@]+}}@_Z3bari
2970 // CHECK3-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
2971 // CHECK3-NEXT:  entry:
2972 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2973 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
2974 // CHECK3-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
2975 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2976 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
2977 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
2978 // CHECK3-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooi(i32 [[TMP0]])
2979 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
2980 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
2981 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2982 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
2983 // CHECK3-NEXT:    [[CALL1:%.*]] = call i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], i32 [[TMP2]])
2984 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
2985 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
2986 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
2987 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
2988 // CHECK3-NEXT:    [[CALL3:%.*]] = call i32 @_ZL7fstatici(i32 [[TMP4]])
2989 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
2990 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
2991 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
2992 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
2993 // CHECK3-NEXT:    [[CALL5:%.*]] = call i32 @_Z9ftemplateIiET_i(i32 [[TMP6]])
2994 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
2995 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
2996 // CHECK3-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
2997 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
2998 // CHECK3-NEXT:    ret i32 [[TMP8]]
2999 //
3000 //
3001 // CHECK3-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
3002 // CHECK3-SAME: (%struct.S1* nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 [[N:%.*]]) #[[ATTR0]] comdat align 2 {
3003 // CHECK3-NEXT:  entry:
3004 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
3005 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3006 // CHECK3-NEXT:    [[B:%.*]] = alloca i32, align 4
3007 // CHECK3-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
3008 // CHECK3-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
3009 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
3010 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
3011 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
3012 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
3013 // CHECK3-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
3014 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
3015 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3016 // CHECK3-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
3017 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3018 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
3019 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
3020 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
3021 // CHECK3-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
3022 // CHECK3-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
3023 // CHECK3-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
3024 // CHECK3-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
3025 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
3026 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
3027 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
3028 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
3029 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
3030 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
3031 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3032 // CHECK3:       omp_if.then:
3033 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
3034 // CHECK3-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
3035 // CHECK3-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
3036 // CHECK3-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
3037 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3038 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
3039 // CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 4
3040 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3041 // CHECK3-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
3042 // CHECK3-NEXT:    store double* [[A]], double** [[TMP13]], align 4
3043 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3044 // CHECK3-NEXT:    store i64 8, i64* [[TMP14]], align 4
3045 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3046 // CHECK3-NEXT:    store i8* null, i8** [[TMP15]], align 4
3047 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3048 // CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
3049 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
3050 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3051 // CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
3052 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
3053 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
3054 // CHECK3-NEXT:    store i64 4, i64* [[TMP20]], align 4
3055 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3056 // CHECK3-NEXT:    store i8* null, i8** [[TMP21]], align 4
3057 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3058 // CHECK3-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
3059 // CHECK3-NEXT:    store i32 2, i32* [[TMP23]], align 4
3060 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3061 // CHECK3-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
3062 // CHECK3-NEXT:    store i32 2, i32* [[TMP25]], align 4
3063 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
3064 // CHECK3-NEXT:    store i64 4, i64* [[TMP26]], align 4
3065 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3066 // CHECK3-NEXT:    store i8* null, i8** [[TMP27]], align 4
3067 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
3068 // CHECK3-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
3069 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
3070 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
3071 // CHECK3-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i32*
3072 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP31]], align 4
3073 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
3074 // CHECK3-NEXT:    store i64 4, i64* [[TMP32]], align 4
3075 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
3076 // CHECK3-NEXT:    store i8* null, i8** [[TMP33]], align 4
3077 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
3078 // CHECK3-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
3079 // CHECK3-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 4
3080 // CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
3081 // CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
3082 // CHECK3-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 4
3083 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
3084 // CHECK3-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 4
3085 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
3086 // CHECK3-NEXT:    store i8* null, i8** [[TMP39]], align 4
3087 // CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3088 // CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3089 // CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3090 // CHECK3-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
3091 // CHECK3-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
3092 // CHECK3-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3093 // CHECK3:       omp_offload.failed:
3094 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
3095 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3096 // CHECK3:       omp_offload.cont:
3097 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
3098 // CHECK3:       omp_if.else:
3099 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
3100 // CHECK3-NEXT:    br label [[OMP_IF_END]]
3101 // CHECK3:       omp_if.end:
3102 // CHECK3-NEXT:    [[TMP45:%.*]] = mul nsw i32 1, [[TMP1]]
3103 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP45]]
3104 // CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
3105 // CHECK3-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
3106 // CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP46]] to i32
3107 // CHECK3-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
3108 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP47]]
3109 // CHECK3-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
3110 // CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
3111 // CHECK3-NEXT:    ret i32 [[ADD3]]
3112 //
3113 //
3114 // CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici
3115 // CHECK3-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
3116 // CHECK3-NEXT:  entry:
3117 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3118 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
3119 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
3120 // CHECK3-NEXT:    [[AAA:%.*]] = alloca i8, align 1
3121 // CHECK3-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
3122 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
3123 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
3124 // CHECK3-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
3125 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
3126 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
3127 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
3128 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3129 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
3130 // CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
3131 // CHECK3-NEXT:    store i8 0, i8* [[AAA]], align 1
3132 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
3133 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
3134 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
3135 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
3136 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
3137 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
3138 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
3139 // CHECK3-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
3140 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
3141 // CHECK3-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
3142 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
3143 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
3144 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
3145 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3146 // CHECK3:       omp_if.then:
3147 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3148 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
3149 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
3150 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3151 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
3152 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
3153 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3154 // CHECK3-NEXT:    store i8* null, i8** [[TMP11]], align 4
3155 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3156 // CHECK3-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
3157 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
3158 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3159 // CHECK3-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
3160 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
3161 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3162 // CHECK3-NEXT:    store i8* null, i8** [[TMP16]], align 4
3163 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3164 // CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
3165 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
3166 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3167 // CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
3168 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
3169 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3170 // CHECK3-NEXT:    store i8* null, i8** [[TMP21]], align 4
3171 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
3172 // CHECK3-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
3173 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
3174 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
3175 // CHECK3-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
3176 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
3177 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
3178 // CHECK3-NEXT:    store i8* null, i8** [[TMP26]], align 4
3179 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3180 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3181 // CHECK3-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
3182 // CHECK3-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
3183 // CHECK3-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3184 // CHECK3:       omp_offload.failed:
3185 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
3186 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3187 // CHECK3:       omp_offload.cont:
3188 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
3189 // CHECK3:       omp_if.else:
3190 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
3191 // CHECK3-NEXT:    br label [[OMP_IF_END]]
3192 // CHECK3:       omp_if.end:
3193 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
3194 // CHECK3-NEXT:    ret i32 [[TMP31]]
3195 //
3196 //
3197 // CHECK3-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
3198 // CHECK3-SAME: (i32 [[N:%.*]]) #[[ATTR0]] comdat {
3199 // CHECK3-NEXT:  entry:
3200 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3201 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
3202 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
3203 // CHECK3-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
3204 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
3205 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
3206 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
3207 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
3208 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
3209 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3210 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
3211 // CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
3212 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
3213 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
3214 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
3215 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
3216 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
3217 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
3218 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
3219 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
3220 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
3221 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3222 // CHECK3:       omp_if.then:
3223 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3224 // CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
3225 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
3226 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3227 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
3228 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
3229 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3230 // CHECK3-NEXT:    store i8* null, i8** [[TMP9]], align 4
3231 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3232 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
3233 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
3234 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3235 // CHECK3-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
3236 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
3237 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3238 // CHECK3-NEXT:    store i8* null, i8** [[TMP14]], align 4
3239 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3240 // CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
3241 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
3242 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3243 // CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
3244 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
3245 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3246 // CHECK3-NEXT:    store i8* null, i8** [[TMP19]], align 4
3247 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3248 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3249 // CHECK3-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
3250 // CHECK3-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
3251 // CHECK3-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3252 // CHECK3:       omp_offload.failed:
3253 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
3254 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3255 // CHECK3:       omp_offload.cont:
3256 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
3257 // CHECK3:       omp_if.else:
3258 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
3259 // CHECK3-NEXT:    br label [[OMP_IF_END]]
3260 // CHECK3:       omp_if.end:
3261 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
3262 // CHECK3-NEXT:    ret i32 [[TMP24]]
3263 //
3264 //
3265 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
3266 // CHECK3-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
3267 // CHECK3-NEXT:  entry:
3268 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
3269 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
3270 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
3271 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
3272 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
3273 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 4
3274 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
3275 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
3276 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3277 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
3278 // CHECK3-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
3279 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
3280 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3281 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
3282 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
3283 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3284 // CHECK3-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 4
3285 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3286 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_ADDR]], align 4
3287 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 4
3288 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
3289 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP7]], align 4
3290 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
3291 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[TMP8]], align 4
3292 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
3293 // CHECK3-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 4
3294 // 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.5*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
3295 // CHECK3-NEXT:    ret void
3296 //
3297 //
3298 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..9
3299 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
3300 // CHECK3-NEXT:  entry:
3301 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3302 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3303 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 4
3304 // CHECK3-NEXT:    [[B:%.*]] = alloca i32, align 4
3305 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3306 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3307 // CHECK3-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 4
3308 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 4
3309 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
3310 // CHECK3-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
3311 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
3312 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
3313 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
3314 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
3315 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
3316 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
3317 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3318 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
3319 // CHECK3-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 4
3320 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
3321 // CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
3322 // CHECK3-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
3323 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
3324 // CHECK3-NEXT:    store double [[ADD]], double* [[A]], align 4
3325 // CHECK3-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
3326 // CHECK3-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 4
3327 // CHECK3-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
3328 // CHECK3-NEXT:    store double [[INC]], double* [[A1]], align 4
3329 // CHECK3-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
3330 // CHECK3-NEXT:    [[TMP13:%.*]] = mul nsw i32 1, [[TMP8]]
3331 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i32 [[TMP13]]
3332 // CHECK3-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
3333 // CHECK3-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
3334 // CHECK3-NEXT:    ret void
3335 //
3336 //
3337 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
3338 // CHECK3-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
3339 // CHECK3-NEXT:  entry:
3340 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3341 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3342 // CHECK3-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
3343 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
3344 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 4
3345 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3346 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3347 // CHECK3-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
3348 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
3349 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3350 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
3351 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
3352 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3353 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
3354 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
3355 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3356 // CHECK3-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
3357 // CHECK3-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
3358 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
3359 // CHECK3-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV1]], align 4
3360 // CHECK3-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
3361 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
3362 // CHECK3-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 4
3363 // 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.6*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
3364 // CHECK3-NEXT:    ret void
3365 //
3366 //
3367 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
3368 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
3369 // CHECK3-NEXT:  entry:
3370 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3371 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3372 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 4
3373 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
3374 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
3375 // CHECK3-NEXT:    [[AAA:%.*]] = alloca i8, align 1
3376 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3377 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3378 // CHECK3-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 4
3379 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 4
3380 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
3381 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3382 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
3383 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
3384 // CHECK3-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
3385 // CHECK3-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
3386 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
3387 // CHECK3-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
3388 // CHECK3-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
3389 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
3390 // CHECK3-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 4
3391 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
3392 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
3393 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
3394 // CHECK3-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
3395 // CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
3396 // CHECK3-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
3397 // CHECK3-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
3398 // CHECK3-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
3399 // CHECK3-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
3400 // CHECK3-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
3401 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
3402 // CHECK3-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
3403 // CHECK3-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
3404 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i32 0, i32 2
3405 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
3406 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
3407 // CHECK3-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
3408 // CHECK3-NEXT:    ret void
3409 //
3410 //
3411 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
3412 // CHECK3-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
3413 // CHECK3-NEXT:  entry:
3414 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3415 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3416 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
3417 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 4
3418 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3419 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3420 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
3421 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3422 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
3423 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3424 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
3425 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
3426 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3427 // CHECK3-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
3428 // CHECK3-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
3429 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
3430 // CHECK3-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 4
3431 // 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.7*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
3432 // CHECK3-NEXT:    ret void
3433 //
3434 //
3435 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..14
3436 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
3437 // CHECK3-NEXT:  entry:
3438 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3439 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3440 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 4
3441 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
3442 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
3443 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3444 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3445 // CHECK3-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 4
3446 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 4
3447 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
3448 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3449 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
3450 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
3451 // CHECK3-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
3452 // CHECK3-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
3453 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
3454 // CHECK3-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 4
3455 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
3456 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
3457 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
3458 // CHECK3-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
3459 // CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
3460 // CHECK3-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
3461 // CHECK3-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
3462 // CHECK3-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
3463 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i32 0, i32 2
3464 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
3465 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
3466 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
3467 // CHECK3-NEXT:    ret void
3468 //
3469 //
3470 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3471 // CHECK3-SAME: () #[[ATTR6:[0-9]+]] {
3472 // CHECK3-NEXT:  entry:
3473 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
3474 // CHECK3-NEXT:    ret void
3475 //
3476 //
3477 // CHECK4-LABEL: define {{[^@]+}}@_Z3fooi
3478 // CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
3479 // CHECK4-NEXT:  entry:
3480 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3481 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
3482 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
3483 // CHECK4-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
3484 // CHECK4-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
3485 // CHECK4-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
3486 // CHECK4-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
3487 // CHECK4-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
3488 // CHECK4-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
3489 // CHECK4-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
3490 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
3491 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
3492 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
3493 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
3494 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
3495 // CHECK4-NEXT:    [[A_CASTED2:%.*]] = alloca i32, align 4
3496 // CHECK4-NEXT:    [[AA_CASTED3:%.*]] = alloca i32, align 4
3497 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4
3498 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4
3499 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4
3500 // CHECK4-NEXT:    [[A_CASTED10:%.*]] = alloca i32, align 4
3501 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [9 x i8*], align 4
3502 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [9 x i8*], align 4
3503 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [9 x i8*], align 4
3504 // CHECK4-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 4
3505 // CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
3506 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3507 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
3508 // CHECK4-NEXT:    store i16 0, i16* [[AA]], align 2
3509 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
3510 // CHECK4-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
3511 // CHECK4-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
3512 // CHECK4-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
3513 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
3514 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
3515 // CHECK4-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
3516 // CHECK4-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
3517 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
3518 // CHECK4-NEXT:    [[TMP5:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i32 20, i32 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
3519 // CHECK4-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to %struct.kmp_task_t_with_privates*
3520 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP6]], i32 0, i32 0
3521 // CHECK4-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP5]])
3522 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
3523 // CHECK4-NEXT:    store i32 [[TMP9]], i32* [[A_CASTED]], align 4
3524 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[A_CASTED]], align 4
3525 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i32 [[TMP10]]) #[[ATTR3:[0-9]+]]
3526 // CHECK4-NEXT:    [[TMP11:%.*]] = load i16, i16* [[AA]], align 2
3527 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
3528 // CHECK4-NEXT:    store i16 [[TMP11]], i16* [[CONV]], align 2
3529 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[AA_CASTED]], align 4
3530 // CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3531 // CHECK4-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
3532 // CHECK4-NEXT:    store i32 [[TMP12]], i32* [[TMP14]], align 4
3533 // CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3534 // CHECK4-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
3535 // CHECK4-NEXT:    store i32 [[TMP12]], i32* [[TMP16]], align 4
3536 // CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3537 // CHECK4-NEXT:    store i8* null, i8** [[TMP17]], align 4
3538 // CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3539 // CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3540 // CHECK4-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP18]], i8** [[TMP19]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
3541 // CHECK4-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
3542 // CHECK4-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3543 // CHECK4:       omp_offload.failed:
3544 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i32 [[TMP12]]) #[[ATTR3]]
3545 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3546 // CHECK4:       omp_offload.cont:
3547 // CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[A]], align 4
3548 // CHECK4-NEXT:    store i32 [[TMP22]], i32* [[A_CASTED2]], align 4
3549 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A_CASTED2]], align 4
3550 // CHECK4-NEXT:    [[TMP24:%.*]] = load i16, i16* [[AA]], align 2
3551 // CHECK4-NEXT:    [[CONV4:%.*]] = bitcast i32* [[AA_CASTED3]] to i16*
3552 // CHECK4-NEXT:    store i16 [[TMP24]], i16* [[CONV4]], align 2
3553 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[AA_CASTED3]], align 4
3554 // CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
3555 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP26]], 10
3556 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3557 // CHECK4:       omp_if.then:
3558 // CHECK4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
3559 // CHECK4-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
3560 // CHECK4-NEXT:    store i32 [[TMP23]], i32* [[TMP28]], align 4
3561 // CHECK4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
3562 // CHECK4-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
3563 // CHECK4-NEXT:    store i32 [[TMP23]], i32* [[TMP30]], align 4
3564 // CHECK4-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
3565 // CHECK4-NEXT:    store i8* null, i8** [[TMP31]], align 4
3566 // CHECK4-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
3567 // CHECK4-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i32*
3568 // CHECK4-NEXT:    store i32 [[TMP25]], i32* [[TMP33]], align 4
3569 // CHECK4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
3570 // CHECK4-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i32*
3571 // CHECK4-NEXT:    store i32 [[TMP25]], i32* [[TMP35]], align 4
3572 // CHECK4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
3573 // CHECK4-NEXT:    store i8* null, i8** [[TMP36]], align 4
3574 // CHECK4-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
3575 // CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
3576 // CHECK4-NEXT:    [[TMP39:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP37]], i8** [[TMP38]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
3577 // CHECK4-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
3578 // CHECK4-NEXT:    br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
3579 // CHECK4:       omp_offload.failed8:
3580 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR3]]
3581 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
3582 // CHECK4:       omp_offload.cont9:
3583 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
3584 // CHECK4:       omp_if.else:
3585 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR3]]
3586 // CHECK4-NEXT:    br label [[OMP_IF_END]]
3587 // CHECK4:       omp_if.end:
3588 // CHECK4-NEXT:    [[TMP41:%.*]] = load i32, i32* [[A]], align 4
3589 // CHECK4-NEXT:    store i32 [[TMP41]], i32* [[A_CASTED10]], align 4
3590 // CHECK4-NEXT:    [[TMP42:%.*]] = load i32, i32* [[A_CASTED10]], align 4
3591 // CHECK4-NEXT:    [[TMP43:%.*]] = load i32, i32* [[N_ADDR]], align 4
3592 // CHECK4-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP43]], 20
3593 // CHECK4-NEXT:    br i1 [[CMP11]], label [[OMP_IF_THEN12:%.*]], label [[OMP_IF_ELSE18:%.*]]
3594 // CHECK4:       omp_if.then12:
3595 // CHECK4-NEXT:    [[TMP44:%.*]] = mul nuw i32 [[TMP1]], 4
3596 // CHECK4-NEXT:    [[TMP45:%.*]] = sext i32 [[TMP44]] to i64
3597 // CHECK4-NEXT:    [[TMP46:%.*]] = mul nuw i32 5, [[TMP3]]
3598 // CHECK4-NEXT:    [[TMP47:%.*]] = mul nuw i32 [[TMP46]], 8
3599 // CHECK4-NEXT:    [[TMP48:%.*]] = sext i32 [[TMP47]] to i64
3600 // CHECK4-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
3601 // CHECK4-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32*
3602 // CHECK4-NEXT:    store i32 [[TMP42]], i32* [[TMP50]], align 4
3603 // CHECK4-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
3604 // CHECK4-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
3605 // CHECK4-NEXT:    store i32 [[TMP42]], i32* [[TMP52]], align 4
3606 // CHECK4-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3607 // CHECK4-NEXT:    store i64 4, i64* [[TMP53]], align 4
3608 // CHECK4-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
3609 // CHECK4-NEXT:    store i8* null, i8** [[TMP54]], align 4
3610 // CHECK4-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 1
3611 // CHECK4-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
3612 // CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 4
3613 // CHECK4-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 1
3614 // CHECK4-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
3615 // CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 4
3616 // CHECK4-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
3617 // CHECK4-NEXT:    store i64 40, i64* [[TMP59]], align 4
3618 // CHECK4-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 1
3619 // CHECK4-NEXT:    store i8* null, i8** [[TMP60]], align 4
3620 // CHECK4-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 2
3621 // CHECK4-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
3622 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP62]], align 4
3623 // CHECK4-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 2
3624 // CHECK4-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i32*
3625 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP64]], align 4
3626 // CHECK4-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
3627 // CHECK4-NEXT:    store i64 4, i64* [[TMP65]], align 4
3628 // CHECK4-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 2
3629 // CHECK4-NEXT:    store i8* null, i8** [[TMP66]], align 4
3630 // CHECK4-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 3
3631 // CHECK4-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
3632 // CHECK4-NEXT:    store float* [[VLA]], float** [[TMP68]], align 4
3633 // CHECK4-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 3
3634 // CHECK4-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
3635 // CHECK4-NEXT:    store float* [[VLA]], float** [[TMP70]], align 4
3636 // CHECK4-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
3637 // CHECK4-NEXT:    store i64 [[TMP45]], i64* [[TMP71]], align 4
3638 // CHECK4-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 3
3639 // CHECK4-NEXT:    store i8* null, i8** [[TMP72]], align 4
3640 // CHECK4-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 4
3641 // CHECK4-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
3642 // CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 4
3643 // CHECK4-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 4
3644 // CHECK4-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
3645 // CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 4
3646 // CHECK4-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
3647 // CHECK4-NEXT:    store i64 400, i64* [[TMP77]], align 4
3648 // CHECK4-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 4
3649 // CHECK4-NEXT:    store i8* null, i8** [[TMP78]], align 4
3650 // CHECK4-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 5
3651 // CHECK4-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i32*
3652 // CHECK4-NEXT:    store i32 5, i32* [[TMP80]], align 4
3653 // CHECK4-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 5
3654 // CHECK4-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i32*
3655 // CHECK4-NEXT:    store i32 5, i32* [[TMP82]], align 4
3656 // CHECK4-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
3657 // CHECK4-NEXT:    store i64 4, i64* [[TMP83]], align 4
3658 // CHECK4-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 5
3659 // CHECK4-NEXT:    store i8* null, i8** [[TMP84]], align 4
3660 // CHECK4-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 6
3661 // CHECK4-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i32*
3662 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP86]], align 4
3663 // CHECK4-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 6
3664 // CHECK4-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
3665 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP88]], align 4
3666 // CHECK4-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
3667 // CHECK4-NEXT:    store i64 4, i64* [[TMP89]], align 4
3668 // CHECK4-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 6
3669 // CHECK4-NEXT:    store i8* null, i8** [[TMP90]], align 4
3670 // CHECK4-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 7
3671 // CHECK4-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
3672 // CHECK4-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 4
3673 // CHECK4-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 7
3674 // CHECK4-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
3675 // CHECK4-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 4
3676 // CHECK4-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
3677 // CHECK4-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 4
3678 // CHECK4-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 7
3679 // CHECK4-NEXT:    store i8* null, i8** [[TMP96]], align 4
3680 // CHECK4-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 8
3681 // CHECK4-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
3682 // CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 4
3683 // CHECK4-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 8
3684 // CHECK4-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
3685 // CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 4
3686 // CHECK4-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
3687 // CHECK4-NEXT:    store i64 12, i64* [[TMP101]], align 4
3688 // CHECK4-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 8
3689 // CHECK4-NEXT:    store i8* null, i8** [[TMP102]], align 4
3690 // CHECK4-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
3691 // CHECK4-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
3692 // CHECK4-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3693 // CHECK4-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
3694 // CHECK4-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
3695 // CHECK4-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
3696 // CHECK4:       omp_offload.failed16:
3697 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
3698 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
3699 // CHECK4:       omp_offload.cont17:
3700 // CHECK4-NEXT:    br label [[OMP_IF_END19:%.*]]
3701 // CHECK4:       omp_if.else18:
3702 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
3703 // CHECK4-NEXT:    br label [[OMP_IF_END19]]
3704 // CHECK4:       omp_if.end19:
3705 // CHECK4-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
3706 // CHECK4-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
3707 // CHECK4-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
3708 // CHECK4-NEXT:    ret i32 [[TMP108]]
3709 //
3710 //
3711 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
3712 // CHECK4-SAME: () #[[ATTR2:[0-9]+]] {
3713 // CHECK4-NEXT:  entry:
3714 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
3715 // 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_]])
3716 // CHECK4-NEXT:    ret void
3717 //
3718 //
3719 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
3720 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
3721 // CHECK4-NEXT:  entry:
3722 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3723 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3724 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
3725 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3726 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3727 // CHECK4-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
3728 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
3729 // CHECK4-NEXT:    ret void
3730 //
3731 //
3732 // CHECK4-LABEL: define {{[^@]+}}@.omp_task_entry.
3733 // CHECK4-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
3734 // CHECK4-NEXT:  entry:
3735 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
3736 // CHECK4-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
3737 // CHECK4-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
3738 // CHECK4-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
3739 // CHECK4-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
3740 // CHECK4-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 4
3741 // CHECK4-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
3742 // CHECK4-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
3743 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
3744 // CHECK4-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
3745 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
3746 // CHECK4-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
3747 // CHECK4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
3748 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
3749 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
3750 // CHECK4-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
3751 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
3752 // CHECK4-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
3753 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
3754 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
3755 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
3756 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
3757 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !22
3758 // CHECK4-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !22
3759 // CHECK4-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
3760 // CHECK4-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
3761 // CHECK4-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !22
3762 // CHECK4-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
3763 // CHECK4-NEXT:    [[TMP10:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
3764 // CHECK4-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0) #[[ATTR3]]
3765 // CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3766 // CHECK4-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
3767 // CHECK4:       omp_offload.failed.i:
3768 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR3]]
3769 // CHECK4-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
3770 // CHECK4:       .omp_outlined..1.exit:
3771 // CHECK4-NEXT:    ret i32 0
3772 //
3773 //
3774 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
3775 // CHECK4-SAME: (i32 [[A:%.*]]) #[[ATTR2]] {
3776 // CHECK4-NEXT:  entry:
3777 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3778 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
3779 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3780 // CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3781 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
3782 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
3783 // 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..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
3784 // CHECK4-NEXT:    ret void
3785 //
3786 //
3787 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
3788 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
3789 // CHECK4-NEXT:  entry:
3790 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3791 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3792 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
3793 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
3794 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3795 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3796 // CHECK4-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
3797 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
3798 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
3799 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3800 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
3801 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
3802 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], 1
3803 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
3804 // CHECK4-NEXT:    ret void
3805 //
3806 //
3807 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
3808 // CHECK4-SAME: (i32 [[AA:%.*]]) #[[ATTR2]] {
3809 // CHECK4-NEXT:  entry:
3810 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3811 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 2
3812 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3813 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3814 // CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3815 // CHECK4-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
3816 // CHECK4-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
3817 // 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..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
3818 // CHECK4-NEXT:    ret void
3819 //
3820 //
3821 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3
3822 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
3823 // CHECK4-NEXT:  entry:
3824 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3825 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3826 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
3827 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
3828 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3829 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3830 // CHECK4-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
3831 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
3832 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
3833 // CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
3834 // CHECK4-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
3835 // CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
3836 // CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
3837 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
3838 // CHECK4-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
3839 // CHECK4-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
3840 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3841 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
3842 // CHECK4-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
3843 // CHECK4-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
3844 // CHECK4-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3845 // CHECK4:       .cancel.exit:
3846 // CHECK4-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
3847 // CHECK4-NEXT:    br label [[DOTCANCEL_CONTINUE]]
3848 // CHECK4:       .cancel.continue:
3849 // CHECK4-NEXT:    ret void
3850 //
3851 //
3852 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
3853 // CHECK4-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR2]] {
3854 // CHECK4-NEXT:  entry:
3855 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3856 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3857 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 4
3858 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3859 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3860 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3861 // CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3862 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
3863 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
3864 // CHECK4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3865 // CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
3866 // CHECK4-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
3867 // 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..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
3868 // CHECK4-NEXT:    ret void
3869 //
3870 //
3871 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..4
3872 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
3873 // CHECK4-NEXT:  entry:
3874 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3875 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3876 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
3877 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
3878 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
3879 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3880 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3881 // CHECK4-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
3882 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
3883 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
3884 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3885 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
3886 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
3887 // CHECK4-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
3888 // CHECK4-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
3889 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
3890 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
3891 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
3892 // CHECK4-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
3893 // CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
3894 // CHECK4-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
3895 // CHECK4-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
3896 // CHECK4-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
3897 // CHECK4-NEXT:    ret void
3898 //
3899 //
3900 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
3901 // CHECK4-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR2]] {
3902 // CHECK4-NEXT:  entry:
3903 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3904 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
3905 // CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
3906 // CHECK4-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
3907 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
3908 // CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
3909 // CHECK4-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
3910 // CHECK4-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
3911 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
3912 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
3913 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3914 // CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
3915 // CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3916 // CHECK4-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
3917 // CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
3918 // CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
3919 // CHECK4-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
3920 // CHECK4-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
3921 // CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
3922 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
3923 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3924 // CHECK4-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
3925 // CHECK4-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
3926 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
3927 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
3928 // CHECK4-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
3929 // CHECK4-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
3930 // CHECK4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3931 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
3932 // CHECK4-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 4
3933 // CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3934 // CHECK4-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 4
3935 // CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
3936 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP11]], align 4
3937 // CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
3938 // CHECK4-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 4
3939 // CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
3940 // CHECK4-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 4
3941 // CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
3942 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[TMP14]], align 4
3943 // CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
3944 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP15]], align 4
3945 // CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
3946 // CHECK4-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 4
3947 // CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
3948 // CHECK4-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 4
3949 // 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..7 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
3950 // CHECK4-NEXT:    ret void
3951 //
3952 //
3953 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..7
3954 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
3955 // CHECK4-NEXT:  entry:
3956 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3957 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3958 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
3959 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
3960 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3961 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3962 // CHECK4-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
3963 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
3964 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
3965 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3966 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
3967 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
3968 // CHECK4-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 4
3969 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
3970 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
3971 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
3972 // CHECK4-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 4
3973 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
3974 // CHECK4-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 4
3975 // CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 5
3976 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
3977 // CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 6
3978 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
3979 // CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 7
3980 // CHECK4-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 4
3981 // CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 8
3982 // CHECK4-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 4
3983 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
3984 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
3985 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
3986 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i32 0, i32 2
3987 // CHECK4-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
3988 // CHECK4-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
3989 // CHECK4-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
3990 // CHECK4-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
3991 // CHECK4-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
3992 // CHECK4-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i32 3
3993 // CHECK4-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
3994 // CHECK4-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
3995 // CHECK4-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
3996 // CHECK4-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
3997 // CHECK4-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
3998 // CHECK4-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i32 0, i32 1
3999 // CHECK4-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i32 0, i32 2
4000 // CHECK4-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
4001 // CHECK4-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
4002 // CHECK4-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
4003 // CHECK4-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP14]]
4004 // CHECK4-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i32 [[TMP23]]
4005 // CHECK4-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i32 3
4006 // CHECK4-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
4007 // CHECK4-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
4008 // CHECK4-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
4009 // CHECK4-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
4010 // CHECK4-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4
4011 // CHECK4-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
4012 // CHECK4-NEXT:    store i64 [[ADD13]], i64* [[X]], align 4
4013 // CHECK4-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
4014 // CHECK4-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4
4015 // CHECK4-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
4016 // CHECK4-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
4017 // CHECK4-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
4018 // CHECK4-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 4
4019 // CHECK4-NEXT:    ret void
4020 //
4021 //
4022 // CHECK4-LABEL: define {{[^@]+}}@_Z3bari
4023 // CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
4024 // CHECK4-NEXT:  entry:
4025 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4026 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
4027 // CHECK4-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
4028 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4029 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
4030 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
4031 // CHECK4-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooi(i32 [[TMP0]])
4032 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
4033 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
4034 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
4035 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
4036 // CHECK4-NEXT:    [[CALL1:%.*]] = call i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], i32 [[TMP2]])
4037 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
4038 // CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
4039 // CHECK4-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
4040 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
4041 // CHECK4-NEXT:    [[CALL3:%.*]] = call i32 @_ZL7fstatici(i32 [[TMP4]])
4042 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
4043 // CHECK4-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
4044 // CHECK4-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
4045 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
4046 // CHECK4-NEXT:    [[CALL5:%.*]] = call i32 @_Z9ftemplateIiET_i(i32 [[TMP6]])
4047 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
4048 // CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
4049 // CHECK4-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
4050 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
4051 // CHECK4-NEXT:    ret i32 [[TMP8]]
4052 //
4053 //
4054 // CHECK4-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
4055 // CHECK4-SAME: (%struct.S1* nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 [[N:%.*]]) #[[ATTR0]] comdat align 2 {
4056 // CHECK4-NEXT:  entry:
4057 // CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
4058 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4059 // CHECK4-NEXT:    [[B:%.*]] = alloca i32, align 4
4060 // CHECK4-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
4061 // CHECK4-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
4062 // CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
4063 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
4064 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
4065 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
4066 // CHECK4-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
4067 // CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
4068 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4069 // CHECK4-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
4070 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
4071 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
4072 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
4073 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
4074 // CHECK4-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
4075 // CHECK4-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
4076 // CHECK4-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
4077 // CHECK4-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
4078 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
4079 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
4080 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
4081 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
4082 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
4083 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
4084 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4085 // CHECK4:       omp_if.then:
4086 // CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
4087 // CHECK4-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
4088 // CHECK4-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
4089 // CHECK4-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
4090 // CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4091 // CHECK4-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
4092 // CHECK4-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 4
4093 // CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4094 // CHECK4-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
4095 // CHECK4-NEXT:    store double* [[A]], double** [[TMP13]], align 4
4096 // CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
4097 // CHECK4-NEXT:    store i64 8, i64* [[TMP14]], align 4
4098 // CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4099 // CHECK4-NEXT:    store i8* null, i8** [[TMP15]], align 4
4100 // CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4101 // CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
4102 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
4103 // CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4104 // CHECK4-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
4105 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
4106 // CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
4107 // CHECK4-NEXT:    store i64 4, i64* [[TMP20]], align 4
4108 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4109 // CHECK4-NEXT:    store i8* null, i8** [[TMP21]], align 4
4110 // CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4111 // CHECK4-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
4112 // CHECK4-NEXT:    store i32 2, i32* [[TMP23]], align 4
4113 // CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4114 // CHECK4-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
4115 // CHECK4-NEXT:    store i32 2, i32* [[TMP25]], align 4
4116 // CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
4117 // CHECK4-NEXT:    store i64 4, i64* [[TMP26]], align 4
4118 // CHECK4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4119 // CHECK4-NEXT:    store i8* null, i8** [[TMP27]], align 4
4120 // CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
4121 // CHECK4-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
4122 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
4123 // CHECK4-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
4124 // CHECK4-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i32*
4125 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP31]], align 4
4126 // CHECK4-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
4127 // CHECK4-NEXT:    store i64 4, i64* [[TMP32]], align 4
4128 // CHECK4-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
4129 // CHECK4-NEXT:    store i8* null, i8** [[TMP33]], align 4
4130 // CHECK4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
4131 // CHECK4-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
4132 // CHECK4-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 4
4133 // CHECK4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
4134 // CHECK4-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
4135 // CHECK4-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 4
4136 // CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
4137 // CHECK4-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 4
4138 // CHECK4-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
4139 // CHECK4-NEXT:    store i8* null, i8** [[TMP39]], align 4
4140 // CHECK4-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4141 // CHECK4-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4142 // CHECK4-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
4143 // CHECK4-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
4144 // CHECK4-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
4145 // CHECK4-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4146 // CHECK4:       omp_offload.failed:
4147 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
4148 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4149 // CHECK4:       omp_offload.cont:
4150 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
4151 // CHECK4:       omp_if.else:
4152 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
4153 // CHECK4-NEXT:    br label [[OMP_IF_END]]
4154 // CHECK4:       omp_if.end:
4155 // CHECK4-NEXT:    [[TMP45:%.*]] = mul nsw i32 1, [[TMP1]]
4156 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP45]]
4157 // CHECK4-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
4158 // CHECK4-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
4159 // CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP46]] to i32
4160 // CHECK4-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
4161 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP47]]
4162 // CHECK4-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
4163 // CHECK4-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
4164 // CHECK4-NEXT:    ret i32 [[ADD3]]
4165 //
4166 //
4167 // CHECK4-LABEL: define {{[^@]+}}@_ZL7fstatici
4168 // CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
4169 // CHECK4-NEXT:  entry:
4170 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4171 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
4172 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
4173 // CHECK4-NEXT:    [[AAA:%.*]] = alloca i8, align 1
4174 // CHECK4-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
4175 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4176 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4177 // CHECK4-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
4178 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
4179 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
4180 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
4181 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4182 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
4183 // CHECK4-NEXT:    store i16 0, i16* [[AA]], align 2
4184 // CHECK4-NEXT:    store i8 0, i8* [[AAA]], align 1
4185 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
4186 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
4187 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
4188 // CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
4189 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4190 // CHECK4-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
4191 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4192 // CHECK4-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
4193 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
4194 // CHECK4-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
4195 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
4196 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
4197 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
4198 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4199 // CHECK4:       omp_if.then:
4200 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4201 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
4202 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
4203 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4204 // CHECK4-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
4205 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
4206 // CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4207 // CHECK4-NEXT:    store i8* null, i8** [[TMP11]], align 4
4208 // CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4209 // CHECK4-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
4210 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
4211 // CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4212 // CHECK4-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
4213 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
4214 // CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4215 // CHECK4-NEXT:    store i8* null, i8** [[TMP16]], align 4
4216 // CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4217 // CHECK4-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
4218 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
4219 // CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4220 // CHECK4-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
4221 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
4222 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4223 // CHECK4-NEXT:    store i8* null, i8** [[TMP21]], align 4
4224 // CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
4225 // CHECK4-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
4226 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
4227 // CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
4228 // CHECK4-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
4229 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
4230 // CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
4231 // CHECK4-NEXT:    store i8* null, i8** [[TMP26]], align 4
4232 // CHECK4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4233 // CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4234 // CHECK4-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
4235 // CHECK4-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
4236 // CHECK4-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4237 // CHECK4:       omp_offload.failed:
4238 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
4239 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4240 // CHECK4:       omp_offload.cont:
4241 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
4242 // CHECK4:       omp_if.else:
4243 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
4244 // CHECK4-NEXT:    br label [[OMP_IF_END]]
4245 // CHECK4:       omp_if.end:
4246 // CHECK4-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
4247 // CHECK4-NEXT:    ret i32 [[TMP31]]
4248 //
4249 //
4250 // CHECK4-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
4251 // CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0]] comdat {
4252 // CHECK4-NEXT:  entry:
4253 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4254 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
4255 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
4256 // CHECK4-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
4257 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4258 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4259 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
4260 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
4261 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
4262 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4263 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
4264 // CHECK4-NEXT:    store i16 0, i16* [[AA]], align 2
4265 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
4266 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
4267 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
4268 // CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
4269 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4270 // CHECK4-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
4271 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4272 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
4273 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
4274 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4275 // CHECK4:       omp_if.then:
4276 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4277 // CHECK4-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
4278 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
4279 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4280 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
4281 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
4282 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4283 // CHECK4-NEXT:    store i8* null, i8** [[TMP9]], align 4
4284 // CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4285 // CHECK4-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
4286 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
4287 // CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4288 // CHECK4-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
4289 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
4290 // CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4291 // CHECK4-NEXT:    store i8* null, i8** [[TMP14]], align 4
4292 // CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4293 // CHECK4-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
4294 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
4295 // CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4296 // CHECK4-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
4297 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
4298 // CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4299 // CHECK4-NEXT:    store i8* null, i8** [[TMP19]], align 4
4300 // CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4301 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4302 // CHECK4-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
4303 // CHECK4-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
4304 // CHECK4-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4305 // CHECK4:       omp_offload.failed:
4306 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
4307 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4308 // CHECK4:       omp_offload.cont:
4309 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
4310 // CHECK4:       omp_if.else:
4311 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
4312 // CHECK4-NEXT:    br label [[OMP_IF_END]]
4313 // CHECK4:       omp_if.end:
4314 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
4315 // CHECK4-NEXT:    ret i32 [[TMP24]]
4316 //
4317 //
4318 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
4319 // CHECK4-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
4320 // CHECK4-NEXT:  entry:
4321 // CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
4322 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
4323 // CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
4324 // CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
4325 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
4326 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 4
4327 // CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
4328 // CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
4329 // CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4330 // CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
4331 // CHECK4-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
4332 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
4333 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4334 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
4335 // CHECK4-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
4336 // CHECK4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4337 // CHECK4-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 4
4338 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4339 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_ADDR]], align 4
4340 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 4
4341 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4342 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP7]], align 4
4343 // CHECK4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
4344 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[TMP8]], align 4
4345 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
4346 // CHECK4-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 4
4347 // 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.5*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
4348 // CHECK4-NEXT:    ret void
4349 //
4350 //
4351 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..9
4352 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
4353 // CHECK4-NEXT:  entry:
4354 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4355 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4356 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 4
4357 // CHECK4-NEXT:    [[B:%.*]] = alloca i32, align 4
4358 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4359 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4360 // CHECK4-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 4
4361 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 4
4362 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
4363 // CHECK4-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
4364 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
4365 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4366 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
4367 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
4368 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
4369 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
4370 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
4371 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
4372 // CHECK4-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 4
4373 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
4374 // CHECK4-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
4375 // CHECK4-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
4376 // CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
4377 // CHECK4-NEXT:    store double [[ADD]], double* [[A]], align 4
4378 // CHECK4-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
4379 // CHECK4-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 4
4380 // CHECK4-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
4381 // CHECK4-NEXT:    store double [[INC]], double* [[A1]], align 4
4382 // CHECK4-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
4383 // CHECK4-NEXT:    [[TMP13:%.*]] = mul nsw i32 1, [[TMP8]]
4384 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i32 [[TMP13]]
4385 // CHECK4-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
4386 // CHECK4-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
4387 // CHECK4-NEXT:    ret void
4388 //
4389 //
4390 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
4391 // CHECK4-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
4392 // CHECK4-NEXT:  entry:
4393 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4394 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4395 // CHECK4-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
4396 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
4397 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 4
4398 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4399 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4400 // CHECK4-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
4401 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
4402 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4403 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
4404 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
4405 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4406 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
4407 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
4408 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4409 // CHECK4-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
4410 // CHECK4-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
4411 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4412 // CHECK4-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV1]], align 4
4413 // CHECK4-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
4414 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
4415 // CHECK4-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 4
4416 // 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.6*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
4417 // CHECK4-NEXT:    ret void
4418 //
4419 //
4420 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..11
4421 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
4422 // CHECK4-NEXT:  entry:
4423 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4424 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4425 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 4
4426 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
4427 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
4428 // CHECK4-NEXT:    [[AAA:%.*]] = alloca i8, align 1
4429 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4430 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4431 // CHECK4-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 4
4432 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 4
4433 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
4434 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4435 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
4436 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
4437 // CHECK4-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
4438 // CHECK4-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
4439 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
4440 // CHECK4-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
4441 // CHECK4-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
4442 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
4443 // CHECK4-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 4
4444 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
4445 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
4446 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
4447 // CHECK4-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
4448 // CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
4449 // CHECK4-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
4450 // CHECK4-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
4451 // CHECK4-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
4452 // CHECK4-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
4453 // CHECK4-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
4454 // CHECK4-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
4455 // CHECK4-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
4456 // CHECK4-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
4457 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i32 0, i32 2
4458 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
4459 // CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
4460 // CHECK4-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
4461 // CHECK4-NEXT:    ret void
4462 //
4463 //
4464 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
4465 // CHECK4-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
4466 // CHECK4-NEXT:  entry:
4467 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4468 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4469 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
4470 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 4
4471 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4472 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4473 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
4474 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4475 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
4476 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4477 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
4478 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
4479 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4480 // CHECK4-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
4481 // CHECK4-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
4482 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4483 // CHECK4-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 4
4484 // 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.7*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
4485 // CHECK4-NEXT:    ret void
4486 //
4487 //
4488 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..14
4489 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
4490 // CHECK4-NEXT:  entry:
4491 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4492 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4493 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 4
4494 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
4495 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
4496 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4497 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4498 // CHECK4-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 4
4499 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 4
4500 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
4501 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4502 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
4503 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
4504 // CHECK4-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
4505 // CHECK4-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
4506 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
4507 // CHECK4-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 4
4508 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
4509 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
4510 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
4511 // CHECK4-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
4512 // CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
4513 // CHECK4-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
4514 // CHECK4-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
4515 // CHECK4-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
4516 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i32 0, i32 2
4517 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
4518 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
4519 // CHECK4-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
4520 // CHECK4-NEXT:    ret void
4521 //
4522 //
4523 // CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4524 // CHECK4-SAME: () #[[ATTR6:[0-9]+]] {
4525 // CHECK4-NEXT:  entry:
4526 // CHECK4-NEXT:    call void @__tgt_register_requires(i64 1)
4527 // CHECK4-NEXT:    ret void
4528 //
4529 //
4530 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
4531 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
4532 // CHECK9-NEXT:  entry:
4533 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
4534 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
4535 // CHECK9-NEXT:    ret void
4536 //
4537 //
4538 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
4539 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4540 // CHECK9-NEXT:  entry:
4541 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4542 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4543 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
4544 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4545 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4546 // CHECK9-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
4547 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
4548 // CHECK9-NEXT:    ret void
4549 //
4550 //
4551 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
4552 // CHECK9-SAME: (i64 [[AA:%.*]]) #[[ATTR0]] {
4553 // CHECK9-NEXT:  entry:
4554 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
4555 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 2
4556 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
4557 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
4558 // CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4559 // CHECK9-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
4560 // CHECK9-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
4561 // 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_]])
4562 // CHECK9-NEXT:    ret void
4563 //
4564 //
4565 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
4566 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4567 // CHECK9-NEXT:  entry:
4568 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4569 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4570 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
4571 // CHECK9-NEXT:    [[AA:%.*]] = alloca i16, align 2
4572 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4573 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4574 // CHECK9-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
4575 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
4576 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
4577 // CHECK9-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
4578 // CHECK9-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
4579 // CHECK9-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
4580 // CHECK9-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
4581 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
4582 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
4583 // CHECK9-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
4584 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4585 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
4586 // CHECK9-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
4587 // CHECK9-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
4588 // CHECK9-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
4589 // CHECK9:       .cancel.exit:
4590 // CHECK9-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
4591 // CHECK9-NEXT:    br label [[DOTCANCEL_CONTINUE]]
4592 // CHECK9:       .cancel.continue:
4593 // CHECK9-NEXT:    ret void
4594 //
4595 //
4596 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
4597 // CHECK9-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
4598 // CHECK9-NEXT:  entry:
4599 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4600 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
4601 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
4602 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4603 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
4604 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4605 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
4606 // CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4607 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
4608 // CHECK9-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
4609 // CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4610 // CHECK9-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
4611 // CHECK9-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
4612 // 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_]])
4613 // CHECK9-NEXT:    ret void
4614 //
4615 //
4616 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
4617 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4618 // CHECK9-NEXT:  entry:
4619 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4620 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4621 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
4622 // CHECK9-NEXT:    [[A:%.*]] = alloca i32, align 4
4623 // CHECK9-NEXT:    [[AA:%.*]] = alloca i16, align 2
4624 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4625 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4626 // CHECK9-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
4627 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
4628 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
4629 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4630 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
4631 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
4632 // CHECK9-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
4633 // CHECK9-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
4634 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
4635 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
4636 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
4637 // CHECK9-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
4638 // CHECK9-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
4639 // CHECK9-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
4640 // CHECK9-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
4641 // CHECK9-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
4642 // CHECK9-NEXT:    ret void
4643 //
4644 //
4645 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
4646 // CHECK9-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR0]] {
4647 // CHECK9-NEXT:  entry:
4648 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4649 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
4650 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
4651 // CHECK9-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
4652 // CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
4653 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
4654 // CHECK9-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
4655 // CHECK9-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
4656 // CHECK9-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
4657 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
4658 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4659 // CHECK9-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
4660 // CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
4661 // CHECK9-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
4662 // CHECK9-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
4663 // CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
4664 // CHECK9-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
4665 // CHECK9-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
4666 // CHECK9-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
4667 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4668 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
4669 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
4670 // CHECK9-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
4671 // CHECK9-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
4672 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
4673 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
4674 // CHECK9-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
4675 // CHECK9-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
4676 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4677 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
4678 // CHECK9-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 8
4679 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4680 // CHECK9-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 8
4681 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4682 // CHECK9-NEXT:    store i64 [[TMP1]], i64* [[TMP11]], align 8
4683 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
4684 // CHECK9-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 8
4685 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
4686 // CHECK9-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 8
4687 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
4688 // CHECK9-NEXT:    store i64 [[TMP4]], i64* [[TMP14]], align 8
4689 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
4690 // CHECK9-NEXT:    store i64 [[TMP5]], i64* [[TMP15]], align 8
4691 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
4692 // CHECK9-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 8
4693 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
4694 // CHECK9-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 8
4695 // 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_]])
4696 // CHECK9-NEXT:    ret void
4697 //
4698 //
4699 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
4700 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4701 // CHECK9-NEXT:  entry:
4702 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4703 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4704 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
4705 // CHECK9-NEXT:    [[A:%.*]] = alloca i32, align 4
4706 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4707 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4708 // CHECK9-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
4709 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
4710 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
4711 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
4712 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
4713 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
4714 // CHECK9-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 8
4715 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
4716 // CHECK9-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
4717 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
4718 // CHECK9-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 8
4719 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 4
4720 // CHECK9-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 8
4721 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 5
4722 // CHECK9-NEXT:    [[TMP12:%.*]] = load i64, i64* [[TMP11]], align 8
4723 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 6
4724 // CHECK9-NEXT:    [[TMP14:%.*]] = load i64, i64* [[TMP13]], align 8
4725 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 7
4726 // CHECK9-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 8
4727 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 8
4728 // CHECK9-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 8
4729 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
4730 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
4731 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
4732 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i64 0, i64 2
4733 // CHECK9-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
4734 // CHECK9-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
4735 // CHECK9-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
4736 // CHECK9-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
4737 // CHECK9-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
4738 // CHECK9-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 3
4739 // CHECK9-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
4740 // CHECK9-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
4741 // CHECK9-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
4742 // CHECK9-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
4743 // CHECK9-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
4744 // CHECK9-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i64 0, i64 1
4745 // CHECK9-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i64 0, i64 2
4746 // CHECK9-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
4747 // CHECK9-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
4748 // CHECK9-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
4749 // CHECK9-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP14]]
4750 // CHECK9-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i64 [[TMP23]]
4751 // CHECK9-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i64 3
4752 // CHECK9-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
4753 // CHECK9-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
4754 // CHECK9-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
4755 // CHECK9-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
4756 // CHECK9-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8
4757 // CHECK9-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
4758 // CHECK9-NEXT:    store i64 [[ADD13]], i64* [[X]], align 8
4759 // CHECK9-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
4760 // CHECK9-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8
4761 // CHECK9-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
4762 // CHECK9-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
4763 // CHECK9-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
4764 // CHECK9-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 8
4765 // CHECK9-NEXT:    ret void
4766 //
4767 //
4768 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
4769 // CHECK9-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
4770 // CHECK9-NEXT:  entry:
4771 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4772 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
4773 // CHECK9-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
4774 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
4775 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 8
4776 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4777 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
4778 // CHECK9-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
4779 // CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
4780 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4781 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
4782 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
4783 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
4784 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4785 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
4786 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
4787 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4788 // CHECK9-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
4789 // CHECK9-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
4790 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4791 // CHECK9-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV2]], align 8
4792 // CHECK9-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
4793 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
4794 // CHECK9-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 8
4795 // 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_]])
4796 // CHECK9-NEXT:    ret void
4797 //
4798 //
4799 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
4800 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4801 // CHECK9-NEXT:  entry:
4802 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4803 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4804 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
4805 // CHECK9-NEXT:    [[A:%.*]] = alloca i32, align 4
4806 // CHECK9-NEXT:    [[AA:%.*]] = alloca i16, align 2
4807 // CHECK9-NEXT:    [[AAA:%.*]] = alloca i8, align 1
4808 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4809 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4810 // CHECK9-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
4811 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
4812 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
4813 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
4814 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
4815 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
4816 // CHECK9-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
4817 // CHECK9-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
4818 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
4819 // CHECK9-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
4820 // CHECK9-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
4821 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
4822 // CHECK9-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 8
4823 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
4824 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
4825 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
4826 // CHECK9-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
4827 // CHECK9-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
4828 // CHECK9-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
4829 // CHECK9-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
4830 // CHECK9-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
4831 // CHECK9-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
4832 // CHECK9-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
4833 // CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
4834 // CHECK9-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
4835 // CHECK9-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
4836 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i64 0, i64 2
4837 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
4838 // CHECK9-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
4839 // CHECK9-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
4840 // CHECK9-NEXT:    ret void
4841 //
4842 //
4843 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
4844 // CHECK9-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
4845 // CHECK9-NEXT:  entry:
4846 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
4847 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
4848 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
4849 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
4850 // CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
4851 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
4852 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
4853 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
4854 // CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
4855 // CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
4856 // CHECK9-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
4857 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
4858 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
4859 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
4860 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
4861 // CHECK9-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
4862 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4863 // CHECK9-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 8
4864 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4865 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[CONV]], align 8
4866 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
4867 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4868 // CHECK9-NEXT:    store i64 [[TMP1]], i64* [[TMP7]], align 8
4869 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
4870 // CHECK9-NEXT:    store i64 [[TMP2]], i64* [[TMP8]], align 8
4871 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
4872 // CHECK9-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 8
4873 // 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_]])
4874 // CHECK9-NEXT:    ret void
4875 //
4876 //
4877 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
4878 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4879 // CHECK9-NEXT:  entry:
4880 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4881 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4882 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
4883 // CHECK9-NEXT:    [[B:%.*]] = alloca i32, align 4
4884 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4885 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4886 // CHECK9-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
4887 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
4888 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
4889 // CHECK9-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
4890 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
4891 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
4892 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
4893 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
4894 // CHECK9-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
4895 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
4896 // CHECK9-NEXT:    [[TMP8:%.*]] = load i64, i64* [[TMP7]], align 8
4897 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
4898 // CHECK9-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 8
4899 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
4900 // CHECK9-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
4901 // CHECK9-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
4902 // CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
4903 // CHECK9-NEXT:    store double [[ADD]], double* [[A]], align 8
4904 // CHECK9-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
4905 // CHECK9-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 8
4906 // CHECK9-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
4907 // CHECK9-NEXT:    store double [[INC]], double* [[A1]], align 8
4908 // CHECK9-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
4909 // CHECK9-NEXT:    [[TMP13:%.*]] = mul nsw i64 1, [[TMP8]]
4910 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i64 [[TMP13]]
4911 // CHECK9-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
4912 // CHECK9-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
4913 // CHECK9-NEXT:    ret void
4914 //
4915 //
4916 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
4917 // CHECK9-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
4918 // CHECK9-NEXT:  entry:
4919 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4920 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
4921 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
4922 // CHECK9-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
4923 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4924 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
4925 // CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
4926 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4927 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
4928 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
4929 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4930 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
4931 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
4932 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4933 // CHECK9-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
4934 // CHECK9-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
4935 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4936 // CHECK9-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 8
4937 // 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.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
4938 // CHECK9-NEXT:    ret void
4939 //
4940 //
4941 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..6
4942 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4943 // CHECK9-NEXT:  entry:
4944 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4945 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4946 // CHECK9-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
4947 // CHECK9-NEXT:    [[A:%.*]] = alloca i32, align 4
4948 // CHECK9-NEXT:    [[AA:%.*]] = alloca i16, align 2
4949 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4950 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4951 // CHECK9-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
4952 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
4953 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
4954 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
4955 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
4956 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
4957 // CHECK9-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
4958 // CHECK9-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
4959 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
4960 // CHECK9-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 8
4961 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
4962 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
4963 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
4964 // CHECK9-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
4965 // CHECK9-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
4966 // CHECK9-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
4967 // CHECK9-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
4968 // CHECK9-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
4969 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i64 0, i64 2
4970 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
4971 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
4972 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
4973 // CHECK9-NEXT:    ret void
4974 //
4975 //
4976 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
4977 // CHECK10-SAME: () #[[ATTR0:[0-9]+]] {
4978 // CHECK10-NEXT:  entry:
4979 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
4980 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
4981 // CHECK10-NEXT:    ret void
4982 //
4983 //
4984 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined.
4985 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
4986 // CHECK10-NEXT:  entry:
4987 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4988 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4989 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
4990 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4991 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4992 // CHECK10-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
4993 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
4994 // CHECK10-NEXT:    ret void
4995 //
4996 //
4997 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
4998 // CHECK10-SAME: (i64 [[AA:%.*]]) #[[ATTR0]] {
4999 // CHECK10-NEXT:  entry:
5000 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
5001 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 2
5002 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
5003 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
5004 // CHECK10-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5005 // CHECK10-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
5006 // CHECK10-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
5007 // 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_]])
5008 // CHECK10-NEXT:    ret void
5009 //
5010 //
5011 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
5012 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5013 // CHECK10-NEXT:  entry:
5014 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5015 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5016 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
5017 // CHECK10-NEXT:    [[AA:%.*]] = alloca i16, align 2
5018 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5019 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5020 // CHECK10-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
5021 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
5022 // CHECK10-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
5023 // CHECK10-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
5024 // CHECK10-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
5025 // CHECK10-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
5026 // CHECK10-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
5027 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
5028 // CHECK10-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
5029 // CHECK10-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
5030 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5031 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
5032 // CHECK10-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
5033 // CHECK10-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
5034 // CHECK10-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
5035 // CHECK10:       .cancel.exit:
5036 // CHECK10-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
5037 // CHECK10-NEXT:    br label [[DOTCANCEL_CONTINUE]]
5038 // CHECK10:       .cancel.continue:
5039 // CHECK10-NEXT:    ret void
5040 //
5041 //
5042 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
5043 // CHECK10-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
5044 // CHECK10-NEXT:  entry:
5045 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
5046 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
5047 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
5048 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
5049 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
5050 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
5051 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
5052 // CHECK10-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5053 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
5054 // CHECK10-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
5055 // CHECK10-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5056 // CHECK10-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
5057 // CHECK10-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
5058 // 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_]])
5059 // CHECK10-NEXT:    ret void
5060 //
5061 //
5062 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
5063 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5064 // CHECK10-NEXT:  entry:
5065 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5066 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5067 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
5068 // CHECK10-NEXT:    [[A:%.*]] = alloca i32, align 4
5069 // CHECK10-NEXT:    [[AA:%.*]] = alloca i16, align 2
5070 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5071 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5072 // CHECK10-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
5073 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
5074 // CHECK10-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
5075 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5076 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5077 // CHECK10-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
5078 // CHECK10-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
5079 // CHECK10-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
5080 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
5081 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
5082 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5083 // CHECK10-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
5084 // CHECK10-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
5085 // CHECK10-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
5086 // CHECK10-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
5087 // CHECK10-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
5088 // CHECK10-NEXT:    ret void
5089 //
5090 //
5091 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
5092 // CHECK10-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR0]] {
5093 // CHECK10-NEXT:  entry:
5094 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
5095 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
5096 // CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
5097 // CHECK10-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
5098 // CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
5099 // CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
5100 // CHECK10-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
5101 // CHECK10-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
5102 // CHECK10-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
5103 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
5104 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
5105 // CHECK10-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
5106 // CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
5107 // CHECK10-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
5108 // CHECK10-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
5109 // CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
5110 // CHECK10-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
5111 // CHECK10-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
5112 // CHECK10-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
5113 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
5114 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
5115 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
5116 // CHECK10-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
5117 // CHECK10-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
5118 // CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
5119 // CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
5120 // CHECK10-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
5121 // CHECK10-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
5122 // CHECK10-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5123 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
5124 // CHECK10-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 8
5125 // CHECK10-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5126 // CHECK10-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 8
5127 // CHECK10-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
5128 // CHECK10-NEXT:    store i64 [[TMP1]], i64* [[TMP11]], align 8
5129 // CHECK10-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
5130 // CHECK10-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 8
5131 // CHECK10-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
5132 // CHECK10-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 8
5133 // CHECK10-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
5134 // CHECK10-NEXT:    store i64 [[TMP4]], i64* [[TMP14]], align 8
5135 // CHECK10-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
5136 // CHECK10-NEXT:    store i64 [[TMP5]], i64* [[TMP15]], align 8
5137 // CHECK10-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
5138 // CHECK10-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 8
5139 // CHECK10-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
5140 // CHECK10-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 8
5141 // 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_]])
5142 // CHECK10-NEXT:    ret void
5143 //
5144 //
5145 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
5146 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5147 // CHECK10-NEXT:  entry:
5148 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5149 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5150 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
5151 // CHECK10-NEXT:    [[A:%.*]] = alloca i32, align 4
5152 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5153 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5154 // CHECK10-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
5155 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
5156 // CHECK10-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
5157 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
5158 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5159 // CHECK10-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
5160 // CHECK10-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 8
5161 // CHECK10-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
5162 // CHECK10-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
5163 // CHECK10-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
5164 // CHECK10-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 8
5165 // CHECK10-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 4
5166 // CHECK10-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 8
5167 // CHECK10-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 5
5168 // CHECK10-NEXT:    [[TMP12:%.*]] = load i64, i64* [[TMP11]], align 8
5169 // CHECK10-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 6
5170 // CHECK10-NEXT:    [[TMP14:%.*]] = load i64, i64* [[TMP13]], align 8
5171 // CHECK10-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 7
5172 // CHECK10-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 8
5173 // CHECK10-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 8
5174 // CHECK10-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 8
5175 // CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
5176 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
5177 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5178 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i64 0, i64 2
5179 // CHECK10-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
5180 // CHECK10-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
5181 // CHECK10-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
5182 // CHECK10-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
5183 // CHECK10-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
5184 // CHECK10-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 3
5185 // CHECK10-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
5186 // CHECK10-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
5187 // CHECK10-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
5188 // CHECK10-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
5189 // CHECK10-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
5190 // CHECK10-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i64 0, i64 1
5191 // CHECK10-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i64 0, i64 2
5192 // CHECK10-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
5193 // CHECK10-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
5194 // CHECK10-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
5195 // CHECK10-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP14]]
5196 // CHECK10-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i64 [[TMP23]]
5197 // CHECK10-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i64 3
5198 // CHECK10-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
5199 // CHECK10-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
5200 // CHECK10-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
5201 // CHECK10-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
5202 // CHECK10-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8
5203 // CHECK10-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
5204 // CHECK10-NEXT:    store i64 [[ADD13]], i64* [[X]], align 8
5205 // CHECK10-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
5206 // CHECK10-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8
5207 // CHECK10-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
5208 // CHECK10-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
5209 // CHECK10-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
5210 // CHECK10-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 8
5211 // CHECK10-NEXT:    ret void
5212 //
5213 //
5214 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
5215 // CHECK10-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
5216 // CHECK10-NEXT:  entry:
5217 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
5218 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
5219 // CHECK10-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
5220 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
5221 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 8
5222 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
5223 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
5224 // CHECK10-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
5225 // CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
5226 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
5227 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
5228 // CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
5229 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
5230 // CHECK10-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5231 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
5232 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
5233 // CHECK10-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5234 // CHECK10-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
5235 // CHECK10-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
5236 // CHECK10-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
5237 // CHECK10-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV2]], align 8
5238 // CHECK10-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
5239 // CHECK10-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
5240 // CHECK10-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 8
5241 // 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_]])
5242 // CHECK10-NEXT:    ret void
5243 //
5244 //
5245 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..4
5246 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5247 // CHECK10-NEXT:  entry:
5248 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5249 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5250 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
5251 // CHECK10-NEXT:    [[A:%.*]] = alloca i32, align 4
5252 // CHECK10-NEXT:    [[AA:%.*]] = alloca i16, align 2
5253 // CHECK10-NEXT:    [[AAA:%.*]] = alloca i8, align 1
5254 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5255 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5256 // CHECK10-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
5257 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
5258 // CHECK10-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
5259 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
5260 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5261 // CHECK10-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
5262 // CHECK10-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
5263 // CHECK10-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
5264 // CHECK10-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
5265 // CHECK10-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
5266 // CHECK10-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
5267 // CHECK10-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
5268 // CHECK10-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 8
5269 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
5270 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
5271 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5272 // CHECK10-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
5273 // CHECK10-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
5274 // CHECK10-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
5275 // CHECK10-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
5276 // CHECK10-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
5277 // CHECK10-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
5278 // CHECK10-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
5279 // CHECK10-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
5280 // CHECK10-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
5281 // CHECK10-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
5282 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i64 0, i64 2
5283 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
5284 // CHECK10-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
5285 // CHECK10-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
5286 // CHECK10-NEXT:    ret void
5287 //
5288 //
5289 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
5290 // CHECK10-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
5291 // CHECK10-NEXT:  entry:
5292 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
5293 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
5294 // CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
5295 // CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
5296 // CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
5297 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
5298 // CHECK10-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
5299 // CHECK10-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
5300 // CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
5301 // CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
5302 // CHECK10-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
5303 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
5304 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
5305 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
5306 // CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
5307 // CHECK10-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
5308 // CHECK10-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5309 // CHECK10-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 8
5310 // CHECK10-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5311 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[CONV]], align 8
5312 // CHECK10-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
5313 // CHECK10-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
5314 // CHECK10-NEXT:    store i64 [[TMP1]], i64* [[TMP7]], align 8
5315 // CHECK10-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
5316 // CHECK10-NEXT:    store i64 [[TMP2]], i64* [[TMP8]], align 8
5317 // CHECK10-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
5318 // CHECK10-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 8
5319 // 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_]])
5320 // CHECK10-NEXT:    ret void
5321 //
5322 //
5323 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..5
5324 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5325 // CHECK10-NEXT:  entry:
5326 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5327 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5328 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
5329 // CHECK10-NEXT:    [[B:%.*]] = alloca i32, align 4
5330 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5331 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5332 // CHECK10-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
5333 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
5334 // CHECK10-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
5335 // CHECK10-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
5336 // CHECK10-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
5337 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
5338 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
5339 // CHECK10-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
5340 // CHECK10-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
5341 // CHECK10-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
5342 // CHECK10-NEXT:    [[TMP8:%.*]] = load i64, i64* [[TMP7]], align 8
5343 // CHECK10-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
5344 // CHECK10-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 8
5345 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
5346 // CHECK10-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
5347 // CHECK10-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
5348 // CHECK10-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
5349 // CHECK10-NEXT:    store double [[ADD]], double* [[A]], align 8
5350 // CHECK10-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
5351 // CHECK10-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 8
5352 // CHECK10-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
5353 // CHECK10-NEXT:    store double [[INC]], double* [[A1]], align 8
5354 // CHECK10-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
5355 // CHECK10-NEXT:    [[TMP13:%.*]] = mul nsw i64 1, [[TMP8]]
5356 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i64 [[TMP13]]
5357 // CHECK10-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
5358 // CHECK10-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
5359 // CHECK10-NEXT:    ret void
5360 //
5361 //
5362 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
5363 // CHECK10-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
5364 // CHECK10-NEXT:  entry:
5365 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
5366 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
5367 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
5368 // CHECK10-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
5369 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
5370 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
5371 // CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
5372 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
5373 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
5374 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
5375 // CHECK10-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5376 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
5377 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
5378 // CHECK10-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5379 // CHECK10-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
5380 // CHECK10-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
5381 // CHECK10-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
5382 // CHECK10-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 8
5383 // 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.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
5384 // CHECK10-NEXT:    ret void
5385 //
5386 //
5387 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..6
5388 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5389 // CHECK10-NEXT:  entry:
5390 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5391 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5392 // CHECK10-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
5393 // CHECK10-NEXT:    [[A:%.*]] = alloca i32, align 4
5394 // CHECK10-NEXT:    [[AA:%.*]] = alloca i16, align 2
5395 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5396 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5397 // CHECK10-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
5398 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
5399 // CHECK10-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
5400 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
5401 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5402 // CHECK10-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
5403 // CHECK10-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
5404 // CHECK10-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
5405 // CHECK10-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
5406 // CHECK10-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 8
5407 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
5408 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
5409 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5410 // CHECK10-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
5411 // CHECK10-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
5412 // CHECK10-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
5413 // CHECK10-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
5414 // CHECK10-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
5415 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i64 0, i64 2
5416 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
5417 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
5418 // CHECK10-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
5419 // CHECK10-NEXT:    ret void
5420 //
5421 //
5422 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
5423 // CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
5424 // CHECK11-NEXT:  entry:
5425 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
5426 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
5427 // CHECK11-NEXT:    ret void
5428 //
5429 //
5430 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
5431 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5432 // CHECK11-NEXT:  entry:
5433 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5434 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5435 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
5436 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5437 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5438 // CHECK11-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
5439 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
5440 // CHECK11-NEXT:    ret void
5441 //
5442 //
5443 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
5444 // CHECK11-SAME: (i32 [[AA:%.*]]) #[[ATTR0]] {
5445 // CHECK11-NEXT:  entry:
5446 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5447 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 2
5448 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5449 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5450 // CHECK11-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5451 // CHECK11-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
5452 // CHECK11-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
5453 // 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_]])
5454 // CHECK11-NEXT:    ret void
5455 //
5456 //
5457 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
5458 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5459 // CHECK11-NEXT:  entry:
5460 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5461 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5462 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4
5463 // CHECK11-NEXT:    [[AA:%.*]] = alloca i16, align 2
5464 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5465 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5466 // CHECK11-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4
5467 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4
5468 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
5469 // CHECK11-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
5470 // CHECK11-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
5471 // CHECK11-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
5472 // CHECK11-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
5473 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
5474 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
5475 // CHECK11-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
5476 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5477 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
5478 // CHECK11-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
5479 // CHECK11-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
5480 // CHECK11-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
5481 // CHECK11:       .cancel.exit:
5482 // CHECK11-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
5483 // CHECK11-NEXT:    br label [[DOTCANCEL_CONTINUE]]
5484 // CHECK11:       .cancel.continue:
5485 // CHECK11-NEXT:    ret void
5486 //
5487 //
5488 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
5489 // CHECK11-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
5490 // CHECK11-NEXT:  entry:
5491 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5492 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5493 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
5494 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5495 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5496 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5497 // CHECK11-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5498 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
5499 // CHECK11-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
5500 // CHECK11-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5501 // CHECK11-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
5502 // CHECK11-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
5503 // 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_]])
5504 // CHECK11-NEXT:    ret void
5505 //
5506 //
5507 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
5508 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5509 // CHECK11-NEXT:  entry:
5510 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5511 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5512 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
5513 // CHECK11-NEXT:    [[A:%.*]] = alloca i32, align 4
5514 // CHECK11-NEXT:    [[AA:%.*]] = alloca i16, align 2
5515 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5516 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5517 // CHECK11-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
5518 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
5519 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
5520 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5521 // CHECK11-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5522 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
5523 // CHECK11-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
5524 // CHECK11-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
5525 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
5526 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
5527 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5528 // CHECK11-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
5529 // CHECK11-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
5530 // CHECK11-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
5531 // CHECK11-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
5532 // CHECK11-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
5533 // CHECK11-NEXT:    ret void
5534 //
5535 //
5536 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
5537 // CHECK11-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR0]] {
5538 // CHECK11-NEXT:  entry:
5539 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5540 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
5541 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
5542 // CHECK11-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
5543 // CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
5544 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
5545 // CHECK11-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
5546 // CHECK11-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
5547 // CHECK11-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
5548 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 4
5549 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5550 // CHECK11-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
5551 // CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
5552 // CHECK11-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
5553 // CHECK11-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
5554 // CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
5555 // CHECK11-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
5556 // CHECK11-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
5557 // CHECK11-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
5558 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
5559 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
5560 // CHECK11-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
5561 // CHECK11-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
5562 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
5563 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
5564 // CHECK11-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
5565 // CHECK11-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
5566 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5567 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
5568 // CHECK11-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 4
5569 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5570 // CHECK11-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 4
5571 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
5572 // CHECK11-NEXT:    store i32 [[TMP1]], i32* [[TMP11]], align 4
5573 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
5574 // CHECK11-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 4
5575 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
5576 // CHECK11-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 4
5577 // CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
5578 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[TMP14]], align 4
5579 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
5580 // CHECK11-NEXT:    store i32 [[TMP5]], i32* [[TMP15]], align 4
5581 // CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
5582 // CHECK11-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 4
5583 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
5584 // CHECK11-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 4
5585 // 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_]])
5586 // CHECK11-NEXT:    ret void
5587 //
5588 //
5589 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
5590 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5591 // CHECK11-NEXT:  entry:
5592 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5593 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5594 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
5595 // CHECK11-NEXT:    [[A:%.*]] = alloca i32, align 4
5596 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5597 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5598 // CHECK11-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
5599 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
5600 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
5601 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5602 // CHECK11-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5603 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
5604 // CHECK11-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 4
5605 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
5606 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
5607 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
5608 // CHECK11-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 4
5609 // CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 4
5610 // CHECK11-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 4
5611 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 5
5612 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
5613 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 6
5614 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
5615 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 7
5616 // CHECK11-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 4
5617 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 8
5618 // CHECK11-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 4
5619 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
5620 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
5621 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5622 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i32 0, i32 2
5623 // CHECK11-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
5624 // CHECK11-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
5625 // CHECK11-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
5626 // CHECK11-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
5627 // CHECK11-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
5628 // CHECK11-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i32 3
5629 // CHECK11-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
5630 // CHECK11-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
5631 // CHECK11-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
5632 // CHECK11-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
5633 // CHECK11-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
5634 // CHECK11-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i32 0, i32 1
5635 // CHECK11-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i32 0, i32 2
5636 // CHECK11-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
5637 // CHECK11-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
5638 // CHECK11-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
5639 // CHECK11-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP14]]
5640 // CHECK11-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i32 [[TMP23]]
5641 // CHECK11-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i32 3
5642 // CHECK11-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
5643 // CHECK11-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
5644 // CHECK11-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
5645 // CHECK11-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
5646 // CHECK11-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4
5647 // CHECK11-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
5648 // CHECK11-NEXT:    store i64 [[ADD13]], i64* [[X]], align 4
5649 // CHECK11-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
5650 // CHECK11-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4
5651 // CHECK11-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
5652 // CHECK11-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
5653 // CHECK11-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
5654 // CHECK11-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 4
5655 // CHECK11-NEXT:    ret void
5656 //
5657 //
5658 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
5659 // CHECK11-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
5660 // CHECK11-NEXT:  entry:
5661 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5662 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5663 // CHECK11-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
5664 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
5665 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 4
5666 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5667 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5668 // CHECK11-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
5669 // CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
5670 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5671 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
5672 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
5673 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5674 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
5675 // CHECK11-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
5676 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5677 // CHECK11-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
5678 // CHECK11-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
5679 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
5680 // CHECK11-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV1]], align 4
5681 // CHECK11-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
5682 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
5683 // CHECK11-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 4
5684 // 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_]])
5685 // CHECK11-NEXT:    ret void
5686 //
5687 //
5688 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
5689 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5690 // CHECK11-NEXT:  entry:
5691 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5692 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5693 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
5694 // CHECK11-NEXT:    [[A:%.*]] = alloca i32, align 4
5695 // CHECK11-NEXT:    [[AA:%.*]] = alloca i16, align 2
5696 // CHECK11-NEXT:    [[AAA:%.*]] = alloca i8, align 1
5697 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5698 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5699 // CHECK11-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
5700 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
5701 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
5702 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5703 // CHECK11-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5704 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
5705 // CHECK11-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
5706 // CHECK11-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
5707 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
5708 // CHECK11-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
5709 // CHECK11-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
5710 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
5711 // CHECK11-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 4
5712 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
5713 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
5714 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5715 // CHECK11-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
5716 // CHECK11-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
5717 // CHECK11-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
5718 // CHECK11-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
5719 // CHECK11-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
5720 // CHECK11-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
5721 // CHECK11-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
5722 // CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
5723 // CHECK11-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
5724 // CHECK11-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
5725 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i32 0, i32 2
5726 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
5727 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
5728 // CHECK11-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
5729 // CHECK11-NEXT:    ret void
5730 //
5731 //
5732 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
5733 // CHECK11-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
5734 // CHECK11-NEXT:  entry:
5735 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
5736 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
5737 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
5738 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
5739 // CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
5740 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
5741 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
5742 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
5743 // CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
5744 // CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
5745 // CHECK11-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
5746 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
5747 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
5748 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
5749 // CHECK11-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
5750 // CHECK11-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5751 // CHECK11-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 4
5752 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5753 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_ADDR]], align 4
5754 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 4
5755 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
5756 // CHECK11-NEXT:    store i32 [[TMP1]], i32* [[TMP7]], align 4
5757 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
5758 // CHECK11-NEXT:    store i32 [[TMP2]], i32* [[TMP8]], align 4
5759 // CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
5760 // CHECK11-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 4
5761 // 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_]])
5762 // CHECK11-NEXT:    ret void
5763 //
5764 //
5765 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
5766 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5767 // CHECK11-NEXT:  entry:
5768 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5769 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5770 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
5771 // CHECK11-NEXT:    [[B:%.*]] = alloca i32, align 4
5772 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5773 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5774 // CHECK11-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
5775 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
5776 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
5777 // CHECK11-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
5778 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
5779 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
5780 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
5781 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
5782 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
5783 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
5784 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
5785 // CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
5786 // CHECK11-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 4
5787 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
5788 // CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
5789 // CHECK11-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
5790 // CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
5791 // CHECK11-NEXT:    store double [[ADD]], double* [[A]], align 4
5792 // CHECK11-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
5793 // CHECK11-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 4
5794 // CHECK11-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
5795 // CHECK11-NEXT:    store double [[INC]], double* [[A1]], align 4
5796 // CHECK11-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
5797 // CHECK11-NEXT:    [[TMP13:%.*]] = mul nsw i32 1, [[TMP8]]
5798 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i32 [[TMP13]]
5799 // CHECK11-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
5800 // CHECK11-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
5801 // CHECK11-NEXT:    ret void
5802 //
5803 //
5804 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
5805 // CHECK11-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
5806 // CHECK11-NEXT:  entry:
5807 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5808 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5809 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
5810 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 4
5811 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5812 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5813 // CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
5814 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5815 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
5816 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5817 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
5818 // CHECK11-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
5819 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5820 // CHECK11-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
5821 // CHECK11-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
5822 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
5823 // CHECK11-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 4
5824 // 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.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
5825 // CHECK11-NEXT:    ret void
5826 //
5827 //
5828 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..6
5829 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5830 // CHECK11-NEXT:  entry:
5831 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5832 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5833 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 4
5834 // CHECK11-NEXT:    [[A:%.*]] = alloca i32, align 4
5835 // CHECK11-NEXT:    [[AA:%.*]] = alloca i16, align 2
5836 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5837 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5838 // CHECK11-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 4
5839 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 4
5840 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
5841 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5842 // CHECK11-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5843 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
5844 // CHECK11-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
5845 // CHECK11-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
5846 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
5847 // CHECK11-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 4
5848 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
5849 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
5850 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5851 // CHECK11-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
5852 // CHECK11-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
5853 // CHECK11-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
5854 // CHECK11-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
5855 // CHECK11-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
5856 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i32 0, i32 2
5857 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
5858 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
5859 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
5860 // CHECK11-NEXT:    ret void
5861 //
5862 //
5863 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
5864 // CHECK12-SAME: () #[[ATTR0:[0-9]+]] {
5865 // CHECK12-NEXT:  entry:
5866 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
5867 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
5868 // CHECK12-NEXT:    ret void
5869 //
5870 //
5871 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined.
5872 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5873 // CHECK12-NEXT:  entry:
5874 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5875 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5876 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
5877 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5878 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5879 // CHECK12-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
5880 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
5881 // CHECK12-NEXT:    ret void
5882 //
5883 //
5884 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
5885 // CHECK12-SAME: (i32 [[AA:%.*]]) #[[ATTR0]] {
5886 // CHECK12-NEXT:  entry:
5887 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5888 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 2
5889 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5890 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5891 // CHECK12-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5892 // CHECK12-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
5893 // CHECK12-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
5894 // 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_]])
5895 // CHECK12-NEXT:    ret void
5896 //
5897 //
5898 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..1
5899 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5900 // CHECK12-NEXT:  entry:
5901 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5902 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5903 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4
5904 // CHECK12-NEXT:    [[AA:%.*]] = alloca i16, align 2
5905 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5906 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5907 // CHECK12-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4
5908 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4
5909 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
5910 // CHECK12-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
5911 // CHECK12-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
5912 // CHECK12-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
5913 // CHECK12-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
5914 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
5915 // CHECK12-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
5916 // CHECK12-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
5917 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5918 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
5919 // CHECK12-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
5920 // CHECK12-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
5921 // CHECK12-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
5922 // CHECK12:       .cancel.exit:
5923 // CHECK12-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
5924 // CHECK12-NEXT:    br label [[DOTCANCEL_CONTINUE]]
5925 // CHECK12:       .cancel.continue:
5926 // CHECK12-NEXT:    ret void
5927 //
5928 //
5929 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
5930 // CHECK12-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
5931 // CHECK12-NEXT:  entry:
5932 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5933 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5934 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
5935 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5936 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5937 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5938 // CHECK12-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
5939 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
5940 // CHECK12-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
5941 // CHECK12-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
5942 // CHECK12-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
5943 // CHECK12-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
5944 // 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_]])
5945 // CHECK12-NEXT:    ret void
5946 //
5947 //
5948 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..2
5949 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
5950 // CHECK12-NEXT:  entry:
5951 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5952 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5953 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
5954 // CHECK12-NEXT:    [[A:%.*]] = alloca i32, align 4
5955 // CHECK12-NEXT:    [[AA:%.*]] = alloca i16, align 2
5956 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5957 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5958 // CHECK12-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
5959 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
5960 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
5961 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5962 // CHECK12-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
5963 // CHECK12-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
5964 // CHECK12-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
5965 // CHECK12-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
5966 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
5967 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
5968 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5969 // CHECK12-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
5970 // CHECK12-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
5971 // CHECK12-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
5972 // CHECK12-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
5973 // CHECK12-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
5974 // CHECK12-NEXT:    ret void
5975 //
5976 //
5977 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
5978 // CHECK12-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR0]] {
5979 // CHECK12-NEXT:  entry:
5980 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5981 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
5982 // CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
5983 // CHECK12-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
5984 // CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
5985 // CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
5986 // CHECK12-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
5987 // CHECK12-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
5988 // CHECK12-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
5989 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 4
5990 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5991 // CHECK12-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
5992 // CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
5993 // CHECK12-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
5994 // CHECK12-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
5995 // CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
5996 // CHECK12-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
5997 // CHECK12-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
5998 // CHECK12-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
5999 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
6000 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6001 // CHECK12-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
6002 // CHECK12-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
6003 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
6004 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
6005 // CHECK12-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
6006 // CHECK12-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
6007 // CHECK12-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6008 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
6009 // CHECK12-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 4
6010 // CHECK12-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
6011 // CHECK12-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 4
6012 // CHECK12-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
6013 // CHECK12-NEXT:    store i32 [[TMP1]], i32* [[TMP11]], align 4
6014 // CHECK12-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
6015 // CHECK12-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 4
6016 // CHECK12-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
6017 // CHECK12-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 4
6018 // CHECK12-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
6019 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[TMP14]], align 4
6020 // CHECK12-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
6021 // CHECK12-NEXT:    store i32 [[TMP5]], i32* [[TMP15]], align 4
6022 // CHECK12-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
6023 // CHECK12-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 4
6024 // CHECK12-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
6025 // CHECK12-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 4
6026 // 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_]])
6027 // CHECK12-NEXT:    ret void
6028 //
6029 //
6030 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..3
6031 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
6032 // CHECK12-NEXT:  entry:
6033 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6034 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6035 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
6036 // CHECK12-NEXT:    [[A:%.*]] = alloca i32, align 4
6037 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6038 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6039 // CHECK12-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
6040 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
6041 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
6042 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6043 // CHECK12-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
6044 // CHECK12-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
6045 // CHECK12-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 4
6046 // CHECK12-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
6047 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
6048 // CHECK12-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
6049 // CHECK12-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 4
6050 // CHECK12-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 4
6051 // CHECK12-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 4
6052 // CHECK12-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 5
6053 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
6054 // CHECK12-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 6
6055 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
6056 // CHECK12-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 7
6057 // CHECK12-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 4
6058 // CHECK12-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 8
6059 // CHECK12-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 4
6060 // CHECK12-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
6061 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
6062 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
6063 // CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i32 0, i32 2
6064 // CHECK12-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
6065 // CHECK12-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
6066 // CHECK12-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
6067 // CHECK12-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
6068 // CHECK12-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
6069 // CHECK12-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i32 3
6070 // CHECK12-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
6071 // CHECK12-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
6072 // CHECK12-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
6073 // CHECK12-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
6074 // CHECK12-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
6075 // CHECK12-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i32 0, i32 1
6076 // CHECK12-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i32 0, i32 2
6077 // CHECK12-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
6078 // CHECK12-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
6079 // CHECK12-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
6080 // CHECK12-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP14]]
6081 // CHECK12-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i32 [[TMP23]]
6082 // CHECK12-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i32 3
6083 // CHECK12-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
6084 // CHECK12-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
6085 // CHECK12-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
6086 // CHECK12-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
6087 // CHECK12-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4
6088 // CHECK12-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
6089 // CHECK12-NEXT:    store i64 [[ADD13]], i64* [[X]], align 4
6090 // CHECK12-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
6091 // CHECK12-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4
6092 // CHECK12-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
6093 // CHECK12-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
6094 // CHECK12-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
6095 // CHECK12-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 4
6096 // CHECK12-NEXT:    ret void
6097 //
6098 //
6099 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
6100 // CHECK12-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
6101 // CHECK12-NEXT:  entry:
6102 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6103 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6104 // CHECK12-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
6105 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
6106 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 4
6107 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6108 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6109 // CHECK12-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
6110 // CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
6111 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6112 // CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
6113 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
6114 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6115 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
6116 // CHECK12-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
6117 // CHECK12-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
6118 // CHECK12-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
6119 // CHECK12-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
6120 // CHECK12-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
6121 // CHECK12-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV1]], align 4
6122 // CHECK12-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
6123 // CHECK12-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
6124 // CHECK12-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 4
6125 // 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_]])
6126 // CHECK12-NEXT:    ret void
6127 //
6128 //
6129 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..4
6130 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
6131 // CHECK12-NEXT:  entry:
6132 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6133 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6134 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
6135 // CHECK12-NEXT:    [[A:%.*]] = alloca i32, align 4
6136 // CHECK12-NEXT:    [[AA:%.*]] = alloca i16, align 2
6137 // CHECK12-NEXT:    [[AAA:%.*]] = alloca i8, align 1
6138 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6139 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6140 // CHECK12-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
6141 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
6142 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
6143 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6144 // CHECK12-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
6145 // CHECK12-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
6146 // CHECK12-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
6147 // CHECK12-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
6148 // CHECK12-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
6149 // CHECK12-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
6150 // CHECK12-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
6151 // CHECK12-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
6152 // CHECK12-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 4
6153 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
6154 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
6155 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
6156 // CHECK12-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
6157 // CHECK12-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
6158 // CHECK12-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
6159 // CHECK12-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
6160 // CHECK12-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
6161 // CHECK12-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
6162 // CHECK12-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
6163 // CHECK12-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
6164 // CHECK12-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
6165 // CHECK12-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
6166 // CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i32 0, i32 2
6167 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
6168 // CHECK12-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
6169 // CHECK12-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
6170 // CHECK12-NEXT:    ret void
6171 //
6172 //
6173 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
6174 // CHECK12-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
6175 // CHECK12-NEXT:  entry:
6176 // CHECK12-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
6177 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
6178 // CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
6179 // CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
6180 // CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
6181 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
6182 // CHECK12-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
6183 // CHECK12-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
6184 // CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6185 // CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
6186 // CHECK12-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
6187 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
6188 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6189 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
6190 // CHECK12-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
6191 // CHECK12-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6192 // CHECK12-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 4
6193 // CHECK12-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
6194 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_ADDR]], align 4
6195 // CHECK12-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 4
6196 // CHECK12-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
6197 // CHECK12-NEXT:    store i32 [[TMP1]], i32* [[TMP7]], align 4
6198 // CHECK12-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
6199 // CHECK12-NEXT:    store i32 [[TMP2]], i32* [[TMP8]], align 4
6200 // CHECK12-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
6201 // CHECK12-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 4
6202 // 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_]])
6203 // CHECK12-NEXT:    ret void
6204 //
6205 //
6206 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..5
6207 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
6208 // CHECK12-NEXT:  entry:
6209 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6210 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6211 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
6212 // CHECK12-NEXT:    [[B:%.*]] = alloca i32, align 4
6213 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6214 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6215 // CHECK12-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
6216 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
6217 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
6218 // CHECK12-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
6219 // CHECK12-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
6220 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
6221 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
6222 // CHECK12-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
6223 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
6224 // CHECK12-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
6225 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
6226 // CHECK12-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
6227 // CHECK12-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 4
6228 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
6229 // CHECK12-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
6230 // CHECK12-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
6231 // CHECK12-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
6232 // CHECK12-NEXT:    store double [[ADD]], double* [[A]], align 4
6233 // CHECK12-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
6234 // CHECK12-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 4
6235 // CHECK12-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
6236 // CHECK12-NEXT:    store double [[INC]], double* [[A1]], align 4
6237 // CHECK12-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
6238 // CHECK12-NEXT:    [[TMP13:%.*]] = mul nsw i32 1, [[TMP8]]
6239 // CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i32 [[TMP13]]
6240 // CHECK12-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
6241 // CHECK12-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
6242 // CHECK12-NEXT:    ret void
6243 //
6244 //
6245 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
6246 // CHECK12-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
6247 // CHECK12-NEXT:  entry:
6248 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6249 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6250 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
6251 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 4
6252 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6253 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6254 // CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
6255 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6256 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
6257 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6258 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
6259 // CHECK12-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
6260 // CHECK12-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
6261 // CHECK12-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
6262 // CHECK12-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
6263 // CHECK12-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
6264 // CHECK12-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 4
6265 // 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.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
6266 // CHECK12-NEXT:    ret void
6267 //
6268 //
6269 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..6
6270 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
6271 // CHECK12-NEXT:  entry:
6272 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6273 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6274 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 4
6275 // CHECK12-NEXT:    [[A:%.*]] = alloca i32, align 4
6276 // CHECK12-NEXT:    [[AA:%.*]] = alloca i16, align 2
6277 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6278 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6279 // CHECK12-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 4
6280 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 4
6281 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
6282 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6283 // CHECK12-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
6284 // CHECK12-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
6285 // CHECK12-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
6286 // CHECK12-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
6287 // CHECK12-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
6288 // CHECK12-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 4
6289 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
6290 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
6291 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
6292 // CHECK12-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
6293 // CHECK12-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
6294 // CHECK12-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
6295 // CHECK12-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
6296 // CHECK12-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
6297 // CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i32 0, i32 2
6298 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
6299 // CHECK12-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
6300 // CHECK12-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
6301 // CHECK12-NEXT:    ret void
6302 //
6303 //
6304 // CHECK17-LABEL: define {{[^@]+}}@_Z3fooi
6305 // CHECK17-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
6306 // CHECK17-NEXT:  entry:
6307 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6308 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
6309 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
6310 // CHECK17-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
6311 // CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
6312 // CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
6313 // CHECK17-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
6314 // CHECK17-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
6315 // CHECK17-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
6316 // CHECK17-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
6317 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
6318 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
6319 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
6320 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
6321 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
6322 // CHECK17-NEXT:    [[A_CASTED3:%.*]] = alloca i64, align 8
6323 // CHECK17-NEXT:    [[AA_CASTED5:%.*]] = alloca i64, align 8
6324 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
6325 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
6326 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
6327 // CHECK17-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
6328 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [9 x i8*], align 8
6329 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [9 x i8*], align 8
6330 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [9 x i8*], align 8
6331 // CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 8
6332 // CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
6333 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6334 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
6335 // CHECK17-NEXT:    store i16 0, i16* [[AA]], align 2
6336 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
6337 // CHECK17-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
6338 // CHECK17-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
6339 // CHECK17-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
6340 // CHECK17-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
6341 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
6342 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
6343 // CHECK17-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
6344 // CHECK17-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
6345 // CHECK17-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
6346 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
6347 // CHECK17-NEXT:    [[TMP7:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
6348 // CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.kmp_task_t_with_privates*
6349 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP8]], i32 0, i32 0
6350 // CHECK17-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP7]])
6351 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A]], align 4
6352 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
6353 // CHECK17-NEXT:    store i32 [[TMP11]], i32* [[CONV]], align 4
6354 // CHECK17-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED]], align 8
6355 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i64 [[TMP12]]) #[[ATTR3:[0-9]+]]
6356 // CHECK17-NEXT:    [[TMP13:%.*]] = load i16, i16* [[AA]], align 2
6357 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
6358 // CHECK17-NEXT:    store i16 [[TMP13]], i16* [[CONV2]], align 2
6359 // CHECK17-NEXT:    [[TMP14:%.*]] = load i64, i64* [[AA_CASTED]], align 8
6360 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6361 // CHECK17-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
6362 // CHECK17-NEXT:    store i64 [[TMP14]], i64* [[TMP16]], align 8
6363 // CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6364 // CHECK17-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
6365 // CHECK17-NEXT:    store i64 [[TMP14]], i64* [[TMP18]], align 8
6366 // CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
6367 // CHECK17-NEXT:    store i8* null, i8** [[TMP19]], align 8
6368 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6369 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6370 // CHECK17-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
6371 // CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
6372 // CHECK17-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6373 // CHECK17:       omp_offload.failed:
6374 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i64 [[TMP14]]) #[[ATTR3]]
6375 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
6376 // CHECK17:       omp_offload.cont:
6377 // CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
6378 // CHECK17-NEXT:    [[CONV4:%.*]] = bitcast i64* [[A_CASTED3]] to i32*
6379 // CHECK17-NEXT:    store i32 [[TMP24]], i32* [[CONV4]], align 4
6380 // CHECK17-NEXT:    [[TMP25:%.*]] = load i64, i64* [[A_CASTED3]], align 8
6381 // CHECK17-NEXT:    [[TMP26:%.*]] = load i16, i16* [[AA]], align 2
6382 // CHECK17-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED5]] to i16*
6383 // CHECK17-NEXT:    store i16 [[TMP26]], i16* [[CONV6]], align 2
6384 // CHECK17-NEXT:    [[TMP27:%.*]] = load i64, i64* [[AA_CASTED5]], align 8
6385 // CHECK17-NEXT:    [[TMP28:%.*]] = load i32, i32* [[N_ADDR]], align 4
6386 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP28]], 10
6387 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6388 // CHECK17:       omp_if.then:
6389 // CHECK17-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
6390 // CHECK17-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i64*
6391 // CHECK17-NEXT:    store i64 [[TMP25]], i64* [[TMP30]], align 8
6392 // CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
6393 // CHECK17-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
6394 // CHECK17-NEXT:    store i64 [[TMP25]], i64* [[TMP32]], align 8
6395 // CHECK17-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
6396 // CHECK17-NEXT:    store i8* null, i8** [[TMP33]], align 8
6397 // CHECK17-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
6398 // CHECK17-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i64*
6399 // CHECK17-NEXT:    store i64 [[TMP27]], i64* [[TMP35]], align 8
6400 // CHECK17-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
6401 // CHECK17-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
6402 // CHECK17-NEXT:    store i64 [[TMP27]], i64* [[TMP37]], align 8
6403 // CHECK17-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
6404 // CHECK17-NEXT:    store i8* null, i8** [[TMP38]], align 8
6405 // CHECK17-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
6406 // CHECK17-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
6407 // CHECK17-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
6408 // CHECK17-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
6409 // CHECK17-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED10:%.*]], label [[OMP_OFFLOAD_CONT11:%.*]]
6410 // CHECK17:       omp_offload.failed10:
6411 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR3]]
6412 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT11]]
6413 // CHECK17:       omp_offload.cont11:
6414 // CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
6415 // CHECK17:       omp_if.else:
6416 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR3]]
6417 // CHECK17-NEXT:    br label [[OMP_IF_END]]
6418 // CHECK17:       omp_if.end:
6419 // CHECK17-NEXT:    [[TMP43:%.*]] = load i32, i32* [[A]], align 4
6420 // CHECK17-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
6421 // CHECK17-NEXT:    store i32 [[TMP43]], i32* [[CONV13]], align 4
6422 // CHECK17-NEXT:    [[TMP44:%.*]] = load i64, i64* [[A_CASTED12]], align 8
6423 // CHECK17-NEXT:    [[TMP45:%.*]] = load i32, i32* [[N_ADDR]], align 4
6424 // CHECK17-NEXT:    [[CMP14:%.*]] = icmp sgt i32 [[TMP45]], 20
6425 // CHECK17-NEXT:    br i1 [[CMP14]], label [[OMP_IF_THEN15:%.*]], label [[OMP_IF_ELSE21:%.*]]
6426 // CHECK17:       omp_if.then15:
6427 // CHECK17-NEXT:    [[TMP46:%.*]] = mul nuw i64 [[TMP2]], 4
6428 // CHECK17-NEXT:    [[TMP47:%.*]] = mul nuw i64 5, [[TMP5]]
6429 // CHECK17-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], 8
6430 // CHECK17-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
6431 // CHECK17-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i64*
6432 // CHECK17-NEXT:    store i64 [[TMP44]], i64* [[TMP50]], align 8
6433 // CHECK17-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
6434 // CHECK17-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i64*
6435 // CHECK17-NEXT:    store i64 [[TMP44]], i64* [[TMP52]], align 8
6436 // CHECK17-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
6437 // CHECK17-NEXT:    store i64 4, i64* [[TMP53]], align 8
6438 // CHECK17-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
6439 // CHECK17-NEXT:    store i8* null, i8** [[TMP54]], align 8
6440 // CHECK17-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
6441 // CHECK17-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
6442 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 8
6443 // CHECK17-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
6444 // CHECK17-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
6445 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 8
6446 // CHECK17-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
6447 // CHECK17-NEXT:    store i64 40, i64* [[TMP59]], align 8
6448 // CHECK17-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
6449 // CHECK17-NEXT:    store i8* null, i8** [[TMP60]], align 8
6450 // CHECK17-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 2
6451 // CHECK17-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
6452 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP62]], align 8
6453 // CHECK17-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 2
6454 // CHECK17-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
6455 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP64]], align 8
6456 // CHECK17-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
6457 // CHECK17-NEXT:    store i64 8, i64* [[TMP65]], align 8
6458 // CHECK17-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 2
6459 // CHECK17-NEXT:    store i8* null, i8** [[TMP66]], align 8
6460 // CHECK17-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 3
6461 // CHECK17-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
6462 // CHECK17-NEXT:    store float* [[VLA]], float** [[TMP68]], align 8
6463 // CHECK17-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 3
6464 // CHECK17-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
6465 // CHECK17-NEXT:    store float* [[VLA]], float** [[TMP70]], align 8
6466 // CHECK17-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
6467 // CHECK17-NEXT:    store i64 [[TMP46]], i64* [[TMP71]], align 8
6468 // CHECK17-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 3
6469 // CHECK17-NEXT:    store i8* null, i8** [[TMP72]], align 8
6470 // CHECK17-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 4
6471 // CHECK17-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
6472 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 8
6473 // CHECK17-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 4
6474 // CHECK17-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
6475 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 8
6476 // CHECK17-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
6477 // CHECK17-NEXT:    store i64 400, i64* [[TMP77]], align 8
6478 // CHECK17-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 4
6479 // CHECK17-NEXT:    store i8* null, i8** [[TMP78]], align 8
6480 // CHECK17-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 5
6481 // CHECK17-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i64*
6482 // CHECK17-NEXT:    store i64 5, i64* [[TMP80]], align 8
6483 // CHECK17-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 5
6484 // CHECK17-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
6485 // CHECK17-NEXT:    store i64 5, i64* [[TMP82]], align 8
6486 // CHECK17-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
6487 // CHECK17-NEXT:    store i64 8, i64* [[TMP83]], align 8
6488 // CHECK17-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 5
6489 // CHECK17-NEXT:    store i8* null, i8** [[TMP84]], align 8
6490 // CHECK17-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 6
6491 // CHECK17-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
6492 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP86]], align 8
6493 // CHECK17-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 6
6494 // CHECK17-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
6495 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP88]], align 8
6496 // CHECK17-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
6497 // CHECK17-NEXT:    store i64 8, i64* [[TMP89]], align 8
6498 // CHECK17-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 6
6499 // CHECK17-NEXT:    store i8* null, i8** [[TMP90]], align 8
6500 // CHECK17-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 7
6501 // CHECK17-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
6502 // CHECK17-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 8
6503 // CHECK17-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 7
6504 // CHECK17-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
6505 // CHECK17-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 8
6506 // CHECK17-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
6507 // CHECK17-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 8
6508 // CHECK17-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 7
6509 // CHECK17-NEXT:    store i8* null, i8** [[TMP96]], align 8
6510 // CHECK17-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 8
6511 // CHECK17-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
6512 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 8
6513 // CHECK17-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 8
6514 // CHECK17-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
6515 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 8
6516 // CHECK17-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
6517 // CHECK17-NEXT:    store i64 16, i64* [[TMP101]], align 8
6518 // CHECK17-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 8
6519 // CHECK17-NEXT:    store i8* null, i8** [[TMP102]], align 8
6520 // CHECK17-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
6521 // CHECK17-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
6522 // CHECK17-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
6523 // CHECK17-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
6524 // CHECK17-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
6525 // CHECK17-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
6526 // CHECK17:       omp_offload.failed19:
6527 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
6528 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
6529 // CHECK17:       omp_offload.cont20:
6530 // CHECK17-NEXT:    br label [[OMP_IF_END22:%.*]]
6531 // CHECK17:       omp_if.else21:
6532 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
6533 // CHECK17-NEXT:    br label [[OMP_IF_END22]]
6534 // CHECK17:       omp_if.end22:
6535 // CHECK17-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
6536 // CHECK17-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
6537 // CHECK17-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
6538 // CHECK17-NEXT:    ret i32 [[TMP108]]
6539 //
6540 //
6541 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
6542 // CHECK17-SAME: () #[[ATTR2:[0-9]+]] {
6543 // CHECK17-NEXT:  entry:
6544 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
6545 // 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_]])
6546 // CHECK17-NEXT:    ret void
6547 //
6548 //
6549 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
6550 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
6551 // CHECK17-NEXT:  entry:
6552 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6553 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6554 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
6555 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6556 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6557 // CHECK17-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
6558 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
6559 // CHECK17-NEXT:    ret void
6560 //
6561 //
6562 // CHECK17-LABEL: define {{[^@]+}}@.omp_task_entry.
6563 // CHECK17-SAME: (i32 signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
6564 // CHECK17-NEXT:  entry:
6565 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
6566 // CHECK17-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
6567 // CHECK17-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
6568 // CHECK17-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
6569 // CHECK17-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
6570 // CHECK17-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 8
6571 // CHECK17-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
6572 // CHECK17-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
6573 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
6574 // CHECK17-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
6575 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
6576 // CHECK17-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
6577 // CHECK17-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
6578 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
6579 // CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
6580 // CHECK17-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
6581 // CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
6582 // CHECK17-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
6583 // CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
6584 // CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
6585 // CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
6586 // CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
6587 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !21
6588 // CHECK17-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !21
6589 // CHECK17-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
6590 // CHECK17-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
6591 // CHECK17-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !21
6592 // CHECK17-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
6593 // CHECK17-NEXT:    [[TMP10:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
6594 // CHECK17-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0) #[[ATTR3]]
6595 // CHECK17-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
6596 // CHECK17-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
6597 // CHECK17:       omp_offload.failed.i:
6598 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR3]]
6599 // CHECK17-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
6600 // CHECK17:       .omp_outlined..1.exit:
6601 // CHECK17-NEXT:    ret i32 0
6602 //
6603 //
6604 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
6605 // CHECK17-SAME: (i64 [[A:%.*]]) #[[ATTR2]] {
6606 // CHECK17-NEXT:  entry:
6607 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6608 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
6609 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6610 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6611 // CHECK17-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6612 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
6613 // CHECK17-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
6614 // 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..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
6615 // CHECK17-NEXT:    ret void
6616 //
6617 //
6618 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..2
6619 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
6620 // CHECK17-NEXT:  entry:
6621 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6622 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6623 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
6624 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
6625 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6626 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6627 // CHECK17-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
6628 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
6629 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
6630 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6631 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
6632 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
6633 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], 1
6634 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
6635 // CHECK17-NEXT:    ret void
6636 //
6637 //
6638 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
6639 // CHECK17-SAME: (i64 [[AA:%.*]]) #[[ATTR2]] {
6640 // CHECK17-NEXT:  entry:
6641 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6642 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 2
6643 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6644 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6645 // CHECK17-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6646 // CHECK17-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
6647 // CHECK17-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
6648 // 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..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
6649 // CHECK17-NEXT:    ret void
6650 //
6651 //
6652 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..3
6653 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
6654 // CHECK17-NEXT:  entry:
6655 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6656 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6657 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
6658 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
6659 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6660 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6661 // CHECK17-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
6662 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
6663 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
6664 // CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
6665 // CHECK17-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
6666 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
6667 // CHECK17-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
6668 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
6669 // CHECK17-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
6670 // CHECK17-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
6671 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6672 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
6673 // CHECK17-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
6674 // CHECK17-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
6675 // CHECK17-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
6676 // CHECK17:       .cancel.exit:
6677 // CHECK17-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
6678 // CHECK17-NEXT:    br label [[DOTCANCEL_CONTINUE]]
6679 // CHECK17:       .cancel.continue:
6680 // CHECK17-NEXT:    ret void
6681 //
6682 //
6683 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
6684 // CHECK17-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR2]] {
6685 // CHECK17-NEXT:  entry:
6686 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6687 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6688 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 4
6689 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6690 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6691 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6692 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6693 // CHECK17-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6694 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
6695 // CHECK17-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
6696 // CHECK17-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
6697 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
6698 // CHECK17-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
6699 // 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..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
6700 // CHECK17-NEXT:    ret void
6701 //
6702 //
6703 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..4
6704 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
6705 // CHECK17-NEXT:  entry:
6706 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6707 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6708 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
6709 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
6710 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
6711 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6712 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6713 // CHECK17-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
6714 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
6715 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
6716 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6717 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
6718 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
6719 // CHECK17-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
6720 // CHECK17-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
6721 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
6722 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
6723 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
6724 // CHECK17-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
6725 // CHECK17-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
6726 // CHECK17-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
6727 // CHECK17-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
6728 // CHECK17-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
6729 // CHECK17-NEXT:    ret void
6730 //
6731 //
6732 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
6733 // CHECK17-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR2]] {
6734 // CHECK17-NEXT:  entry:
6735 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6736 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
6737 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
6738 // CHECK17-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
6739 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
6740 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
6741 // CHECK17-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
6742 // CHECK17-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
6743 // CHECK17-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
6744 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
6745 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6746 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
6747 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
6748 // CHECK17-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
6749 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
6750 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
6751 // CHECK17-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
6752 // CHECK17-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
6753 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
6754 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6755 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
6756 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
6757 // CHECK17-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
6758 // CHECK17-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
6759 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
6760 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
6761 // CHECK17-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
6762 // CHECK17-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
6763 // CHECK17-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6764 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
6765 // CHECK17-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 8
6766 // CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
6767 // CHECK17-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 8
6768 // CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
6769 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP11]], align 8
6770 // CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
6771 // CHECK17-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 8
6772 // CHECK17-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
6773 // CHECK17-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 8
6774 // CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
6775 // CHECK17-NEXT:    store i64 [[TMP4]], i64* [[TMP14]], align 8
6776 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
6777 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP15]], align 8
6778 // CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
6779 // CHECK17-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 8
6780 // CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
6781 // CHECK17-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 8
6782 // 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..7 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
6783 // CHECK17-NEXT:    ret void
6784 //
6785 //
6786 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..7
6787 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
6788 // CHECK17-NEXT:  entry:
6789 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6790 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6791 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
6792 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
6793 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6794 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6795 // CHECK17-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
6796 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
6797 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
6798 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
6799 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
6800 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
6801 // CHECK17-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 8
6802 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
6803 // CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
6804 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
6805 // CHECK17-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 8
6806 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
6807 // CHECK17-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 8
6808 // CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 5
6809 // CHECK17-NEXT:    [[TMP12:%.*]] = load i64, i64* [[TMP11]], align 8
6810 // CHECK17-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 6
6811 // CHECK17-NEXT:    [[TMP14:%.*]] = load i64, i64* [[TMP13]], align 8
6812 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 7
6813 // CHECK17-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 8
6814 // CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 8
6815 // CHECK17-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 8
6816 // CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
6817 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
6818 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
6819 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i64 0, i64 2
6820 // CHECK17-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
6821 // CHECK17-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
6822 // CHECK17-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
6823 // CHECK17-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
6824 // CHECK17-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
6825 // CHECK17-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 3
6826 // CHECK17-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
6827 // CHECK17-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
6828 // CHECK17-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
6829 // CHECK17-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
6830 // CHECK17-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
6831 // CHECK17-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i64 0, i64 1
6832 // CHECK17-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i64 0, i64 2
6833 // CHECK17-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
6834 // CHECK17-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
6835 // CHECK17-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
6836 // CHECK17-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP14]]
6837 // CHECK17-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i64 [[TMP23]]
6838 // CHECK17-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i64 3
6839 // CHECK17-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
6840 // CHECK17-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
6841 // CHECK17-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
6842 // CHECK17-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
6843 // CHECK17-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8
6844 // CHECK17-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
6845 // CHECK17-NEXT:    store i64 [[ADD13]], i64* [[X]], align 8
6846 // CHECK17-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
6847 // CHECK17-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8
6848 // CHECK17-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
6849 // CHECK17-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
6850 // CHECK17-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
6851 // CHECK17-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 8
6852 // CHECK17-NEXT:    ret void
6853 //
6854 //
6855 // CHECK17-LABEL: define {{[^@]+}}@_Z3bari
6856 // CHECK17-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
6857 // CHECK17-NEXT:  entry:
6858 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6859 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
6860 // CHECK17-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
6861 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6862 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
6863 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
6864 // CHECK17-NEXT:    [[CALL:%.*]] = call signext i32 @_Z3fooi(i32 signext [[TMP0]])
6865 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
6866 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
6867 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
6868 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
6869 // CHECK17-NEXT:    [[CALL1:%.*]] = call signext i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], i32 signext [[TMP2]])
6870 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
6871 // CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
6872 // CHECK17-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
6873 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
6874 // CHECK17-NEXT:    [[CALL3:%.*]] = call signext i32 @_ZL7fstatici(i32 signext [[TMP4]])
6875 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
6876 // CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
6877 // CHECK17-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
6878 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
6879 // CHECK17-NEXT:    [[CALL5:%.*]] = call signext i32 @_Z9ftemplateIiET_i(i32 signext [[TMP6]])
6880 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
6881 // CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
6882 // CHECK17-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
6883 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
6884 // CHECK17-NEXT:    ret i32 [[TMP8]]
6885 //
6886 //
6887 // CHECK17-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
6888 // CHECK17-SAME: (%struct.S1* nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
6889 // CHECK17-NEXT:  entry:
6890 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
6891 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6892 // CHECK17-NEXT:    [[B:%.*]] = alloca i32, align 4
6893 // CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
6894 // CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
6895 // CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
6896 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
6897 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
6898 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
6899 // CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
6900 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
6901 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6902 // CHECK17-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
6903 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
6904 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
6905 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
6906 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
6907 // CHECK17-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
6908 // CHECK17-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
6909 // CHECK17-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
6910 // CHECK17-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
6911 // CHECK17-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
6912 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
6913 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
6914 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
6915 // CHECK17-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
6916 // CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
6917 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
6918 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
6919 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6920 // CHECK17:       omp_if.then:
6921 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
6922 // CHECK17-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
6923 // CHECK17-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
6924 // CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6925 // CHECK17-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
6926 // CHECK17-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 8
6927 // CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6928 // CHECK17-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
6929 // CHECK17-NEXT:    store double* [[A]], double** [[TMP13]], align 8
6930 // CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
6931 // CHECK17-NEXT:    store i64 8, i64* [[TMP14]], align 8
6932 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
6933 // CHECK17-NEXT:    store i8* null, i8** [[TMP15]], align 8
6934 // CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
6935 // CHECK17-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
6936 // CHECK17-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
6937 // CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
6938 // CHECK17-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
6939 // CHECK17-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
6940 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
6941 // CHECK17-NEXT:    store i64 4, i64* [[TMP20]], align 8
6942 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
6943 // CHECK17-NEXT:    store i8* null, i8** [[TMP21]], align 8
6944 // CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
6945 // CHECK17-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
6946 // CHECK17-NEXT:    store i64 2, i64* [[TMP23]], align 8
6947 // CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
6948 // CHECK17-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
6949 // CHECK17-NEXT:    store i64 2, i64* [[TMP25]], align 8
6950 // CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
6951 // CHECK17-NEXT:    store i64 8, i64* [[TMP26]], align 8
6952 // CHECK17-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
6953 // CHECK17-NEXT:    store i8* null, i8** [[TMP27]], align 8
6954 // CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
6955 // CHECK17-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
6956 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
6957 // CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
6958 // CHECK17-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
6959 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP31]], align 8
6960 // CHECK17-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
6961 // CHECK17-NEXT:    store i64 8, i64* [[TMP32]], align 8
6962 // CHECK17-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
6963 // CHECK17-NEXT:    store i8* null, i8** [[TMP33]], align 8
6964 // CHECK17-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
6965 // CHECK17-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
6966 // CHECK17-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 8
6967 // CHECK17-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
6968 // CHECK17-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
6969 // CHECK17-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 8
6970 // CHECK17-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
6971 // CHECK17-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 8
6972 // CHECK17-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
6973 // CHECK17-NEXT:    store i8* null, i8** [[TMP39]], align 8
6974 // CHECK17-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6975 // CHECK17-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6976 // CHECK17-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
6977 // CHECK17-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
6978 // CHECK17-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
6979 // CHECK17-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6980 // CHECK17:       omp_offload.failed:
6981 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
6982 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
6983 // CHECK17:       omp_offload.cont:
6984 // CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
6985 // CHECK17:       omp_if.else:
6986 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
6987 // CHECK17-NEXT:    br label [[OMP_IF_END]]
6988 // CHECK17:       omp_if.end:
6989 // CHECK17-NEXT:    [[TMP45:%.*]] = mul nsw i64 1, [[TMP2]]
6990 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP45]]
6991 // CHECK17-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
6992 // CHECK17-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
6993 // CHECK17-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP46]] to i32
6994 // CHECK17-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
6995 // CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP47]]
6996 // CHECK17-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
6997 // CHECK17-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
6998 // CHECK17-NEXT:    ret i32 [[ADD4]]
6999 //
7000 //
7001 // CHECK17-LABEL: define {{[^@]+}}@_ZL7fstatici
7002 // CHECK17-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
7003 // CHECK17-NEXT:  entry:
7004 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7005 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
7006 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
7007 // CHECK17-NEXT:    [[AAA:%.*]] = alloca i8, align 1
7008 // CHECK17-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
7009 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
7010 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
7011 // CHECK17-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
7012 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
7013 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
7014 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
7015 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7016 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
7017 // CHECK17-NEXT:    store i16 0, i16* [[AA]], align 2
7018 // CHECK17-NEXT:    store i8 0, i8* [[AAA]], align 1
7019 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
7020 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
7021 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
7022 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
7023 // CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
7024 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
7025 // CHECK17-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
7026 // CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
7027 // CHECK17-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
7028 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
7029 // CHECK17-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
7030 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
7031 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
7032 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
7033 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
7034 // CHECK17:       omp_if.then:
7035 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7036 // CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
7037 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
7038 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7039 // CHECK17-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
7040 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
7041 // CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
7042 // CHECK17-NEXT:    store i8* null, i8** [[TMP11]], align 8
7043 // CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
7044 // CHECK17-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
7045 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
7046 // CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
7047 // CHECK17-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
7048 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
7049 // CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
7050 // CHECK17-NEXT:    store i8* null, i8** [[TMP16]], align 8
7051 // CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
7052 // CHECK17-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
7053 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
7054 // CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
7055 // CHECK17-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
7056 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
7057 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
7058 // CHECK17-NEXT:    store i8* null, i8** [[TMP21]], align 8
7059 // CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
7060 // CHECK17-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
7061 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
7062 // CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
7063 // CHECK17-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
7064 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
7065 // CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
7066 // CHECK17-NEXT:    store i8* null, i8** [[TMP26]], align 8
7067 // CHECK17-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7068 // CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7069 // CHECK17-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
7070 // CHECK17-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
7071 // CHECK17-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
7072 // CHECK17:       omp_offload.failed:
7073 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
7074 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
7075 // CHECK17:       omp_offload.cont:
7076 // CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
7077 // CHECK17:       omp_if.else:
7078 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
7079 // CHECK17-NEXT:    br label [[OMP_IF_END]]
7080 // CHECK17:       omp_if.end:
7081 // CHECK17-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
7082 // CHECK17-NEXT:    ret i32 [[TMP31]]
7083 //
7084 //
7085 // CHECK17-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
7086 // CHECK17-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] comdat {
7087 // CHECK17-NEXT:  entry:
7088 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7089 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
7090 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
7091 // CHECK17-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
7092 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
7093 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
7094 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
7095 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
7096 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
7097 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7098 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
7099 // CHECK17-NEXT:    store i16 0, i16* [[AA]], align 2
7100 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
7101 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
7102 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
7103 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
7104 // CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
7105 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
7106 // CHECK17-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
7107 // CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
7108 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
7109 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
7110 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
7111 // CHECK17:       omp_if.then:
7112 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7113 // CHECK17-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
7114 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
7115 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7116 // CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
7117 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
7118 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
7119 // CHECK17-NEXT:    store i8* null, i8** [[TMP9]], align 8
7120 // CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
7121 // CHECK17-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
7122 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
7123 // CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
7124 // CHECK17-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
7125 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
7126 // CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
7127 // CHECK17-NEXT:    store i8* null, i8** [[TMP14]], align 8
7128 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
7129 // CHECK17-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
7130 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
7131 // CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
7132 // CHECK17-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
7133 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
7134 // CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
7135 // CHECK17-NEXT:    store i8* null, i8** [[TMP19]], align 8
7136 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7137 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7138 // CHECK17-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
7139 // CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
7140 // CHECK17-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
7141 // CHECK17:       omp_offload.failed:
7142 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
7143 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
7144 // CHECK17:       omp_offload.cont:
7145 // CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
7146 // CHECK17:       omp_if.else:
7147 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
7148 // CHECK17-NEXT:    br label [[OMP_IF_END]]
7149 // CHECK17:       omp_if.end:
7150 // CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
7151 // CHECK17-NEXT:    ret i32 [[TMP24]]
7152 //
7153 //
7154 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
7155 // CHECK17-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
7156 // CHECK17-NEXT:  entry:
7157 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
7158 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
7159 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
7160 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
7161 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
7162 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
7163 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
7164 // CHECK17-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
7165 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
7166 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
7167 // CHECK17-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
7168 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
7169 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
7170 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
7171 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
7172 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
7173 // CHECK17-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7174 // CHECK17-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 8
7175 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
7176 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[CONV]], align 8
7177 // CHECK17-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
7178 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
7179 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP7]], align 8
7180 // CHECK17-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
7181 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP8]], align 8
7182 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
7183 // CHECK17-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 8
7184 // 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.5*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
7185 // CHECK17-NEXT:    ret void
7186 //
7187 //
7188 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..9
7189 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
7190 // CHECK17-NEXT:  entry:
7191 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7192 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7193 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
7194 // CHECK17-NEXT:    [[B:%.*]] = alloca i32, align 4
7195 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7196 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7197 // CHECK17-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
7198 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
7199 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
7200 // CHECK17-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
7201 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
7202 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
7203 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
7204 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
7205 // CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
7206 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
7207 // CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[TMP7]], align 8
7208 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
7209 // CHECK17-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 8
7210 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
7211 // CHECK17-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
7212 // CHECK17-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
7213 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
7214 // CHECK17-NEXT:    store double [[ADD]], double* [[A]], align 8
7215 // CHECK17-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
7216 // CHECK17-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 8
7217 // CHECK17-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
7218 // CHECK17-NEXT:    store double [[INC]], double* [[A1]], align 8
7219 // CHECK17-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
7220 // CHECK17-NEXT:    [[TMP13:%.*]] = mul nsw i64 1, [[TMP8]]
7221 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i64 [[TMP13]]
7222 // CHECK17-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
7223 // CHECK17-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
7224 // CHECK17-NEXT:    ret void
7225 //
7226 //
7227 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
7228 // CHECK17-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
7229 // CHECK17-NEXT:  entry:
7230 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7231 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7232 // CHECK17-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
7233 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
7234 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 8
7235 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7236 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7237 // CHECK17-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
7238 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
7239 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7240 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7241 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
7242 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
7243 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7244 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
7245 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
7246 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
7247 // CHECK17-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
7248 // CHECK17-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
7249 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
7250 // CHECK17-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV2]], align 8
7251 // CHECK17-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
7252 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
7253 // CHECK17-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 8
7254 // 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.6*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
7255 // CHECK17-NEXT:    ret void
7256 //
7257 //
7258 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..11
7259 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
7260 // CHECK17-NEXT:  entry:
7261 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7262 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7263 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 8
7264 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
7265 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
7266 // CHECK17-NEXT:    [[AAA:%.*]] = alloca i8, align 1
7267 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7268 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7269 // CHECK17-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 8
7270 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 8
7271 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
7272 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
7273 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
7274 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
7275 // CHECK17-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
7276 // CHECK17-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
7277 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
7278 // CHECK17-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
7279 // CHECK17-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
7280 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
7281 // CHECK17-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 8
7282 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
7283 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
7284 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
7285 // CHECK17-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
7286 // CHECK17-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
7287 // CHECK17-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
7288 // CHECK17-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
7289 // CHECK17-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
7290 // CHECK17-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
7291 // CHECK17-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
7292 // CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
7293 // CHECK17-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
7294 // CHECK17-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
7295 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i64 0, i64 2
7296 // CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
7297 // CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
7298 // CHECK17-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
7299 // CHECK17-NEXT:    ret void
7300 //
7301 //
7302 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
7303 // CHECK17-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
7304 // CHECK17-NEXT:  entry:
7305 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7306 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7307 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
7308 // CHECK17-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 8
7309 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7310 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7311 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
7312 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7313 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7314 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
7315 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7316 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
7317 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
7318 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
7319 // CHECK17-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
7320 // CHECK17-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
7321 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
7322 // CHECK17-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 8
7323 // 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.7*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
7324 // CHECK17-NEXT:    ret void
7325 //
7326 //
7327 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..14
7328 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
7329 // CHECK17-NEXT:  entry:
7330 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7331 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7332 // CHECK17-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 8
7333 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
7334 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
7335 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7336 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7337 // CHECK17-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 8
7338 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 8
7339 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
7340 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
7341 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
7342 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
7343 // CHECK17-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
7344 // CHECK17-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
7345 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
7346 // CHECK17-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 8
7347 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
7348 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
7349 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
7350 // CHECK17-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
7351 // CHECK17-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
7352 // CHECK17-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
7353 // CHECK17-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
7354 // CHECK17-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
7355 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i64 0, i64 2
7356 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
7357 // CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
7358 // CHECK17-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
7359 // CHECK17-NEXT:    ret void
7360 //
7361 //
7362 // CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
7363 // CHECK17-SAME: () #[[ATTR6:[0-9]+]] {
7364 // CHECK17-NEXT:  entry:
7365 // CHECK17-NEXT:    call void @__tgt_register_requires(i64 1)
7366 // CHECK17-NEXT:    ret void
7367 //
7368 //
7369 // CHECK18-LABEL: define {{[^@]+}}@_Z3fooi
7370 // CHECK18-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
7371 // CHECK18-NEXT:  entry:
7372 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7373 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
7374 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
7375 // CHECK18-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
7376 // CHECK18-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
7377 // CHECK18-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
7378 // CHECK18-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
7379 // CHECK18-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
7380 // CHECK18-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
7381 // CHECK18-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
7382 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
7383 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
7384 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
7385 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
7386 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
7387 // CHECK18-NEXT:    [[A_CASTED3:%.*]] = alloca i64, align 8
7388 // CHECK18-NEXT:    [[AA_CASTED5:%.*]] = alloca i64, align 8
7389 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
7390 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
7391 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
7392 // CHECK18-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
7393 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [9 x i8*], align 8
7394 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [9 x i8*], align 8
7395 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [9 x i8*], align 8
7396 // CHECK18-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 8
7397 // CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
7398 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7399 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
7400 // CHECK18-NEXT:    store i16 0, i16* [[AA]], align 2
7401 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
7402 // CHECK18-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
7403 // CHECK18-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
7404 // CHECK18-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
7405 // CHECK18-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
7406 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
7407 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
7408 // CHECK18-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
7409 // CHECK18-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
7410 // CHECK18-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
7411 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
7412 // CHECK18-NEXT:    [[TMP7:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
7413 // CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.kmp_task_t_with_privates*
7414 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP8]], i32 0, i32 0
7415 // CHECK18-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP7]])
7416 // CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A]], align 4
7417 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
7418 // CHECK18-NEXT:    store i32 [[TMP11]], i32* [[CONV]], align 4
7419 // CHECK18-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED]], align 8
7420 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i64 [[TMP12]]) #[[ATTR3:[0-9]+]]
7421 // CHECK18-NEXT:    [[TMP13:%.*]] = load i16, i16* [[AA]], align 2
7422 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
7423 // CHECK18-NEXT:    store i16 [[TMP13]], i16* [[CONV2]], align 2
7424 // CHECK18-NEXT:    [[TMP14:%.*]] = load i64, i64* [[AA_CASTED]], align 8
7425 // CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7426 // CHECK18-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
7427 // CHECK18-NEXT:    store i64 [[TMP14]], i64* [[TMP16]], align 8
7428 // CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7429 // CHECK18-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
7430 // CHECK18-NEXT:    store i64 [[TMP14]], i64* [[TMP18]], align 8
7431 // CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
7432 // CHECK18-NEXT:    store i8* null, i8** [[TMP19]], align 8
7433 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7434 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7435 // CHECK18-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
7436 // CHECK18-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
7437 // CHECK18-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
7438 // CHECK18:       omp_offload.failed:
7439 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i64 [[TMP14]]) #[[ATTR3]]
7440 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
7441 // CHECK18:       omp_offload.cont:
7442 // CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
7443 // CHECK18-NEXT:    [[CONV4:%.*]] = bitcast i64* [[A_CASTED3]] to i32*
7444 // CHECK18-NEXT:    store i32 [[TMP24]], i32* [[CONV4]], align 4
7445 // CHECK18-NEXT:    [[TMP25:%.*]] = load i64, i64* [[A_CASTED3]], align 8
7446 // CHECK18-NEXT:    [[TMP26:%.*]] = load i16, i16* [[AA]], align 2
7447 // CHECK18-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED5]] to i16*
7448 // CHECK18-NEXT:    store i16 [[TMP26]], i16* [[CONV6]], align 2
7449 // CHECK18-NEXT:    [[TMP27:%.*]] = load i64, i64* [[AA_CASTED5]], align 8
7450 // CHECK18-NEXT:    [[TMP28:%.*]] = load i32, i32* [[N_ADDR]], align 4
7451 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP28]], 10
7452 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
7453 // CHECK18:       omp_if.then:
7454 // CHECK18-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
7455 // CHECK18-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i64*
7456 // CHECK18-NEXT:    store i64 [[TMP25]], i64* [[TMP30]], align 8
7457 // CHECK18-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
7458 // CHECK18-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
7459 // CHECK18-NEXT:    store i64 [[TMP25]], i64* [[TMP32]], align 8
7460 // CHECK18-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
7461 // CHECK18-NEXT:    store i8* null, i8** [[TMP33]], align 8
7462 // CHECK18-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
7463 // CHECK18-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i64*
7464 // CHECK18-NEXT:    store i64 [[TMP27]], i64* [[TMP35]], align 8
7465 // CHECK18-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
7466 // CHECK18-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
7467 // CHECK18-NEXT:    store i64 [[TMP27]], i64* [[TMP37]], align 8
7468 // CHECK18-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
7469 // CHECK18-NEXT:    store i8* null, i8** [[TMP38]], align 8
7470 // CHECK18-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
7471 // CHECK18-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
7472 // CHECK18-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
7473 // CHECK18-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
7474 // CHECK18-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED10:%.*]], label [[OMP_OFFLOAD_CONT11:%.*]]
7475 // CHECK18:       omp_offload.failed10:
7476 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR3]]
7477 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT11]]
7478 // CHECK18:       omp_offload.cont11:
7479 // CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
7480 // CHECK18:       omp_if.else:
7481 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR3]]
7482 // CHECK18-NEXT:    br label [[OMP_IF_END]]
7483 // CHECK18:       omp_if.end:
7484 // CHECK18-NEXT:    [[TMP43:%.*]] = load i32, i32* [[A]], align 4
7485 // CHECK18-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
7486 // CHECK18-NEXT:    store i32 [[TMP43]], i32* [[CONV13]], align 4
7487 // CHECK18-NEXT:    [[TMP44:%.*]] = load i64, i64* [[A_CASTED12]], align 8
7488 // CHECK18-NEXT:    [[TMP45:%.*]] = load i32, i32* [[N_ADDR]], align 4
7489 // CHECK18-NEXT:    [[CMP14:%.*]] = icmp sgt i32 [[TMP45]], 20
7490 // CHECK18-NEXT:    br i1 [[CMP14]], label [[OMP_IF_THEN15:%.*]], label [[OMP_IF_ELSE21:%.*]]
7491 // CHECK18:       omp_if.then15:
7492 // CHECK18-NEXT:    [[TMP46:%.*]] = mul nuw i64 [[TMP2]], 4
7493 // CHECK18-NEXT:    [[TMP47:%.*]] = mul nuw i64 5, [[TMP5]]
7494 // CHECK18-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], 8
7495 // CHECK18-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
7496 // CHECK18-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i64*
7497 // CHECK18-NEXT:    store i64 [[TMP44]], i64* [[TMP50]], align 8
7498 // CHECK18-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
7499 // CHECK18-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i64*
7500 // CHECK18-NEXT:    store i64 [[TMP44]], i64* [[TMP52]], align 8
7501 // CHECK18-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
7502 // CHECK18-NEXT:    store i64 4, i64* [[TMP53]], align 8
7503 // CHECK18-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
7504 // CHECK18-NEXT:    store i8* null, i8** [[TMP54]], align 8
7505 // CHECK18-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
7506 // CHECK18-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
7507 // CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 8
7508 // CHECK18-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
7509 // CHECK18-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
7510 // CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 8
7511 // CHECK18-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
7512 // CHECK18-NEXT:    store i64 40, i64* [[TMP59]], align 8
7513 // CHECK18-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
7514 // CHECK18-NEXT:    store i8* null, i8** [[TMP60]], align 8
7515 // CHECK18-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 2
7516 // CHECK18-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
7517 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP62]], align 8
7518 // CHECK18-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 2
7519 // CHECK18-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
7520 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP64]], align 8
7521 // CHECK18-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
7522 // CHECK18-NEXT:    store i64 8, i64* [[TMP65]], align 8
7523 // CHECK18-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 2
7524 // CHECK18-NEXT:    store i8* null, i8** [[TMP66]], align 8
7525 // CHECK18-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 3
7526 // CHECK18-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
7527 // CHECK18-NEXT:    store float* [[VLA]], float** [[TMP68]], align 8
7528 // CHECK18-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 3
7529 // CHECK18-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
7530 // CHECK18-NEXT:    store float* [[VLA]], float** [[TMP70]], align 8
7531 // CHECK18-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
7532 // CHECK18-NEXT:    store i64 [[TMP46]], i64* [[TMP71]], align 8
7533 // CHECK18-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 3
7534 // CHECK18-NEXT:    store i8* null, i8** [[TMP72]], align 8
7535 // CHECK18-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 4
7536 // CHECK18-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
7537 // CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 8
7538 // CHECK18-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 4
7539 // CHECK18-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
7540 // CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 8
7541 // CHECK18-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
7542 // CHECK18-NEXT:    store i64 400, i64* [[TMP77]], align 8
7543 // CHECK18-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 4
7544 // CHECK18-NEXT:    store i8* null, i8** [[TMP78]], align 8
7545 // CHECK18-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 5
7546 // CHECK18-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i64*
7547 // CHECK18-NEXT:    store i64 5, i64* [[TMP80]], align 8
7548 // CHECK18-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 5
7549 // CHECK18-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
7550 // CHECK18-NEXT:    store i64 5, i64* [[TMP82]], align 8
7551 // CHECK18-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
7552 // CHECK18-NEXT:    store i64 8, i64* [[TMP83]], align 8
7553 // CHECK18-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 5
7554 // CHECK18-NEXT:    store i8* null, i8** [[TMP84]], align 8
7555 // CHECK18-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 6
7556 // CHECK18-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
7557 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP86]], align 8
7558 // CHECK18-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 6
7559 // CHECK18-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
7560 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP88]], align 8
7561 // CHECK18-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
7562 // CHECK18-NEXT:    store i64 8, i64* [[TMP89]], align 8
7563 // CHECK18-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 6
7564 // CHECK18-NEXT:    store i8* null, i8** [[TMP90]], align 8
7565 // CHECK18-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 7
7566 // CHECK18-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
7567 // CHECK18-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 8
7568 // CHECK18-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 7
7569 // CHECK18-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
7570 // CHECK18-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 8
7571 // CHECK18-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
7572 // CHECK18-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 8
7573 // CHECK18-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 7
7574 // CHECK18-NEXT:    store i8* null, i8** [[TMP96]], align 8
7575 // CHECK18-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 8
7576 // CHECK18-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
7577 // CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 8
7578 // CHECK18-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 8
7579 // CHECK18-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
7580 // CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 8
7581 // CHECK18-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
7582 // CHECK18-NEXT:    store i64 16, i64* [[TMP101]], align 8
7583 // CHECK18-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 8
7584 // CHECK18-NEXT:    store i8* null, i8** [[TMP102]], align 8
7585 // CHECK18-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
7586 // CHECK18-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
7587 // CHECK18-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
7588 // CHECK18-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
7589 // CHECK18-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
7590 // CHECK18-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
7591 // CHECK18:       omp_offload.failed19:
7592 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
7593 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
7594 // CHECK18:       omp_offload.cont20:
7595 // CHECK18-NEXT:    br label [[OMP_IF_END22:%.*]]
7596 // CHECK18:       omp_if.else21:
7597 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i64 [[TMP44]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
7598 // CHECK18-NEXT:    br label [[OMP_IF_END22]]
7599 // CHECK18:       omp_if.end22:
7600 // CHECK18-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
7601 // CHECK18-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
7602 // CHECK18-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
7603 // CHECK18-NEXT:    ret i32 [[TMP108]]
7604 //
7605 //
7606 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
7607 // CHECK18-SAME: () #[[ATTR2:[0-9]+]] {
7608 // CHECK18-NEXT:  entry:
7609 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
7610 // 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_]])
7611 // CHECK18-NEXT:    ret void
7612 //
7613 //
7614 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined.
7615 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
7616 // CHECK18-NEXT:  entry:
7617 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7618 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7619 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
7620 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7621 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7622 // CHECK18-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
7623 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
7624 // CHECK18-NEXT:    ret void
7625 //
7626 //
7627 // CHECK18-LABEL: define {{[^@]+}}@.omp_task_entry.
7628 // CHECK18-SAME: (i32 signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
7629 // CHECK18-NEXT:  entry:
7630 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
7631 // CHECK18-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
7632 // CHECK18-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
7633 // CHECK18-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
7634 // CHECK18-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
7635 // CHECK18-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 8
7636 // CHECK18-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
7637 // CHECK18-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
7638 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
7639 // CHECK18-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
7640 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
7641 // CHECK18-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
7642 // CHECK18-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
7643 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
7644 // CHECK18-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
7645 // CHECK18-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
7646 // CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
7647 // CHECK18-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
7648 // CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
7649 // CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
7650 // CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
7651 // CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
7652 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !21
7653 // CHECK18-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !21
7654 // CHECK18-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
7655 // CHECK18-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
7656 // CHECK18-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !21
7657 // CHECK18-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
7658 // CHECK18-NEXT:    [[TMP10:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
7659 // CHECK18-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0) #[[ATTR3]]
7660 // CHECK18-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
7661 // CHECK18-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
7662 // CHECK18:       omp_offload.failed.i:
7663 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR3]]
7664 // CHECK18-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
7665 // CHECK18:       .omp_outlined..1.exit:
7666 // CHECK18-NEXT:    ret i32 0
7667 //
7668 //
7669 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
7670 // CHECK18-SAME: (i64 [[A:%.*]]) #[[ATTR2]] {
7671 // CHECK18-NEXT:  entry:
7672 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7673 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
7674 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7675 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7676 // CHECK18-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7677 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
7678 // CHECK18-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
7679 // 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..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
7680 // CHECK18-NEXT:    ret void
7681 //
7682 //
7683 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..2
7684 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
7685 // CHECK18-NEXT:  entry:
7686 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7687 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7688 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
7689 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
7690 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7691 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7692 // CHECK18-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
7693 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
7694 // CHECK18-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
7695 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
7696 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
7697 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
7698 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], 1
7699 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
7700 // CHECK18-NEXT:    ret void
7701 //
7702 //
7703 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
7704 // CHECK18-SAME: (i64 [[AA:%.*]]) #[[ATTR2]] {
7705 // CHECK18-NEXT:  entry:
7706 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7707 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 2
7708 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7709 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7710 // CHECK18-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7711 // CHECK18-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
7712 // CHECK18-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
7713 // 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..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
7714 // CHECK18-NEXT:    ret void
7715 //
7716 //
7717 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..3
7718 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
7719 // CHECK18-NEXT:  entry:
7720 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7721 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7722 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
7723 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
7724 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7725 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7726 // CHECK18-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
7727 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
7728 // CHECK18-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
7729 // CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
7730 // CHECK18-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
7731 // CHECK18-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
7732 // CHECK18-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
7733 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
7734 // CHECK18-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
7735 // CHECK18-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
7736 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7737 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
7738 // CHECK18-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
7739 // CHECK18-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
7740 // CHECK18-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
7741 // CHECK18:       .cancel.exit:
7742 // CHECK18-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
7743 // CHECK18-NEXT:    br label [[DOTCANCEL_CONTINUE]]
7744 // CHECK18:       .cancel.continue:
7745 // CHECK18-NEXT:    ret void
7746 //
7747 //
7748 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
7749 // CHECK18-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR2]] {
7750 // CHECK18-NEXT:  entry:
7751 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7752 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7753 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 4
7754 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7755 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7756 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7757 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7758 // CHECK18-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7759 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
7760 // CHECK18-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
7761 // CHECK18-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
7762 // CHECK18-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
7763 // CHECK18-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
7764 // 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..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
7765 // CHECK18-NEXT:    ret void
7766 //
7767 //
7768 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..4
7769 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
7770 // CHECK18-NEXT:  entry:
7771 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7772 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7773 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
7774 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
7775 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
7776 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7777 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7778 // CHECK18-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
7779 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
7780 // CHECK18-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
7781 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
7782 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
7783 // CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
7784 // CHECK18-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
7785 // CHECK18-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
7786 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
7787 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
7788 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
7789 // CHECK18-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
7790 // CHECK18-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
7791 // CHECK18-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
7792 // CHECK18-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
7793 // CHECK18-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
7794 // CHECK18-NEXT:    ret void
7795 //
7796 //
7797 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
7798 // CHECK18-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR2]] {
7799 // CHECK18-NEXT:  entry:
7800 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7801 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
7802 // CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
7803 // CHECK18-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
7804 // CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
7805 // CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
7806 // CHECK18-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
7807 // CHECK18-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
7808 // CHECK18-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
7809 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
7810 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7811 // CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
7812 // CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
7813 // CHECK18-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
7814 // CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
7815 // CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
7816 // CHECK18-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
7817 // CHECK18-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
7818 // CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
7819 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7820 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
7821 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
7822 // CHECK18-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
7823 // CHECK18-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
7824 // CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
7825 // CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
7826 // CHECK18-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
7827 // CHECK18-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
7828 // CHECK18-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7829 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
7830 // CHECK18-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 8
7831 // CHECK18-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
7832 // CHECK18-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 8
7833 // CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
7834 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP11]], align 8
7835 // CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
7836 // CHECK18-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 8
7837 // CHECK18-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
7838 // CHECK18-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 8
7839 // CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
7840 // CHECK18-NEXT:    store i64 [[TMP4]], i64* [[TMP14]], align 8
7841 // CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
7842 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP15]], align 8
7843 // CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
7844 // CHECK18-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 8
7845 // CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
7846 // CHECK18-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 8
7847 // 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..7 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
7848 // CHECK18-NEXT:    ret void
7849 //
7850 //
7851 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..7
7852 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
7853 // CHECK18-NEXT:  entry:
7854 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7855 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7856 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
7857 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
7858 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7859 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7860 // CHECK18-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
7861 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
7862 // CHECK18-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
7863 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
7864 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
7865 // CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
7866 // CHECK18-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 8
7867 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
7868 // CHECK18-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
7869 // CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
7870 // CHECK18-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 8
7871 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
7872 // CHECK18-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 8
7873 // CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 5
7874 // CHECK18-NEXT:    [[TMP12:%.*]] = load i64, i64* [[TMP11]], align 8
7875 // CHECK18-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 6
7876 // CHECK18-NEXT:    [[TMP14:%.*]] = load i64, i64* [[TMP13]], align 8
7877 // CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 7
7878 // CHECK18-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 8
7879 // CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 8
7880 // CHECK18-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 8
7881 // CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
7882 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
7883 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
7884 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i64 0, i64 2
7885 // CHECK18-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
7886 // CHECK18-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
7887 // CHECK18-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
7888 // CHECK18-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
7889 // CHECK18-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
7890 // CHECK18-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 3
7891 // CHECK18-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
7892 // CHECK18-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
7893 // CHECK18-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
7894 // CHECK18-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
7895 // CHECK18-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
7896 // CHECK18-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i64 0, i64 1
7897 // CHECK18-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i64 0, i64 2
7898 // CHECK18-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
7899 // CHECK18-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
7900 // CHECK18-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
7901 // CHECK18-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP14]]
7902 // CHECK18-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i64 [[TMP23]]
7903 // CHECK18-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i64 3
7904 // CHECK18-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
7905 // CHECK18-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
7906 // CHECK18-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
7907 // CHECK18-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
7908 // CHECK18-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8
7909 // CHECK18-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
7910 // CHECK18-NEXT:    store i64 [[ADD13]], i64* [[X]], align 8
7911 // CHECK18-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
7912 // CHECK18-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8
7913 // CHECK18-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
7914 // CHECK18-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
7915 // CHECK18-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
7916 // CHECK18-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 8
7917 // CHECK18-NEXT:    ret void
7918 //
7919 //
7920 // CHECK18-LABEL: define {{[^@]+}}@_Z3bari
7921 // CHECK18-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
7922 // CHECK18-NEXT:  entry:
7923 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7924 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
7925 // CHECK18-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
7926 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7927 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
7928 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
7929 // CHECK18-NEXT:    [[CALL:%.*]] = call signext i32 @_Z3fooi(i32 signext [[TMP0]])
7930 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
7931 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
7932 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
7933 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
7934 // CHECK18-NEXT:    [[CALL1:%.*]] = call signext i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 8 dereferenceable(8) [[S]], i32 signext [[TMP2]])
7935 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
7936 // CHECK18-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
7937 // CHECK18-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
7938 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
7939 // CHECK18-NEXT:    [[CALL3:%.*]] = call signext i32 @_ZL7fstatici(i32 signext [[TMP4]])
7940 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
7941 // CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
7942 // CHECK18-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
7943 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
7944 // CHECK18-NEXT:    [[CALL5:%.*]] = call signext i32 @_Z9ftemplateIiET_i(i32 signext [[TMP6]])
7945 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
7946 // CHECK18-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
7947 // CHECK18-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
7948 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
7949 // CHECK18-NEXT:    ret i32 [[TMP8]]
7950 //
7951 //
7952 // CHECK18-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
7953 // CHECK18-SAME: (%struct.S1* nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
7954 // CHECK18-NEXT:  entry:
7955 // CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
7956 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7957 // CHECK18-NEXT:    [[B:%.*]] = alloca i32, align 4
7958 // CHECK18-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
7959 // CHECK18-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
7960 // CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
7961 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
7962 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
7963 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
7964 // CHECK18-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
7965 // CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
7966 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7967 // CHECK18-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
7968 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
7969 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
7970 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
7971 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
7972 // CHECK18-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
7973 // CHECK18-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
7974 // CHECK18-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
7975 // CHECK18-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
7976 // CHECK18-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
7977 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
7978 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
7979 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
7980 // CHECK18-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
7981 // CHECK18-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
7982 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
7983 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
7984 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
7985 // CHECK18:       omp_if.then:
7986 // CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
7987 // CHECK18-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
7988 // CHECK18-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
7989 // CHECK18-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7990 // CHECK18-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
7991 // CHECK18-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 8
7992 // CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7993 // CHECK18-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
7994 // CHECK18-NEXT:    store double* [[A]], double** [[TMP13]], align 8
7995 // CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
7996 // CHECK18-NEXT:    store i64 8, i64* [[TMP14]], align 8
7997 // CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
7998 // CHECK18-NEXT:    store i8* null, i8** [[TMP15]], align 8
7999 // CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
8000 // CHECK18-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
8001 // CHECK18-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
8002 // CHECK18-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
8003 // CHECK18-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
8004 // CHECK18-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
8005 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
8006 // CHECK18-NEXT:    store i64 4, i64* [[TMP20]], align 8
8007 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
8008 // CHECK18-NEXT:    store i8* null, i8** [[TMP21]], align 8
8009 // CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
8010 // CHECK18-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
8011 // CHECK18-NEXT:    store i64 2, i64* [[TMP23]], align 8
8012 // CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
8013 // CHECK18-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
8014 // CHECK18-NEXT:    store i64 2, i64* [[TMP25]], align 8
8015 // CHECK18-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
8016 // CHECK18-NEXT:    store i64 8, i64* [[TMP26]], align 8
8017 // CHECK18-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
8018 // CHECK18-NEXT:    store i8* null, i8** [[TMP27]], align 8
8019 // CHECK18-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
8020 // CHECK18-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
8021 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
8022 // CHECK18-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
8023 // CHECK18-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
8024 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP31]], align 8
8025 // CHECK18-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
8026 // CHECK18-NEXT:    store i64 8, i64* [[TMP32]], align 8
8027 // CHECK18-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
8028 // CHECK18-NEXT:    store i8* null, i8** [[TMP33]], align 8
8029 // CHECK18-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
8030 // CHECK18-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
8031 // CHECK18-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 8
8032 // CHECK18-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
8033 // CHECK18-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
8034 // CHECK18-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 8
8035 // CHECK18-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
8036 // CHECK18-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 8
8037 // CHECK18-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
8038 // CHECK18-NEXT:    store i8* null, i8** [[TMP39]], align 8
8039 // CHECK18-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8040 // CHECK18-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8041 // CHECK18-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
8042 // CHECK18-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
8043 // CHECK18-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
8044 // CHECK18-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8045 // CHECK18:       omp_offload.failed:
8046 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
8047 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
8048 // CHECK18:       omp_offload.cont:
8049 // CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
8050 // CHECK18:       omp_if.else:
8051 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
8052 // CHECK18-NEXT:    br label [[OMP_IF_END]]
8053 // CHECK18:       omp_if.end:
8054 // CHECK18-NEXT:    [[TMP45:%.*]] = mul nsw i64 1, [[TMP2]]
8055 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP45]]
8056 // CHECK18-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
8057 // CHECK18-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
8058 // CHECK18-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP46]] to i32
8059 // CHECK18-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
8060 // CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP47]]
8061 // CHECK18-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
8062 // CHECK18-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
8063 // CHECK18-NEXT:    ret i32 [[ADD4]]
8064 //
8065 //
8066 // CHECK18-LABEL: define {{[^@]+}}@_ZL7fstatici
8067 // CHECK18-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] {
8068 // CHECK18-NEXT:  entry:
8069 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8070 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
8071 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
8072 // CHECK18-NEXT:    [[AAA:%.*]] = alloca i8, align 1
8073 // CHECK18-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
8074 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8075 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
8076 // CHECK18-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
8077 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
8078 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
8079 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
8080 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8081 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
8082 // CHECK18-NEXT:    store i16 0, i16* [[AA]], align 2
8083 // CHECK18-NEXT:    store i8 0, i8* [[AAA]], align 1
8084 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
8085 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8086 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
8087 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
8088 // CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
8089 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
8090 // CHECK18-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
8091 // CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
8092 // CHECK18-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
8093 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
8094 // CHECK18-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
8095 // CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
8096 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
8097 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
8098 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
8099 // CHECK18:       omp_if.then:
8100 // CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8101 // CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
8102 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
8103 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8104 // CHECK18-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
8105 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
8106 // CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
8107 // CHECK18-NEXT:    store i8* null, i8** [[TMP11]], align 8
8108 // CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
8109 // CHECK18-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
8110 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
8111 // CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
8112 // CHECK18-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
8113 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
8114 // CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
8115 // CHECK18-NEXT:    store i8* null, i8** [[TMP16]], align 8
8116 // CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
8117 // CHECK18-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
8118 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
8119 // CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
8120 // CHECK18-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
8121 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
8122 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
8123 // CHECK18-NEXT:    store i8* null, i8** [[TMP21]], align 8
8124 // CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
8125 // CHECK18-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
8126 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
8127 // CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
8128 // CHECK18-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
8129 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
8130 // CHECK18-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
8131 // CHECK18-NEXT:    store i8* null, i8** [[TMP26]], align 8
8132 // CHECK18-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8133 // CHECK18-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8134 // CHECK18-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
8135 // CHECK18-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
8136 // CHECK18-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8137 // CHECK18:       omp_offload.failed:
8138 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
8139 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
8140 // CHECK18:       omp_offload.cont:
8141 // CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
8142 // CHECK18:       omp_if.else:
8143 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
8144 // CHECK18-NEXT:    br label [[OMP_IF_END]]
8145 // CHECK18:       omp_if.end:
8146 // CHECK18-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
8147 // CHECK18-NEXT:    ret i32 [[TMP31]]
8148 //
8149 //
8150 // CHECK18-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
8151 // CHECK18-SAME: (i32 signext [[N:%.*]]) #[[ATTR0]] comdat {
8152 // CHECK18-NEXT:  entry:
8153 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8154 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
8155 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
8156 // CHECK18-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
8157 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8158 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
8159 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
8160 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
8161 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
8162 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8163 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
8164 // CHECK18-NEXT:    store i16 0, i16* [[AA]], align 2
8165 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
8166 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8167 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
8168 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
8169 // CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
8170 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
8171 // CHECK18-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
8172 // CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
8173 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
8174 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
8175 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
8176 // CHECK18:       omp_if.then:
8177 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8178 // CHECK18-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
8179 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
8180 // CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8181 // CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
8182 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
8183 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
8184 // CHECK18-NEXT:    store i8* null, i8** [[TMP9]], align 8
8185 // CHECK18-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
8186 // CHECK18-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
8187 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
8188 // CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
8189 // CHECK18-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
8190 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
8191 // CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
8192 // CHECK18-NEXT:    store i8* null, i8** [[TMP14]], align 8
8193 // CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
8194 // CHECK18-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
8195 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
8196 // CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
8197 // CHECK18-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
8198 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
8199 // CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
8200 // CHECK18-NEXT:    store i8* null, i8** [[TMP19]], align 8
8201 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8202 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8203 // CHECK18-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
8204 // CHECK18-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
8205 // CHECK18-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8206 // CHECK18:       omp_offload.failed:
8207 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
8208 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
8209 // CHECK18:       omp_offload.cont:
8210 // CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
8211 // CHECK18:       omp_if.else:
8212 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
8213 // CHECK18-NEXT:    br label [[OMP_IF_END]]
8214 // CHECK18:       omp_if.end:
8215 // CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
8216 // CHECK18-NEXT:    ret i32 [[TMP24]]
8217 //
8218 //
8219 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
8220 // CHECK18-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
8221 // CHECK18-NEXT:  entry:
8222 // CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
8223 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
8224 // CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
8225 // CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
8226 // CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
8227 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
8228 // CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
8229 // CHECK18-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
8230 // CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8231 // CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
8232 // CHECK18-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
8233 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
8234 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
8235 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8236 // CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
8237 // CHECK18-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
8238 // CHECK18-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
8239 // CHECK18-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 8
8240 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
8241 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[CONV]], align 8
8242 // CHECK18-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
8243 // CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
8244 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP7]], align 8
8245 // CHECK18-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
8246 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP8]], align 8
8247 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
8248 // CHECK18-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 8
8249 // 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.5*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
8250 // CHECK18-NEXT:    ret void
8251 //
8252 //
8253 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..9
8254 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
8255 // CHECK18-NEXT:  entry:
8256 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8257 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8258 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
8259 // CHECK18-NEXT:    [[B:%.*]] = alloca i32, align 4
8260 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8261 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8262 // CHECK18-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
8263 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
8264 // CHECK18-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
8265 // CHECK18-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
8266 // CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
8267 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
8268 // CHECK18-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
8269 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
8270 // CHECK18-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
8271 // CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
8272 // CHECK18-NEXT:    [[TMP8:%.*]] = load i64, i64* [[TMP7]], align 8
8273 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
8274 // CHECK18-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 8
8275 // CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
8276 // CHECK18-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
8277 // CHECK18-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
8278 // CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
8279 // CHECK18-NEXT:    store double [[ADD]], double* [[A]], align 8
8280 // CHECK18-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
8281 // CHECK18-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 8
8282 // CHECK18-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
8283 // CHECK18-NEXT:    store double [[INC]], double* [[A1]], align 8
8284 // CHECK18-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
8285 // CHECK18-NEXT:    [[TMP13:%.*]] = mul nsw i64 1, [[TMP8]]
8286 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i64 [[TMP13]]
8287 // CHECK18-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
8288 // CHECK18-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
8289 // CHECK18-NEXT:    ret void
8290 //
8291 //
8292 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
8293 // CHECK18-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
8294 // CHECK18-NEXT:  entry:
8295 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8296 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8297 // CHECK18-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
8298 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
8299 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 8
8300 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8301 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8302 // CHECK18-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
8303 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
8304 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8305 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8306 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
8307 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
8308 // CHECK18-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
8309 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
8310 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
8311 // CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
8312 // CHECK18-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
8313 // CHECK18-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
8314 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
8315 // CHECK18-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV2]], align 8
8316 // CHECK18-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
8317 // CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
8318 // CHECK18-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 8
8319 // 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.6*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
8320 // CHECK18-NEXT:    ret void
8321 //
8322 //
8323 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..11
8324 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
8325 // CHECK18-NEXT:  entry:
8326 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8327 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8328 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 8
8329 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
8330 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
8331 // CHECK18-NEXT:    [[AAA:%.*]] = alloca i8, align 1
8332 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8333 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8334 // CHECK18-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 8
8335 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 8
8336 // CHECK18-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
8337 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
8338 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
8339 // CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
8340 // CHECK18-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
8341 // CHECK18-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
8342 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
8343 // CHECK18-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
8344 // CHECK18-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
8345 // CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
8346 // CHECK18-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 8
8347 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
8348 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
8349 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
8350 // CHECK18-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
8351 // CHECK18-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
8352 // CHECK18-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
8353 // CHECK18-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
8354 // CHECK18-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
8355 // CHECK18-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
8356 // CHECK18-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
8357 // CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
8358 // CHECK18-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
8359 // CHECK18-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
8360 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i64 0, i64 2
8361 // CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
8362 // CHECK18-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
8363 // CHECK18-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
8364 // CHECK18-NEXT:    ret void
8365 //
8366 //
8367 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
8368 // CHECK18-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
8369 // CHECK18-NEXT:  entry:
8370 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8371 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8372 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
8373 // CHECK18-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 8
8374 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8375 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8376 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
8377 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8378 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8379 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
8380 // CHECK18-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
8381 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
8382 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
8383 // CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
8384 // CHECK18-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
8385 // CHECK18-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
8386 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
8387 // CHECK18-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 8
8388 // 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.7*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
8389 // CHECK18-NEXT:    ret void
8390 //
8391 //
8392 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..14
8393 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
8394 // CHECK18-NEXT:  entry:
8395 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8396 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8397 // CHECK18-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 8
8398 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
8399 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
8400 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8401 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8402 // CHECK18-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 8
8403 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 8
8404 // CHECK18-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
8405 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
8406 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
8407 // CHECK18-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
8408 // CHECK18-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
8409 // CHECK18-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
8410 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
8411 // CHECK18-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 8
8412 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
8413 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
8414 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
8415 // CHECK18-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
8416 // CHECK18-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
8417 // CHECK18-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
8418 // CHECK18-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
8419 // CHECK18-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
8420 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i64 0, i64 2
8421 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
8422 // CHECK18-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
8423 // CHECK18-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
8424 // CHECK18-NEXT:    ret void
8425 //
8426 //
8427 // CHECK18-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
8428 // CHECK18-SAME: () #[[ATTR6:[0-9]+]] {
8429 // CHECK18-NEXT:  entry:
8430 // CHECK18-NEXT:    call void @__tgt_register_requires(i64 1)
8431 // CHECK18-NEXT:    ret void
8432 //
8433 //
8434 // CHECK19-LABEL: define {{[^@]+}}@_Z3fooi
8435 // CHECK19-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
8436 // CHECK19-NEXT:  entry:
8437 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8438 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
8439 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
8440 // CHECK19-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
8441 // CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
8442 // CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
8443 // CHECK19-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
8444 // CHECK19-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
8445 // CHECK19-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
8446 // CHECK19-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
8447 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
8448 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
8449 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
8450 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
8451 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
8452 // CHECK19-NEXT:    [[A_CASTED2:%.*]] = alloca i32, align 4
8453 // CHECK19-NEXT:    [[AA_CASTED3:%.*]] = alloca i32, align 4
8454 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4
8455 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4
8456 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4
8457 // CHECK19-NEXT:    [[A_CASTED10:%.*]] = alloca i32, align 4
8458 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [9 x i8*], align 4
8459 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [9 x i8*], align 4
8460 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [9 x i8*], align 4
8461 // CHECK19-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 4
8462 // CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
8463 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8464 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
8465 // CHECK19-NEXT:    store i16 0, i16* [[AA]], align 2
8466 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
8467 // CHECK19-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
8468 // CHECK19-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
8469 // CHECK19-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
8470 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
8471 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
8472 // CHECK19-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
8473 // CHECK19-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
8474 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
8475 // CHECK19-NEXT:    [[TMP5:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i32 20, i32 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
8476 // CHECK19-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to %struct.kmp_task_t_with_privates*
8477 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP6]], i32 0, i32 0
8478 // CHECK19-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP5]])
8479 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
8480 // CHECK19-NEXT:    store i32 [[TMP9]], i32* [[A_CASTED]], align 4
8481 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[A_CASTED]], align 4
8482 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i32 [[TMP10]]) #[[ATTR3:[0-9]+]]
8483 // CHECK19-NEXT:    [[TMP11:%.*]] = load i16, i16* [[AA]], align 2
8484 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
8485 // CHECK19-NEXT:    store i16 [[TMP11]], i16* [[CONV]], align 2
8486 // CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[AA_CASTED]], align 4
8487 // CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8488 // CHECK19-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
8489 // CHECK19-NEXT:    store i32 [[TMP12]], i32* [[TMP14]], align 4
8490 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8491 // CHECK19-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
8492 // CHECK19-NEXT:    store i32 [[TMP12]], i32* [[TMP16]], align 4
8493 // CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
8494 // CHECK19-NEXT:    store i8* null, i8** [[TMP17]], align 4
8495 // CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8496 // CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8497 // CHECK19-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP18]], i8** [[TMP19]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
8498 // CHECK19-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
8499 // CHECK19-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8500 // CHECK19:       omp_offload.failed:
8501 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i32 [[TMP12]]) #[[ATTR3]]
8502 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
8503 // CHECK19:       omp_offload.cont:
8504 // CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[A]], align 4
8505 // CHECK19-NEXT:    store i32 [[TMP22]], i32* [[A_CASTED2]], align 4
8506 // CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A_CASTED2]], align 4
8507 // CHECK19-NEXT:    [[TMP24:%.*]] = load i16, i16* [[AA]], align 2
8508 // CHECK19-NEXT:    [[CONV4:%.*]] = bitcast i32* [[AA_CASTED3]] to i16*
8509 // CHECK19-NEXT:    store i16 [[TMP24]], i16* [[CONV4]], align 2
8510 // CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[AA_CASTED3]], align 4
8511 // CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
8512 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP26]], 10
8513 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
8514 // CHECK19:       omp_if.then:
8515 // CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
8516 // CHECK19-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
8517 // CHECK19-NEXT:    store i32 [[TMP23]], i32* [[TMP28]], align 4
8518 // CHECK19-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
8519 // CHECK19-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
8520 // CHECK19-NEXT:    store i32 [[TMP23]], i32* [[TMP30]], align 4
8521 // CHECK19-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
8522 // CHECK19-NEXT:    store i8* null, i8** [[TMP31]], align 4
8523 // CHECK19-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
8524 // CHECK19-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i32*
8525 // CHECK19-NEXT:    store i32 [[TMP25]], i32* [[TMP33]], align 4
8526 // CHECK19-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
8527 // CHECK19-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i32*
8528 // CHECK19-NEXT:    store i32 [[TMP25]], i32* [[TMP35]], align 4
8529 // CHECK19-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
8530 // CHECK19-NEXT:    store i8* null, i8** [[TMP36]], align 4
8531 // CHECK19-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
8532 // CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
8533 // CHECK19-NEXT:    [[TMP39:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP37]], i8** [[TMP38]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
8534 // CHECK19-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
8535 // CHECK19-NEXT:    br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
8536 // CHECK19:       omp_offload.failed8:
8537 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR3]]
8538 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
8539 // CHECK19:       omp_offload.cont9:
8540 // CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
8541 // CHECK19:       omp_if.else:
8542 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR3]]
8543 // CHECK19-NEXT:    br label [[OMP_IF_END]]
8544 // CHECK19:       omp_if.end:
8545 // CHECK19-NEXT:    [[TMP41:%.*]] = load i32, i32* [[A]], align 4
8546 // CHECK19-NEXT:    store i32 [[TMP41]], i32* [[A_CASTED10]], align 4
8547 // CHECK19-NEXT:    [[TMP42:%.*]] = load i32, i32* [[A_CASTED10]], align 4
8548 // CHECK19-NEXT:    [[TMP43:%.*]] = load i32, i32* [[N_ADDR]], align 4
8549 // CHECK19-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP43]], 20
8550 // CHECK19-NEXT:    br i1 [[CMP11]], label [[OMP_IF_THEN12:%.*]], label [[OMP_IF_ELSE18:%.*]]
8551 // CHECK19:       omp_if.then12:
8552 // CHECK19-NEXT:    [[TMP44:%.*]] = mul nuw i32 [[TMP1]], 4
8553 // CHECK19-NEXT:    [[TMP45:%.*]] = sext i32 [[TMP44]] to i64
8554 // CHECK19-NEXT:    [[TMP46:%.*]] = mul nuw i32 5, [[TMP3]]
8555 // CHECK19-NEXT:    [[TMP47:%.*]] = mul nuw i32 [[TMP46]], 8
8556 // CHECK19-NEXT:    [[TMP48:%.*]] = sext i32 [[TMP47]] to i64
8557 // CHECK19-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
8558 // CHECK19-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32*
8559 // CHECK19-NEXT:    store i32 [[TMP42]], i32* [[TMP50]], align 4
8560 // CHECK19-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
8561 // CHECK19-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
8562 // CHECK19-NEXT:    store i32 [[TMP42]], i32* [[TMP52]], align 4
8563 // CHECK19-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
8564 // CHECK19-NEXT:    store i64 4, i64* [[TMP53]], align 4
8565 // CHECK19-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
8566 // CHECK19-NEXT:    store i8* null, i8** [[TMP54]], align 4
8567 // CHECK19-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 1
8568 // CHECK19-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
8569 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 4
8570 // CHECK19-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 1
8571 // CHECK19-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
8572 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 4
8573 // CHECK19-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
8574 // CHECK19-NEXT:    store i64 40, i64* [[TMP59]], align 4
8575 // CHECK19-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 1
8576 // CHECK19-NEXT:    store i8* null, i8** [[TMP60]], align 4
8577 // CHECK19-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 2
8578 // CHECK19-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
8579 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP62]], align 4
8580 // CHECK19-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 2
8581 // CHECK19-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i32*
8582 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP64]], align 4
8583 // CHECK19-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
8584 // CHECK19-NEXT:    store i64 4, i64* [[TMP65]], align 4
8585 // CHECK19-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 2
8586 // CHECK19-NEXT:    store i8* null, i8** [[TMP66]], align 4
8587 // CHECK19-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 3
8588 // CHECK19-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
8589 // CHECK19-NEXT:    store float* [[VLA]], float** [[TMP68]], align 4
8590 // CHECK19-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 3
8591 // CHECK19-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
8592 // CHECK19-NEXT:    store float* [[VLA]], float** [[TMP70]], align 4
8593 // CHECK19-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
8594 // CHECK19-NEXT:    store i64 [[TMP45]], i64* [[TMP71]], align 4
8595 // CHECK19-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 3
8596 // CHECK19-NEXT:    store i8* null, i8** [[TMP72]], align 4
8597 // CHECK19-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 4
8598 // CHECK19-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
8599 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 4
8600 // CHECK19-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 4
8601 // CHECK19-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
8602 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 4
8603 // CHECK19-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
8604 // CHECK19-NEXT:    store i64 400, i64* [[TMP77]], align 4
8605 // CHECK19-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 4
8606 // CHECK19-NEXT:    store i8* null, i8** [[TMP78]], align 4
8607 // CHECK19-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 5
8608 // CHECK19-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i32*
8609 // CHECK19-NEXT:    store i32 5, i32* [[TMP80]], align 4
8610 // CHECK19-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 5
8611 // CHECK19-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i32*
8612 // CHECK19-NEXT:    store i32 5, i32* [[TMP82]], align 4
8613 // CHECK19-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
8614 // CHECK19-NEXT:    store i64 4, i64* [[TMP83]], align 4
8615 // CHECK19-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 5
8616 // CHECK19-NEXT:    store i8* null, i8** [[TMP84]], align 4
8617 // CHECK19-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 6
8618 // CHECK19-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i32*
8619 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP86]], align 4
8620 // CHECK19-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 6
8621 // CHECK19-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
8622 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP88]], align 4
8623 // CHECK19-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
8624 // CHECK19-NEXT:    store i64 4, i64* [[TMP89]], align 4
8625 // CHECK19-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 6
8626 // CHECK19-NEXT:    store i8* null, i8** [[TMP90]], align 4
8627 // CHECK19-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 7
8628 // CHECK19-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
8629 // CHECK19-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 4
8630 // CHECK19-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 7
8631 // CHECK19-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
8632 // CHECK19-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 4
8633 // CHECK19-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
8634 // CHECK19-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 4
8635 // CHECK19-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 7
8636 // CHECK19-NEXT:    store i8* null, i8** [[TMP96]], align 4
8637 // CHECK19-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 8
8638 // CHECK19-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
8639 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 4
8640 // CHECK19-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 8
8641 // CHECK19-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
8642 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 4
8643 // CHECK19-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
8644 // CHECK19-NEXT:    store i64 12, i64* [[TMP101]], align 4
8645 // CHECK19-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 8
8646 // CHECK19-NEXT:    store i8* null, i8** [[TMP102]], align 4
8647 // CHECK19-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
8648 // CHECK19-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
8649 // CHECK19-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
8650 // CHECK19-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
8651 // CHECK19-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
8652 // CHECK19-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
8653 // CHECK19:       omp_offload.failed16:
8654 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
8655 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
8656 // CHECK19:       omp_offload.cont17:
8657 // CHECK19-NEXT:    br label [[OMP_IF_END19:%.*]]
8658 // CHECK19:       omp_if.else18:
8659 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
8660 // CHECK19-NEXT:    br label [[OMP_IF_END19]]
8661 // CHECK19:       omp_if.end19:
8662 // CHECK19-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
8663 // CHECK19-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
8664 // CHECK19-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
8665 // CHECK19-NEXT:    ret i32 [[TMP108]]
8666 //
8667 //
8668 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
8669 // CHECK19-SAME: () #[[ATTR2:[0-9]+]] {
8670 // CHECK19-NEXT:  entry:
8671 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
8672 // 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_]])
8673 // CHECK19-NEXT:    ret void
8674 //
8675 //
8676 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined.
8677 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
8678 // CHECK19-NEXT:  entry:
8679 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8680 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8681 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
8682 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8683 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8684 // CHECK19-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
8685 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
8686 // CHECK19-NEXT:    ret void
8687 //
8688 //
8689 // CHECK19-LABEL: define {{[^@]+}}@.omp_task_entry.
8690 // CHECK19-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
8691 // CHECK19-NEXT:  entry:
8692 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
8693 // CHECK19-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
8694 // CHECK19-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
8695 // CHECK19-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
8696 // CHECK19-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
8697 // CHECK19-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 4
8698 // CHECK19-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
8699 // CHECK19-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
8700 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
8701 // CHECK19-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
8702 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
8703 // CHECK19-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
8704 // CHECK19-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
8705 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
8706 // CHECK19-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
8707 // CHECK19-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
8708 // CHECK19-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
8709 // CHECK19-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
8710 // CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
8711 // CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
8712 // CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
8713 // CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
8714 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !22
8715 // CHECK19-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !22
8716 // CHECK19-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
8717 // CHECK19-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
8718 // CHECK19-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !22
8719 // CHECK19-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
8720 // CHECK19-NEXT:    [[TMP10:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
8721 // CHECK19-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0) #[[ATTR3]]
8722 // CHECK19-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
8723 // CHECK19-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
8724 // CHECK19:       omp_offload.failed.i:
8725 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR3]]
8726 // CHECK19-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
8727 // CHECK19:       .omp_outlined..1.exit:
8728 // CHECK19-NEXT:    ret i32 0
8729 //
8730 //
8731 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
8732 // CHECK19-SAME: (i32 [[A:%.*]]) #[[ATTR2]] {
8733 // CHECK19-NEXT:  entry:
8734 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
8735 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
8736 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
8737 // CHECK19-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
8738 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
8739 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
8740 // 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..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
8741 // CHECK19-NEXT:    ret void
8742 //
8743 //
8744 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..2
8745 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
8746 // CHECK19-NEXT:  entry:
8747 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8748 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8749 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
8750 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
8751 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8752 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8753 // CHECK19-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
8754 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
8755 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
8756 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
8757 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
8758 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
8759 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], 1
8760 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
8761 // CHECK19-NEXT:    ret void
8762 //
8763 //
8764 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
8765 // CHECK19-SAME: (i32 [[AA:%.*]]) #[[ATTR2]] {
8766 // CHECK19-NEXT:  entry:
8767 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
8768 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 2
8769 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
8770 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
8771 // CHECK19-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
8772 // CHECK19-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
8773 // CHECK19-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
8774 // 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..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
8775 // CHECK19-NEXT:    ret void
8776 //
8777 //
8778 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..3
8779 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
8780 // CHECK19-NEXT:  entry:
8781 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8782 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8783 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
8784 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
8785 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8786 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8787 // CHECK19-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
8788 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
8789 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
8790 // CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
8791 // CHECK19-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
8792 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
8793 // CHECK19-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
8794 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
8795 // CHECK19-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
8796 // CHECK19-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
8797 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8798 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
8799 // CHECK19-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
8800 // CHECK19-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
8801 // CHECK19-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
8802 // CHECK19:       .cancel.exit:
8803 // CHECK19-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
8804 // CHECK19-NEXT:    br label [[DOTCANCEL_CONTINUE]]
8805 // CHECK19:       .cancel.continue:
8806 // CHECK19-NEXT:    ret void
8807 //
8808 //
8809 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
8810 // CHECK19-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR2]] {
8811 // CHECK19-NEXT:  entry:
8812 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
8813 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
8814 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 4
8815 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
8816 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
8817 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
8818 // CHECK19-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
8819 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
8820 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
8821 // CHECK19-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
8822 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
8823 // CHECK19-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
8824 // 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..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
8825 // CHECK19-NEXT:    ret void
8826 //
8827 //
8828 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..4
8829 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
8830 // CHECK19-NEXT:  entry:
8831 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8832 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8833 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
8834 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
8835 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
8836 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8837 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8838 // CHECK19-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
8839 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
8840 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
8841 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
8842 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
8843 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
8844 // CHECK19-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
8845 // CHECK19-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
8846 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
8847 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
8848 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
8849 // CHECK19-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
8850 // CHECK19-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
8851 // CHECK19-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
8852 // CHECK19-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
8853 // CHECK19-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
8854 // CHECK19-NEXT:    ret void
8855 //
8856 //
8857 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
8858 // CHECK19-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR2]] {
8859 // CHECK19-NEXT:  entry:
8860 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
8861 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
8862 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
8863 // CHECK19-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
8864 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
8865 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
8866 // CHECK19-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
8867 // CHECK19-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
8868 // CHECK19-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
8869 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
8870 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
8871 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
8872 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
8873 // CHECK19-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
8874 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
8875 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
8876 // CHECK19-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
8877 // CHECK19-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
8878 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
8879 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
8880 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
8881 // CHECK19-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
8882 // CHECK19-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
8883 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
8884 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
8885 // CHECK19-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
8886 // CHECK19-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
8887 // CHECK19-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
8888 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
8889 // CHECK19-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 4
8890 // CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
8891 // CHECK19-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 4
8892 // CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
8893 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP11]], align 4
8894 // CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
8895 // CHECK19-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 4
8896 // CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
8897 // CHECK19-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 4
8898 // CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
8899 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[TMP14]], align 4
8900 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
8901 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP15]], align 4
8902 // CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
8903 // CHECK19-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 4
8904 // CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
8905 // CHECK19-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 4
8906 // 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..7 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
8907 // CHECK19-NEXT:    ret void
8908 //
8909 //
8910 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..7
8911 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
8912 // CHECK19-NEXT:  entry:
8913 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8914 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8915 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
8916 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
8917 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8918 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8919 // CHECK19-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
8920 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
8921 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
8922 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
8923 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
8924 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
8925 // CHECK19-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 4
8926 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
8927 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
8928 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
8929 // CHECK19-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 4
8930 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
8931 // CHECK19-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 4
8932 // CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 5
8933 // CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
8934 // CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 6
8935 // CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
8936 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 7
8937 // CHECK19-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 4
8938 // CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 8
8939 // CHECK19-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 4
8940 // CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
8941 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
8942 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
8943 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i32 0, i32 2
8944 // CHECK19-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
8945 // CHECK19-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
8946 // CHECK19-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
8947 // CHECK19-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
8948 // CHECK19-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
8949 // CHECK19-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i32 3
8950 // CHECK19-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
8951 // CHECK19-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
8952 // CHECK19-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
8953 // CHECK19-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
8954 // CHECK19-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
8955 // CHECK19-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i32 0, i32 1
8956 // CHECK19-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i32 0, i32 2
8957 // CHECK19-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
8958 // CHECK19-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
8959 // CHECK19-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
8960 // CHECK19-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP14]]
8961 // CHECK19-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i32 [[TMP23]]
8962 // CHECK19-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i32 3
8963 // CHECK19-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
8964 // CHECK19-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
8965 // CHECK19-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
8966 // CHECK19-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
8967 // CHECK19-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4
8968 // CHECK19-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
8969 // CHECK19-NEXT:    store i64 [[ADD13]], i64* [[X]], align 4
8970 // CHECK19-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
8971 // CHECK19-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4
8972 // CHECK19-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
8973 // CHECK19-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
8974 // CHECK19-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
8975 // CHECK19-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 4
8976 // CHECK19-NEXT:    ret void
8977 //
8978 //
8979 // CHECK19-LABEL: define {{[^@]+}}@_Z3bari
8980 // CHECK19-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
8981 // CHECK19-NEXT:  entry:
8982 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8983 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
8984 // CHECK19-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
8985 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8986 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
8987 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
8988 // CHECK19-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooi(i32 [[TMP0]])
8989 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
8990 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
8991 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
8992 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
8993 // CHECK19-NEXT:    [[CALL1:%.*]] = call i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], i32 [[TMP2]])
8994 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
8995 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
8996 // CHECK19-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
8997 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
8998 // CHECK19-NEXT:    [[CALL3:%.*]] = call i32 @_ZL7fstatici(i32 [[TMP4]])
8999 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
9000 // CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
9001 // CHECK19-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
9002 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
9003 // CHECK19-NEXT:    [[CALL5:%.*]] = call i32 @_Z9ftemplateIiET_i(i32 [[TMP6]])
9004 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
9005 // CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
9006 // CHECK19-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
9007 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
9008 // CHECK19-NEXT:    ret i32 [[TMP8]]
9009 //
9010 //
9011 // CHECK19-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
9012 // CHECK19-SAME: (%struct.S1* nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 [[N:%.*]]) #[[ATTR0]] comdat align 2 {
9013 // CHECK19-NEXT:  entry:
9014 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
9015 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
9016 // CHECK19-NEXT:    [[B:%.*]] = alloca i32, align 4
9017 // CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
9018 // CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
9019 // CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
9020 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
9021 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
9022 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
9023 // CHECK19-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
9024 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
9025 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
9026 // CHECK19-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
9027 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
9028 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
9029 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
9030 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
9031 // CHECK19-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
9032 // CHECK19-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
9033 // CHECK19-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
9034 // CHECK19-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
9035 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
9036 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
9037 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
9038 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
9039 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
9040 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
9041 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
9042 // CHECK19:       omp_if.then:
9043 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
9044 // CHECK19-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
9045 // CHECK19-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
9046 // CHECK19-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
9047 // CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9048 // CHECK19-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
9049 // CHECK19-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 4
9050 // CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9051 // CHECK19-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
9052 // CHECK19-NEXT:    store double* [[A]], double** [[TMP13]], align 4
9053 // CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
9054 // CHECK19-NEXT:    store i64 8, i64* [[TMP14]], align 4
9055 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
9056 // CHECK19-NEXT:    store i8* null, i8** [[TMP15]], align 4
9057 // CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
9058 // CHECK19-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
9059 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
9060 // CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
9061 // CHECK19-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
9062 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
9063 // CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
9064 // CHECK19-NEXT:    store i64 4, i64* [[TMP20]], align 4
9065 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
9066 // CHECK19-NEXT:    store i8* null, i8** [[TMP21]], align 4
9067 // CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
9068 // CHECK19-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
9069 // CHECK19-NEXT:    store i32 2, i32* [[TMP23]], align 4
9070 // CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
9071 // CHECK19-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
9072 // CHECK19-NEXT:    store i32 2, i32* [[TMP25]], align 4
9073 // CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
9074 // CHECK19-NEXT:    store i64 4, i64* [[TMP26]], align 4
9075 // CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
9076 // CHECK19-NEXT:    store i8* null, i8** [[TMP27]], align 4
9077 // CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
9078 // CHECK19-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
9079 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
9080 // CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
9081 // CHECK19-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i32*
9082 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP31]], align 4
9083 // CHECK19-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
9084 // CHECK19-NEXT:    store i64 4, i64* [[TMP32]], align 4
9085 // CHECK19-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
9086 // CHECK19-NEXT:    store i8* null, i8** [[TMP33]], align 4
9087 // CHECK19-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
9088 // CHECK19-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
9089 // CHECK19-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 4
9090 // CHECK19-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
9091 // CHECK19-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
9092 // CHECK19-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 4
9093 // CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
9094 // CHECK19-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 4
9095 // CHECK19-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
9096 // CHECK19-NEXT:    store i8* null, i8** [[TMP39]], align 4
9097 // CHECK19-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9098 // CHECK19-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9099 // CHECK19-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
9100 // CHECK19-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
9101 // CHECK19-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
9102 // CHECK19-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
9103 // CHECK19:       omp_offload.failed:
9104 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
9105 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
9106 // CHECK19:       omp_offload.cont:
9107 // CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
9108 // CHECK19:       omp_if.else:
9109 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
9110 // CHECK19-NEXT:    br label [[OMP_IF_END]]
9111 // CHECK19:       omp_if.end:
9112 // CHECK19-NEXT:    [[TMP45:%.*]] = mul nsw i32 1, [[TMP1]]
9113 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP45]]
9114 // CHECK19-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
9115 // CHECK19-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
9116 // CHECK19-NEXT:    [[CONV:%.*]] = sext i16 [[TMP46]] to i32
9117 // CHECK19-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
9118 // CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP47]]
9119 // CHECK19-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
9120 // CHECK19-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
9121 // CHECK19-NEXT:    ret i32 [[ADD3]]
9122 //
9123 //
9124 // CHECK19-LABEL: define {{[^@]+}}@_ZL7fstatici
9125 // CHECK19-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
9126 // CHECK19-NEXT:  entry:
9127 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
9128 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
9129 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
9130 // CHECK19-NEXT:    [[AAA:%.*]] = alloca i8, align 1
9131 // CHECK19-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
9132 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
9133 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
9134 // CHECK19-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
9135 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
9136 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
9137 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
9138 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
9139 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
9140 // CHECK19-NEXT:    store i16 0, i16* [[AA]], align 2
9141 // CHECK19-NEXT:    store i8 0, i8* [[AAA]], align 1
9142 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
9143 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
9144 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
9145 // CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
9146 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
9147 // CHECK19-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
9148 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
9149 // CHECK19-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
9150 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
9151 // CHECK19-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
9152 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
9153 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
9154 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
9155 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
9156 // CHECK19:       omp_if.then:
9157 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9158 // CHECK19-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
9159 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
9160 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9161 // CHECK19-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
9162 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
9163 // CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
9164 // CHECK19-NEXT:    store i8* null, i8** [[TMP11]], align 4
9165 // CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
9166 // CHECK19-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
9167 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
9168 // CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
9169 // CHECK19-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
9170 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
9171 // CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
9172 // CHECK19-NEXT:    store i8* null, i8** [[TMP16]], align 4
9173 // CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
9174 // CHECK19-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
9175 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
9176 // CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
9177 // CHECK19-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
9178 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
9179 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
9180 // CHECK19-NEXT:    store i8* null, i8** [[TMP21]], align 4
9181 // CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
9182 // CHECK19-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
9183 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
9184 // CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
9185 // CHECK19-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
9186 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
9187 // CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
9188 // CHECK19-NEXT:    store i8* null, i8** [[TMP26]], align 4
9189 // CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9190 // CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9191 // CHECK19-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
9192 // CHECK19-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
9193 // CHECK19-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
9194 // CHECK19:       omp_offload.failed:
9195 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
9196 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
9197 // CHECK19:       omp_offload.cont:
9198 // CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
9199 // CHECK19:       omp_if.else:
9200 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
9201 // CHECK19-NEXT:    br label [[OMP_IF_END]]
9202 // CHECK19:       omp_if.end:
9203 // CHECK19-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
9204 // CHECK19-NEXT:    ret i32 [[TMP31]]
9205 //
9206 //
9207 // CHECK19-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
9208 // CHECK19-SAME: (i32 [[N:%.*]]) #[[ATTR0]] comdat {
9209 // CHECK19-NEXT:  entry:
9210 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
9211 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
9212 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
9213 // CHECK19-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
9214 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
9215 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
9216 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
9217 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
9218 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
9219 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
9220 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
9221 // CHECK19-NEXT:    store i16 0, i16* [[AA]], align 2
9222 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
9223 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
9224 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
9225 // CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
9226 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
9227 // CHECK19-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
9228 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
9229 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
9230 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
9231 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
9232 // CHECK19:       omp_if.then:
9233 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9234 // CHECK19-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
9235 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
9236 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9237 // CHECK19-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
9238 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
9239 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
9240 // CHECK19-NEXT:    store i8* null, i8** [[TMP9]], align 4
9241 // CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
9242 // CHECK19-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
9243 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
9244 // CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
9245 // CHECK19-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
9246 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
9247 // CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
9248 // CHECK19-NEXT:    store i8* null, i8** [[TMP14]], align 4
9249 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
9250 // CHECK19-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
9251 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
9252 // CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
9253 // CHECK19-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
9254 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
9255 // CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
9256 // CHECK19-NEXT:    store i8* null, i8** [[TMP19]], align 4
9257 // CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9258 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9259 // CHECK19-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
9260 // CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
9261 // CHECK19-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
9262 // CHECK19:       omp_offload.failed:
9263 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
9264 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
9265 // CHECK19:       omp_offload.cont:
9266 // CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
9267 // CHECK19:       omp_if.else:
9268 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
9269 // CHECK19-NEXT:    br label [[OMP_IF_END]]
9270 // CHECK19:       omp_if.end:
9271 // CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
9272 // CHECK19-NEXT:    ret i32 [[TMP24]]
9273 //
9274 //
9275 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
9276 // CHECK19-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
9277 // CHECK19-NEXT:  entry:
9278 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
9279 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
9280 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
9281 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
9282 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
9283 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 4
9284 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
9285 // CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
9286 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
9287 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
9288 // CHECK19-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
9289 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
9290 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
9291 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
9292 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
9293 // CHECK19-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
9294 // CHECK19-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 4
9295 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
9296 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_ADDR]], align 4
9297 // CHECK19-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 4
9298 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
9299 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP7]], align 4
9300 // CHECK19-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
9301 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[TMP8]], align 4
9302 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
9303 // CHECK19-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 4
9304 // 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.5*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
9305 // CHECK19-NEXT:    ret void
9306 //
9307 //
9308 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..9
9309 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
9310 // CHECK19-NEXT:  entry:
9311 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9312 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9313 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 4
9314 // CHECK19-NEXT:    [[B:%.*]] = alloca i32, align 4
9315 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9316 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9317 // CHECK19-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 4
9318 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 4
9319 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
9320 // CHECK19-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
9321 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
9322 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
9323 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
9324 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
9325 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
9326 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
9327 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
9328 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
9329 // CHECK19-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 4
9330 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
9331 // CHECK19-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
9332 // CHECK19-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
9333 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
9334 // CHECK19-NEXT:    store double [[ADD]], double* [[A]], align 4
9335 // CHECK19-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
9336 // CHECK19-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 4
9337 // CHECK19-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
9338 // CHECK19-NEXT:    store double [[INC]], double* [[A1]], align 4
9339 // CHECK19-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
9340 // CHECK19-NEXT:    [[TMP13:%.*]] = mul nsw i32 1, [[TMP8]]
9341 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i32 [[TMP13]]
9342 // CHECK19-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
9343 // CHECK19-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
9344 // CHECK19-NEXT:    ret void
9345 //
9346 //
9347 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
9348 // CHECK19-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
9349 // CHECK19-NEXT:  entry:
9350 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9351 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9352 // CHECK19-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
9353 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
9354 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 4
9355 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9356 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9357 // CHECK19-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
9358 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
9359 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9360 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
9361 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
9362 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
9363 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
9364 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
9365 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
9366 // CHECK19-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
9367 // CHECK19-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
9368 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
9369 // CHECK19-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV1]], align 4
9370 // CHECK19-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
9371 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
9372 // CHECK19-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 4
9373 // 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.6*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
9374 // CHECK19-NEXT:    ret void
9375 //
9376 //
9377 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..11
9378 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
9379 // CHECK19-NEXT:  entry:
9380 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9381 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9382 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 4
9383 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
9384 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
9385 // CHECK19-NEXT:    [[AAA:%.*]] = alloca i8, align 1
9386 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9387 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9388 // CHECK19-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 4
9389 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 4
9390 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
9391 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9392 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
9393 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
9394 // CHECK19-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
9395 // CHECK19-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
9396 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
9397 // CHECK19-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
9398 // CHECK19-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
9399 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
9400 // CHECK19-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 4
9401 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
9402 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
9403 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
9404 // CHECK19-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
9405 // CHECK19-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
9406 // CHECK19-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
9407 // CHECK19-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
9408 // CHECK19-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
9409 // CHECK19-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
9410 // CHECK19-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
9411 // CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
9412 // CHECK19-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
9413 // CHECK19-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
9414 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i32 0, i32 2
9415 // CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
9416 // CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
9417 // CHECK19-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
9418 // CHECK19-NEXT:    ret void
9419 //
9420 //
9421 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
9422 // CHECK19-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
9423 // CHECK19-NEXT:  entry:
9424 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9425 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9426 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
9427 // CHECK19-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 4
9428 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9429 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9430 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
9431 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9432 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
9433 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
9434 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
9435 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
9436 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
9437 // CHECK19-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
9438 // CHECK19-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
9439 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
9440 // CHECK19-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 4
9441 // 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.7*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
9442 // CHECK19-NEXT:    ret void
9443 //
9444 //
9445 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..14
9446 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
9447 // CHECK19-NEXT:  entry:
9448 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9449 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9450 // CHECK19-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 4
9451 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
9452 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
9453 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9454 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9455 // CHECK19-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 4
9456 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 4
9457 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
9458 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9459 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
9460 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
9461 // CHECK19-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
9462 // CHECK19-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
9463 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
9464 // CHECK19-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 4
9465 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
9466 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
9467 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
9468 // CHECK19-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
9469 // CHECK19-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
9470 // CHECK19-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
9471 // CHECK19-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
9472 // CHECK19-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
9473 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i32 0, i32 2
9474 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
9475 // CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
9476 // CHECK19-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
9477 // CHECK19-NEXT:    ret void
9478 //
9479 //
9480 // CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
9481 // CHECK19-SAME: () #[[ATTR6:[0-9]+]] {
9482 // CHECK19-NEXT:  entry:
9483 // CHECK19-NEXT:    call void @__tgt_register_requires(i64 1)
9484 // CHECK19-NEXT:    ret void
9485 //
9486 //
9487 // CHECK20-LABEL: define {{[^@]+}}@_Z3fooi
9488 // CHECK20-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
9489 // CHECK20-NEXT:  entry:
9490 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
9491 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
9492 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
9493 // CHECK20-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
9494 // CHECK20-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
9495 // CHECK20-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
9496 // CHECK20-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
9497 // CHECK20-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
9498 // CHECK20-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
9499 // CHECK20-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 1
9500 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
9501 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
9502 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
9503 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
9504 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
9505 // CHECK20-NEXT:    [[A_CASTED2:%.*]] = alloca i32, align 4
9506 // CHECK20-NEXT:    [[AA_CASTED3:%.*]] = alloca i32, align 4
9507 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4
9508 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4
9509 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4
9510 // CHECK20-NEXT:    [[A_CASTED10:%.*]] = alloca i32, align 4
9511 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [9 x i8*], align 4
9512 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS14:%.*]] = alloca [9 x i8*], align 4
9513 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [9 x i8*], align 4
9514 // CHECK20-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 4
9515 // CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
9516 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
9517 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
9518 // CHECK20-NEXT:    store i16 0, i16* [[AA]], align 2
9519 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
9520 // CHECK20-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
9521 // CHECK20-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
9522 // CHECK20-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
9523 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
9524 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
9525 // CHECK20-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
9526 // CHECK20-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
9527 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
9528 // CHECK20-NEXT:    [[TMP5:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 1, i32 20, i32 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
9529 // CHECK20-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to %struct.kmp_task_t_with_privates*
9530 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP6]], i32 0, i32 0
9531 // CHECK20-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP5]])
9532 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
9533 // CHECK20-NEXT:    store i32 [[TMP9]], i32* [[A_CASTED]], align 4
9534 // CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[A_CASTED]], align 4
9535 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104(i32 [[TMP10]]) #[[ATTR3:[0-9]+]]
9536 // CHECK20-NEXT:    [[TMP11:%.*]] = load i16, i16* [[AA]], align 2
9537 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
9538 // CHECK20-NEXT:    store i16 [[TMP11]], i16* [[CONV]], align 2
9539 // CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[AA_CASTED]], align 4
9540 // CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9541 // CHECK20-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
9542 // CHECK20-NEXT:    store i32 [[TMP12]], i32* [[TMP14]], align 4
9543 // CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9544 // CHECK20-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
9545 // CHECK20-NEXT:    store i32 [[TMP12]], i32* [[TMP16]], align 4
9546 // CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
9547 // CHECK20-NEXT:    store i8* null, i8** [[TMP17]], align 4
9548 // CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9549 // CHECK20-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9550 // CHECK20-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110.region_id, i32 1, i8** [[TMP18]], i8** [[TMP19]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
9551 // CHECK20-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
9552 // CHECK20-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
9553 // CHECK20:       omp_offload.failed:
9554 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110(i32 [[TMP12]]) #[[ATTR3]]
9555 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
9556 // CHECK20:       omp_offload.cont:
9557 // CHECK20-NEXT:    [[TMP22:%.*]] = load i32, i32* [[A]], align 4
9558 // CHECK20-NEXT:    store i32 [[TMP22]], i32* [[A_CASTED2]], align 4
9559 // CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A_CASTED2]], align 4
9560 // CHECK20-NEXT:    [[TMP24:%.*]] = load i16, i16* [[AA]], align 2
9561 // CHECK20-NEXT:    [[CONV4:%.*]] = bitcast i32* [[AA_CASTED3]] to i16*
9562 // CHECK20-NEXT:    store i16 [[TMP24]], i16* [[CONV4]], align 2
9563 // CHECK20-NEXT:    [[TMP25:%.*]] = load i32, i32* [[AA_CASTED3]], align 4
9564 // CHECK20-NEXT:    [[TMP26:%.*]] = load i32, i32* [[N_ADDR]], align 4
9565 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP26]], 10
9566 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
9567 // CHECK20:       omp_if.then:
9568 // CHECK20-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
9569 // CHECK20-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
9570 // CHECK20-NEXT:    store i32 [[TMP23]], i32* [[TMP28]], align 4
9571 // CHECK20-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
9572 // CHECK20-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
9573 // CHECK20-NEXT:    store i32 [[TMP23]], i32* [[TMP30]], align 4
9574 // CHECK20-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
9575 // CHECK20-NEXT:    store i8* null, i8** [[TMP31]], align 4
9576 // CHECK20-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
9577 // CHECK20-NEXT:    [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i32*
9578 // CHECK20-NEXT:    store i32 [[TMP25]], i32* [[TMP33]], align 4
9579 // CHECK20-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
9580 // CHECK20-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i32*
9581 // CHECK20-NEXT:    store i32 [[TMP25]], i32* [[TMP35]], align 4
9582 // CHECK20-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
9583 // CHECK20-NEXT:    store i8* null, i8** [[TMP36]], align 4
9584 // CHECK20-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
9585 // CHECK20-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
9586 // CHECK20-NEXT:    [[TMP39:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119.region_id, i32 2, i8** [[TMP37]], i8** [[TMP38]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
9587 // CHECK20-NEXT:    [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
9588 // CHECK20-NEXT:    br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
9589 // CHECK20:       omp_offload.failed8:
9590 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR3]]
9591 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
9592 // CHECK20:       omp_offload.cont9:
9593 // CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
9594 // CHECK20:       omp_if.else:
9595 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119(i32 [[TMP23]], i32 [[TMP25]]) #[[ATTR3]]
9596 // CHECK20-NEXT:    br label [[OMP_IF_END]]
9597 // CHECK20:       omp_if.end:
9598 // CHECK20-NEXT:    [[TMP41:%.*]] = load i32, i32* [[A]], align 4
9599 // CHECK20-NEXT:    store i32 [[TMP41]], i32* [[A_CASTED10]], align 4
9600 // CHECK20-NEXT:    [[TMP42:%.*]] = load i32, i32* [[A_CASTED10]], align 4
9601 // CHECK20-NEXT:    [[TMP43:%.*]] = load i32, i32* [[N_ADDR]], align 4
9602 // CHECK20-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP43]], 20
9603 // CHECK20-NEXT:    br i1 [[CMP11]], label [[OMP_IF_THEN12:%.*]], label [[OMP_IF_ELSE18:%.*]]
9604 // CHECK20:       omp_if.then12:
9605 // CHECK20-NEXT:    [[TMP44:%.*]] = mul nuw i32 [[TMP1]], 4
9606 // CHECK20-NEXT:    [[TMP45:%.*]] = sext i32 [[TMP44]] to i64
9607 // CHECK20-NEXT:    [[TMP46:%.*]] = mul nuw i32 5, [[TMP3]]
9608 // CHECK20-NEXT:    [[TMP47:%.*]] = mul nuw i32 [[TMP46]], 8
9609 // CHECK20-NEXT:    [[TMP48:%.*]] = sext i32 [[TMP47]] to i64
9610 // CHECK20-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
9611 // CHECK20-NEXT:    [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32*
9612 // CHECK20-NEXT:    store i32 [[TMP42]], i32* [[TMP50]], align 4
9613 // CHECK20-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
9614 // CHECK20-NEXT:    [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
9615 // CHECK20-NEXT:    store i32 [[TMP42]], i32* [[TMP52]], align 4
9616 // CHECK20-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
9617 // CHECK20-NEXT:    store i64 4, i64* [[TMP53]], align 4
9618 // CHECK20-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
9619 // CHECK20-NEXT:    store i8* null, i8** [[TMP54]], align 4
9620 // CHECK20-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 1
9621 // CHECK20-NEXT:    [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [10 x float]**
9622 // CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP56]], align 4
9623 // CHECK20-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 1
9624 // CHECK20-NEXT:    [[TMP58:%.*]] = bitcast i8** [[TMP57]] to [10 x float]**
9625 // CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP58]], align 4
9626 // CHECK20-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
9627 // CHECK20-NEXT:    store i64 40, i64* [[TMP59]], align 4
9628 // CHECK20-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 1
9629 // CHECK20-NEXT:    store i8* null, i8** [[TMP60]], align 4
9630 // CHECK20-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 2
9631 // CHECK20-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
9632 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP62]], align 4
9633 // CHECK20-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 2
9634 // CHECK20-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i32*
9635 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP64]], align 4
9636 // CHECK20-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
9637 // CHECK20-NEXT:    store i64 4, i64* [[TMP65]], align 4
9638 // CHECK20-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 2
9639 // CHECK20-NEXT:    store i8* null, i8** [[TMP66]], align 4
9640 // CHECK20-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 3
9641 // CHECK20-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to float**
9642 // CHECK20-NEXT:    store float* [[VLA]], float** [[TMP68]], align 4
9643 // CHECK20-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 3
9644 // CHECK20-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to float**
9645 // CHECK20-NEXT:    store float* [[VLA]], float** [[TMP70]], align 4
9646 // CHECK20-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
9647 // CHECK20-NEXT:    store i64 [[TMP45]], i64* [[TMP71]], align 4
9648 // CHECK20-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 3
9649 // CHECK20-NEXT:    store i8* null, i8** [[TMP72]], align 4
9650 // CHECK20-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 4
9651 // CHECK20-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to [5 x [10 x double]]**
9652 // CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP74]], align 4
9653 // CHECK20-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 4
9654 // CHECK20-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to [5 x [10 x double]]**
9655 // CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP76]], align 4
9656 // CHECK20-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
9657 // CHECK20-NEXT:    store i64 400, i64* [[TMP77]], align 4
9658 // CHECK20-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 4
9659 // CHECK20-NEXT:    store i8* null, i8** [[TMP78]], align 4
9660 // CHECK20-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 5
9661 // CHECK20-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i32*
9662 // CHECK20-NEXT:    store i32 5, i32* [[TMP80]], align 4
9663 // CHECK20-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 5
9664 // CHECK20-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i32*
9665 // CHECK20-NEXT:    store i32 5, i32* [[TMP82]], align 4
9666 // CHECK20-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 5
9667 // CHECK20-NEXT:    store i64 4, i64* [[TMP83]], align 4
9668 // CHECK20-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 5
9669 // CHECK20-NEXT:    store i8* null, i8** [[TMP84]], align 4
9670 // CHECK20-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 6
9671 // CHECK20-NEXT:    [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i32*
9672 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP86]], align 4
9673 // CHECK20-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 6
9674 // CHECK20-NEXT:    [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
9675 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP88]], align 4
9676 // CHECK20-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 6
9677 // CHECK20-NEXT:    store i64 4, i64* [[TMP89]], align 4
9678 // CHECK20-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 6
9679 // CHECK20-NEXT:    store i8* null, i8** [[TMP90]], align 4
9680 // CHECK20-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 7
9681 // CHECK20-NEXT:    [[TMP92:%.*]] = bitcast i8** [[TMP91]] to double**
9682 // CHECK20-NEXT:    store double* [[VLA1]], double** [[TMP92]], align 4
9683 // CHECK20-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 7
9684 // CHECK20-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to double**
9685 // CHECK20-NEXT:    store double* [[VLA1]], double** [[TMP94]], align 4
9686 // CHECK20-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
9687 // CHECK20-NEXT:    store i64 [[TMP48]], i64* [[TMP95]], align 4
9688 // CHECK20-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 7
9689 // CHECK20-NEXT:    store i8* null, i8** [[TMP96]], align 4
9690 // CHECK20-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 8
9691 // CHECK20-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to %struct.TT**
9692 // CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP98]], align 4
9693 // CHECK20-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 8
9694 // CHECK20-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to %struct.TT**
9695 // CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP100]], align 4
9696 // CHECK20-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 8
9697 // CHECK20-NEXT:    store i64 12, i64* [[TMP101]], align 4
9698 // CHECK20-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 8
9699 // CHECK20-NEXT:    store i8* null, i8** [[TMP102]], align 4
9700 // CHECK20-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
9701 // CHECK20-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
9702 // CHECK20-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
9703 // CHECK20-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144.region_id, i32 9, i8** [[TMP103]], i8** [[TMP104]], i64* [[TMP105]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
9704 // CHECK20-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
9705 // CHECK20-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
9706 // CHECK20:       omp_offload.failed16:
9707 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
9708 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
9709 // CHECK20:       omp_offload.cont17:
9710 // CHECK20-NEXT:    br label [[OMP_IF_END19:%.*]]
9711 // CHECK20:       omp_if.else18:
9712 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144(i32 [[TMP42]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
9713 // CHECK20-NEXT:    br label [[OMP_IF_END19]]
9714 // CHECK20:       omp_if.end19:
9715 // CHECK20-NEXT:    [[TMP108:%.*]] = load i32, i32* [[A]], align 4
9716 // CHECK20-NEXT:    [[TMP109:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
9717 // CHECK20-NEXT:    call void @llvm.stackrestore(i8* [[TMP109]])
9718 // CHECK20-NEXT:    ret i32 [[TMP108]]
9719 //
9720 //
9721 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
9722 // CHECK20-SAME: () #[[ATTR2:[0-9]+]] {
9723 // CHECK20-NEXT:  entry:
9724 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
9725 // 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_]])
9726 // CHECK20-NEXT:    ret void
9727 //
9728 //
9729 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined.
9730 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
9731 // CHECK20-NEXT:  entry:
9732 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9733 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9734 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
9735 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9736 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9737 // CHECK20-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
9738 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
9739 // CHECK20-NEXT:    ret void
9740 //
9741 //
9742 // CHECK20-LABEL: define {{[^@]+}}@.omp_task_entry.
9743 // CHECK20-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
9744 // CHECK20-NEXT:  entry:
9745 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
9746 // CHECK20-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
9747 // CHECK20-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
9748 // CHECK20-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
9749 // CHECK20-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
9750 // CHECK20-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 4
9751 // CHECK20-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
9752 // CHECK20-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
9753 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
9754 // CHECK20-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
9755 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
9756 // CHECK20-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
9757 // CHECK20-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
9758 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
9759 // CHECK20-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
9760 // CHECK20-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
9761 // CHECK20-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
9762 // CHECK20-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
9763 // CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
9764 // CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
9765 // CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
9766 // CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
9767 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !22
9768 // CHECK20-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !22
9769 // CHECK20-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
9770 // CHECK20-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
9771 // CHECK20-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !22
9772 // CHECK20-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
9773 // CHECK20-NEXT:    [[TMP10:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
9774 // CHECK20-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0) #[[ATTR3]]
9775 // CHECK20-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
9776 // CHECK20-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
9777 // CHECK20:       omp_offload.failed.i:
9778 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100() #[[ATTR3]]
9779 // CHECK20-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
9780 // CHECK20:       .omp_outlined..1.exit:
9781 // CHECK20-NEXT:    ret i32 0
9782 //
9783 //
9784 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l104
9785 // CHECK20-SAME: (i32 [[A:%.*]]) #[[ATTR2]] {
9786 // CHECK20-NEXT:  entry:
9787 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9788 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
9789 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9790 // CHECK20-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
9791 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
9792 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
9793 // 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..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
9794 // CHECK20-NEXT:    ret void
9795 //
9796 //
9797 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..2
9798 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
9799 // CHECK20-NEXT:  entry:
9800 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9801 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9802 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
9803 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
9804 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9805 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9806 // CHECK20-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
9807 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
9808 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
9809 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9810 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
9811 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
9812 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], 1
9813 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
9814 // CHECK20-NEXT:    ret void
9815 //
9816 //
9817 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
9818 // CHECK20-SAME: (i32 [[AA:%.*]]) #[[ATTR2]] {
9819 // CHECK20-NEXT:  entry:
9820 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9821 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 2
9822 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9823 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9824 // CHECK20-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
9825 // CHECK20-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
9826 // CHECK20-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
9827 // 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..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
9828 // CHECK20-NEXT:    ret void
9829 //
9830 //
9831 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..3
9832 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
9833 // CHECK20-NEXT:  entry:
9834 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9835 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9836 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
9837 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
9838 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9839 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9840 // CHECK20-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
9841 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
9842 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
9843 // CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
9844 // CHECK20-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
9845 // CHECK20-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
9846 // CHECK20-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
9847 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
9848 // CHECK20-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
9849 // CHECK20-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
9850 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9851 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
9852 // CHECK20-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
9853 // CHECK20-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
9854 // CHECK20-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
9855 // CHECK20:       .cancel.exit:
9856 // CHECK20-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
9857 // CHECK20-NEXT:    br label [[DOTCANCEL_CONTINUE]]
9858 // CHECK20:       .cancel.continue:
9859 // CHECK20-NEXT:    ret void
9860 //
9861 //
9862 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
9863 // CHECK20-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR2]] {
9864 // CHECK20-NEXT:  entry:
9865 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9866 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9867 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 4
9868 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9869 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9870 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9871 // CHECK20-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
9872 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
9873 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
9874 // CHECK20-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
9875 // CHECK20-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
9876 // CHECK20-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
9877 // 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..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
9878 // CHECK20-NEXT:    ret void
9879 //
9880 //
9881 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..4
9882 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
9883 // CHECK20-NEXT:  entry:
9884 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9885 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9886 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
9887 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
9888 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
9889 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9890 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9891 // CHECK20-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
9892 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
9893 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
9894 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9895 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
9896 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
9897 // CHECK20-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
9898 // CHECK20-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
9899 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
9900 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
9901 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
9902 // CHECK20-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
9903 // CHECK20-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
9904 // CHECK20-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
9905 // CHECK20-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
9906 // CHECK20-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
9907 // CHECK20-NEXT:    ret void
9908 //
9909 //
9910 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
9911 // CHECK20-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR2]] {
9912 // CHECK20-NEXT:  entry:
9913 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9914 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
9915 // CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
9916 // CHECK20-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
9917 // CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
9918 // CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
9919 // CHECK20-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
9920 // CHECK20-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
9921 // CHECK20-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
9922 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
9923 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9924 // CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
9925 // CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
9926 // CHECK20-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
9927 // CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
9928 // CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
9929 // CHECK20-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
9930 // CHECK20-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
9931 // CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
9932 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
9933 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
9934 // CHECK20-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
9935 // CHECK20-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
9936 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
9937 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
9938 // CHECK20-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
9939 // CHECK20-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
9940 // CHECK20-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
9941 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
9942 // CHECK20-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 4
9943 // CHECK20-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
9944 // CHECK20-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 4
9945 // CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
9946 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP11]], align 4
9947 // CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
9948 // CHECK20-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 4
9949 // CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
9950 // CHECK20-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 4
9951 // CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
9952 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[TMP14]], align 4
9953 // CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
9954 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP15]], align 4
9955 // CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
9956 // CHECK20-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 4
9957 // CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
9958 // CHECK20-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 4
9959 // 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..7 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
9960 // CHECK20-NEXT:    ret void
9961 //
9962 //
9963 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..7
9964 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
9965 // CHECK20-NEXT:  entry:
9966 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9967 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9968 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
9969 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
9970 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9971 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9972 // CHECK20-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
9973 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
9974 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
9975 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9976 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
9977 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
9978 // CHECK20-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 4
9979 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
9980 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
9981 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
9982 // CHECK20-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 4
9983 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
9984 // CHECK20-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 4
9985 // CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 5
9986 // CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
9987 // CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 6
9988 // CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
9989 // CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 7
9990 // CHECK20-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 4
9991 // CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 8
9992 // CHECK20-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 4
9993 // CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
9994 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
9995 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
9996 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i32 0, i32 2
9997 // CHECK20-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
9998 // CHECK20-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
9999 // CHECK20-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
10000 // CHECK20-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
10001 // CHECK20-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
10002 // CHECK20-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i32 3
10003 // CHECK20-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
10004 // CHECK20-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
10005 // CHECK20-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
10006 // CHECK20-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
10007 // CHECK20-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
10008 // CHECK20-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i32 0, i32 1
10009 // CHECK20-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i32 0, i32 2
10010 // CHECK20-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
10011 // CHECK20-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
10012 // CHECK20-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
10013 // CHECK20-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP14]]
10014 // CHECK20-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i32 [[TMP23]]
10015 // CHECK20-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i32 3
10016 // CHECK20-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
10017 // CHECK20-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
10018 // CHECK20-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
10019 // CHECK20-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
10020 // CHECK20-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4
10021 // CHECK20-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
10022 // CHECK20-NEXT:    store i64 [[ADD13]], i64* [[X]], align 4
10023 // CHECK20-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
10024 // CHECK20-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4
10025 // CHECK20-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
10026 // CHECK20-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
10027 // CHECK20-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
10028 // CHECK20-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 4
10029 // CHECK20-NEXT:    ret void
10030 //
10031 //
10032 // CHECK20-LABEL: define {{[^@]+}}@_Z3bari
10033 // CHECK20-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
10034 // CHECK20-NEXT:  entry:
10035 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
10036 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
10037 // CHECK20-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
10038 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
10039 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
10040 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
10041 // CHECK20-NEXT:    [[CALL:%.*]] = call i32 @_Z3fooi(i32 [[TMP0]])
10042 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
10043 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
10044 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
10045 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
10046 // CHECK20-NEXT:    [[CALL1:%.*]] = call i32 @_ZN2S12r1Ei(%struct.S1* nonnull align 4 dereferenceable(8) [[S]], i32 [[TMP2]])
10047 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
10048 // CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
10049 // CHECK20-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
10050 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
10051 // CHECK20-NEXT:    [[CALL3:%.*]] = call i32 @_ZL7fstatici(i32 [[TMP4]])
10052 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
10053 // CHECK20-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
10054 // CHECK20-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
10055 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
10056 // CHECK20-NEXT:    [[CALL5:%.*]] = call i32 @_Z9ftemplateIiET_i(i32 [[TMP6]])
10057 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
10058 // CHECK20-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
10059 // CHECK20-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
10060 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
10061 // CHECK20-NEXT:    ret i32 [[TMP8]]
10062 //
10063 //
10064 // CHECK20-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
10065 // CHECK20-SAME: (%struct.S1* nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 [[N:%.*]]) #[[ATTR0]] comdat align 2 {
10066 // CHECK20-NEXT:  entry:
10067 // CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
10068 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
10069 // CHECK20-NEXT:    [[B:%.*]] = alloca i32, align 4
10070 // CHECK20-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
10071 // CHECK20-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
10072 // CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
10073 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
10074 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
10075 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
10076 // CHECK20-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
10077 // CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
10078 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
10079 // CHECK20-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
10080 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
10081 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
10082 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
10083 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
10084 // CHECK20-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
10085 // CHECK20-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
10086 // CHECK20-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
10087 // CHECK20-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
10088 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
10089 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
10090 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
10091 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
10092 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
10093 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
10094 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
10095 // CHECK20:       omp_if.then:
10096 // CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
10097 // CHECK20-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
10098 // CHECK20-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
10099 // CHECK20-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
10100 // CHECK20-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10101 // CHECK20-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to %struct.S1**
10102 // CHECK20-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP11]], align 4
10103 // CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10104 // CHECK20-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to double**
10105 // CHECK20-NEXT:    store double* [[A]], double** [[TMP13]], align 4
10106 // CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
10107 // CHECK20-NEXT:    store i64 8, i64* [[TMP14]], align 4
10108 // CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
10109 // CHECK20-NEXT:    store i8* null, i8** [[TMP15]], align 4
10110 // CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
10111 // CHECK20-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
10112 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
10113 // CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
10114 // CHECK20-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
10115 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
10116 // CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
10117 // CHECK20-NEXT:    store i64 4, i64* [[TMP20]], align 4
10118 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
10119 // CHECK20-NEXT:    store i8* null, i8** [[TMP21]], align 4
10120 // CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
10121 // CHECK20-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
10122 // CHECK20-NEXT:    store i32 2, i32* [[TMP23]], align 4
10123 // CHECK20-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
10124 // CHECK20-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
10125 // CHECK20-NEXT:    store i32 2, i32* [[TMP25]], align 4
10126 // CHECK20-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
10127 // CHECK20-NEXT:    store i64 4, i64* [[TMP26]], align 4
10128 // CHECK20-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
10129 // CHECK20-NEXT:    store i8* null, i8** [[TMP27]], align 4
10130 // CHECK20-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
10131 // CHECK20-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
10132 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
10133 // CHECK20-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
10134 // CHECK20-NEXT:    [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i32*
10135 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP31]], align 4
10136 // CHECK20-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
10137 // CHECK20-NEXT:    store i64 4, i64* [[TMP32]], align 4
10138 // CHECK20-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
10139 // CHECK20-NEXT:    store i8* null, i8** [[TMP33]], align 4
10140 // CHECK20-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
10141 // CHECK20-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
10142 // CHECK20-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 4
10143 // CHECK20-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
10144 // CHECK20-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
10145 // CHECK20-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 4
10146 // CHECK20-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
10147 // CHECK20-NEXT:    store i64 [[TMP9]], i64* [[TMP38]], align 4
10148 // CHECK20-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
10149 // CHECK20-NEXT:    store i8* null, i8** [[TMP39]], align 4
10150 // CHECK20-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10151 // CHECK20-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10152 // CHECK20-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
10153 // CHECK20-NEXT:    [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP40]], i8** [[TMP41]], i64* [[TMP42]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
10154 // CHECK20-NEXT:    [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
10155 // CHECK20-NEXT:    br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
10156 // CHECK20:       omp_offload.failed:
10157 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
10158 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
10159 // CHECK20:       omp_offload.cont:
10160 // CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
10161 // CHECK20:       omp_if.else:
10162 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
10163 // CHECK20-NEXT:    br label [[OMP_IF_END]]
10164 // CHECK20:       omp_if.end:
10165 // CHECK20-NEXT:    [[TMP45:%.*]] = mul nsw i32 1, [[TMP1]]
10166 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP45]]
10167 // CHECK20-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
10168 // CHECK20-NEXT:    [[TMP46:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
10169 // CHECK20-NEXT:    [[CONV:%.*]] = sext i16 [[TMP46]] to i32
10170 // CHECK20-NEXT:    [[TMP47:%.*]] = load i32, i32* [[B]], align 4
10171 // CHECK20-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP47]]
10172 // CHECK20-NEXT:    [[TMP48:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
10173 // CHECK20-NEXT:    call void @llvm.stackrestore(i8* [[TMP48]])
10174 // CHECK20-NEXT:    ret i32 [[ADD3]]
10175 //
10176 //
10177 // CHECK20-LABEL: define {{[^@]+}}@_ZL7fstatici
10178 // CHECK20-SAME: (i32 [[N:%.*]]) #[[ATTR0]] {
10179 // CHECK20-NEXT:  entry:
10180 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
10181 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
10182 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
10183 // CHECK20-NEXT:    [[AAA:%.*]] = alloca i8, align 1
10184 // CHECK20-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
10185 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
10186 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
10187 // CHECK20-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
10188 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
10189 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
10190 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
10191 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
10192 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
10193 // CHECK20-NEXT:    store i16 0, i16* [[AA]], align 2
10194 // CHECK20-NEXT:    store i8 0, i8* [[AAA]], align 1
10195 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
10196 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
10197 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
10198 // CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
10199 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
10200 // CHECK20-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
10201 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
10202 // CHECK20-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
10203 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
10204 // CHECK20-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
10205 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
10206 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
10207 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
10208 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
10209 // CHECK20:       omp_if.then:
10210 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10211 // CHECK20-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
10212 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
10213 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10214 // CHECK20-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
10215 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
10216 // CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
10217 // CHECK20-NEXT:    store i8* null, i8** [[TMP11]], align 4
10218 // CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
10219 // CHECK20-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
10220 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
10221 // CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
10222 // CHECK20-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
10223 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
10224 // CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
10225 // CHECK20-NEXT:    store i8* null, i8** [[TMP16]], align 4
10226 // CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
10227 // CHECK20-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
10228 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
10229 // CHECK20-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
10230 // CHECK20-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
10231 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
10232 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
10233 // CHECK20-NEXT:    store i8* null, i8** [[TMP21]], align 4
10234 // CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
10235 // CHECK20-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
10236 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
10237 // CHECK20-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
10238 // CHECK20-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
10239 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
10240 // CHECK20-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
10241 // CHECK20-NEXT:    store i8* null, i8** [[TMP26]], align 4
10242 // CHECK20-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10243 // CHECK20-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10244 // CHECK20-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
10245 // CHECK20-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
10246 // CHECK20-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
10247 // CHECK20:       omp_offload.failed:
10248 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
10249 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
10250 // CHECK20:       omp_offload.cont:
10251 // CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
10252 // CHECK20:       omp_if.else:
10253 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
10254 // CHECK20-NEXT:    br label [[OMP_IF_END]]
10255 // CHECK20:       omp_if.end:
10256 // CHECK20-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
10257 // CHECK20-NEXT:    ret i32 [[TMP31]]
10258 //
10259 //
10260 // CHECK20-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
10261 // CHECK20-SAME: (i32 [[N:%.*]]) #[[ATTR0]] comdat {
10262 // CHECK20-NEXT:  entry:
10263 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
10264 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
10265 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
10266 // CHECK20-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
10267 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
10268 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
10269 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
10270 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
10271 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
10272 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
10273 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
10274 // CHECK20-NEXT:    store i16 0, i16* [[AA]], align 2
10275 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
10276 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
10277 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
10278 // CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
10279 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
10280 // CHECK20-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
10281 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
10282 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
10283 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
10284 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
10285 // CHECK20:       omp_if.then:
10286 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10287 // CHECK20-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
10288 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
10289 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10290 // CHECK20-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
10291 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
10292 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
10293 // CHECK20-NEXT:    store i8* null, i8** [[TMP9]], align 4
10294 // CHECK20-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
10295 // CHECK20-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
10296 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
10297 // CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
10298 // CHECK20-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
10299 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
10300 // CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
10301 // CHECK20-NEXT:    store i8* null, i8** [[TMP14]], align 4
10302 // CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
10303 // CHECK20-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
10304 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
10305 // CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
10306 // CHECK20-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
10307 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
10308 // CHECK20-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
10309 // CHECK20-NEXT:    store i8* null, i8** [[TMP19]], align 4
10310 // CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10311 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10312 // CHECK20-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.15, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
10313 // CHECK20-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
10314 // CHECK20-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
10315 // CHECK20:       omp_offload.failed:
10316 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
10317 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
10318 // CHECK20:       omp_offload.cont:
10319 // CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
10320 // CHECK20:       omp_if.else:
10321 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
10322 // CHECK20-NEXT:    br label [[OMP_IF_END]]
10323 // CHECK20:       omp_if.end:
10324 // CHECK20-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
10325 // CHECK20-NEXT:    ret i32 [[TMP24]]
10326 //
10327 //
10328 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
10329 // CHECK20-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
10330 // CHECK20-NEXT:  entry:
10331 // CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
10332 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
10333 // CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
10334 // CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
10335 // CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
10336 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 4
10337 // CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
10338 // CHECK20-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
10339 // CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
10340 // CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
10341 // CHECK20-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
10342 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
10343 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
10344 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
10345 // CHECK20-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
10346 // CHECK20-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
10347 // CHECK20-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 4
10348 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
10349 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_ADDR]], align 4
10350 // CHECK20-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 4
10351 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
10352 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP7]], align 4
10353 // CHECK20-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
10354 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[TMP8]], align 4
10355 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
10356 // CHECK20-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 4
10357 // 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.5*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
10358 // CHECK20-NEXT:    ret void
10359 //
10360 //
10361 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..9
10362 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
10363 // CHECK20-NEXT:  entry:
10364 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
10365 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
10366 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 4
10367 // CHECK20-NEXT:    [[B:%.*]] = alloca i32, align 4
10368 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
10369 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
10370 // CHECK20-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 4
10371 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 4
10372 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
10373 // CHECK20-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
10374 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
10375 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
10376 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
10377 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
10378 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
10379 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
10380 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
10381 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
10382 // CHECK20-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 4
10383 // CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
10384 // CHECK20-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
10385 // CHECK20-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
10386 // CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
10387 // CHECK20-NEXT:    store double [[ADD]], double* [[A]], align 4
10388 // CHECK20-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
10389 // CHECK20-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 4
10390 // CHECK20-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
10391 // CHECK20-NEXT:    store double [[INC]], double* [[A1]], align 4
10392 // CHECK20-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
10393 // CHECK20-NEXT:    [[TMP13:%.*]] = mul nsw i32 1, [[TMP8]]
10394 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i32 [[TMP13]]
10395 // CHECK20-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
10396 // CHECK20-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
10397 // CHECK20-NEXT:    ret void
10398 //
10399 //
10400 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
10401 // CHECK20-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
10402 // CHECK20-NEXT:  entry:
10403 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
10404 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
10405 // CHECK20-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
10406 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
10407 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 4
10408 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
10409 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
10410 // CHECK20-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
10411 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
10412 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
10413 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
10414 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
10415 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
10416 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
10417 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
10418 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
10419 // CHECK20-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
10420 // CHECK20-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
10421 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
10422 // CHECK20-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV1]], align 4
10423 // CHECK20-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
10424 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
10425 // CHECK20-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 4
10426 // 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.6*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
10427 // CHECK20-NEXT:    ret void
10428 //
10429 //
10430 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..11
10431 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
10432 // CHECK20-NEXT:  entry:
10433 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
10434 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
10435 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 4
10436 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
10437 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
10438 // CHECK20-NEXT:    [[AAA:%.*]] = alloca i8, align 1
10439 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
10440 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
10441 // CHECK20-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 4
10442 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 4
10443 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
10444 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
10445 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
10446 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
10447 // CHECK20-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
10448 // CHECK20-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
10449 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
10450 // CHECK20-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
10451 // CHECK20-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
10452 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
10453 // CHECK20-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 4
10454 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
10455 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
10456 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
10457 // CHECK20-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
10458 // CHECK20-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
10459 // CHECK20-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
10460 // CHECK20-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
10461 // CHECK20-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
10462 // CHECK20-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
10463 // CHECK20-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
10464 // CHECK20-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
10465 // CHECK20-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
10466 // CHECK20-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
10467 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i32 0, i32 2
10468 // CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
10469 // CHECK20-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
10470 // CHECK20-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
10471 // CHECK20-NEXT:    ret void
10472 //
10473 //
10474 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
10475 // CHECK20-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
10476 // CHECK20-NEXT:  entry:
10477 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
10478 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
10479 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
10480 // CHECK20-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 4
10481 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
10482 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
10483 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
10484 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
10485 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
10486 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
10487 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
10488 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
10489 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
10490 // CHECK20-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
10491 // CHECK20-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
10492 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
10493 // CHECK20-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 4
10494 // 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.7*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
10495 // CHECK20-NEXT:    ret void
10496 //
10497 //
10498 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..14
10499 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] {
10500 // CHECK20-NEXT:  entry:
10501 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
10502 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
10503 // CHECK20-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 4
10504 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
10505 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
10506 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
10507 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
10508 // CHECK20-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 4
10509 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 4
10510 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
10511 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
10512 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
10513 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
10514 // CHECK20-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
10515 // CHECK20-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
10516 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
10517 // CHECK20-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 4
10518 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
10519 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
10520 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
10521 // CHECK20-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
10522 // CHECK20-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
10523 // CHECK20-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
10524 // CHECK20-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
10525 // CHECK20-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
10526 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i32 0, i32 2
10527 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
10528 // CHECK20-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
10529 // CHECK20-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
10530 // CHECK20-NEXT:    ret void
10531 //
10532 //
10533 // CHECK20-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
10534 // CHECK20-SAME: () #[[ATTR6:[0-9]+]] {
10535 // CHECK20-NEXT:  entry:
10536 // CHECK20-NEXT:    call void @__tgt_register_requires(i64 1)
10537 // CHECK20-NEXT:    ret void
10538 //
10539 //
10540 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
10541 // CHECK25-SAME: () #[[ATTR0:[0-9]+]] {
10542 // CHECK25-NEXT:  entry:
10543 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
10544 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
10545 // CHECK25-NEXT:    ret void
10546 //
10547 //
10548 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined.
10549 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
10550 // CHECK25-NEXT:  entry:
10551 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10552 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10553 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
10554 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10555 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10556 // CHECK25-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
10557 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
10558 // CHECK25-NEXT:    ret void
10559 //
10560 //
10561 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
10562 // CHECK25-SAME: (i64 [[AA:%.*]]) #[[ATTR0]] {
10563 // CHECK25-NEXT:  entry:
10564 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10565 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 2
10566 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10567 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10568 // CHECK25-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
10569 // CHECK25-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
10570 // CHECK25-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
10571 // 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_]])
10572 // CHECK25-NEXT:    ret void
10573 //
10574 //
10575 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..1
10576 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
10577 // CHECK25-NEXT:  entry:
10578 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10579 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10580 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
10581 // CHECK25-NEXT:    [[AA:%.*]] = alloca i16, align 2
10582 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10583 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10584 // CHECK25-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
10585 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
10586 // CHECK25-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
10587 // CHECK25-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
10588 // CHECK25-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
10589 // CHECK25-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
10590 // CHECK25-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
10591 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
10592 // CHECK25-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
10593 // CHECK25-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
10594 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10595 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
10596 // CHECK25-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
10597 // CHECK25-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
10598 // CHECK25-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
10599 // CHECK25:       .cancel.exit:
10600 // CHECK25-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
10601 // CHECK25-NEXT:    br label [[DOTCANCEL_CONTINUE]]
10602 // CHECK25:       .cancel.continue:
10603 // CHECK25-NEXT:    ret void
10604 //
10605 //
10606 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
10607 // CHECK25-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
10608 // CHECK25-NEXT:  entry:
10609 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10610 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10611 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
10612 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10613 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10614 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10615 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10616 // CHECK25-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
10617 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
10618 // CHECK25-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
10619 // CHECK25-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
10620 // CHECK25-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
10621 // CHECK25-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
10622 // 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_]])
10623 // CHECK25-NEXT:    ret void
10624 //
10625 //
10626 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..2
10627 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
10628 // CHECK25-NEXT:  entry:
10629 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10630 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10631 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
10632 // CHECK25-NEXT:    [[A:%.*]] = alloca i32, align 4
10633 // CHECK25-NEXT:    [[AA:%.*]] = alloca i16, align 2
10634 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10635 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10636 // CHECK25-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
10637 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
10638 // CHECK25-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
10639 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
10640 // CHECK25-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
10641 // CHECK25-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
10642 // CHECK25-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
10643 // CHECK25-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
10644 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
10645 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
10646 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
10647 // CHECK25-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
10648 // CHECK25-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
10649 // CHECK25-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
10650 // CHECK25-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
10651 // CHECK25-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
10652 // CHECK25-NEXT:    ret void
10653 //
10654 //
10655 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
10656 // CHECK25-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR0]] {
10657 // CHECK25-NEXT:  entry:
10658 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10659 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
10660 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
10661 // CHECK25-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
10662 // CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
10663 // CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
10664 // CHECK25-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
10665 // CHECK25-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
10666 // CHECK25-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
10667 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
10668 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10669 // CHECK25-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
10670 // CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10671 // CHECK25-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
10672 // CHECK25-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
10673 // CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
10674 // CHECK25-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
10675 // CHECK25-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
10676 // CHECK25-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
10677 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10678 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
10679 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10680 // CHECK25-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
10681 // CHECK25-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
10682 // CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
10683 // CHECK25-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
10684 // CHECK25-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
10685 // CHECK25-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
10686 // CHECK25-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
10687 // CHECK25-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
10688 // CHECK25-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 8
10689 // CHECK25-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
10690 // CHECK25-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 8
10691 // CHECK25-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
10692 // CHECK25-NEXT:    store i64 [[TMP1]], i64* [[TMP11]], align 8
10693 // CHECK25-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
10694 // CHECK25-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 8
10695 // CHECK25-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
10696 // CHECK25-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 8
10697 // CHECK25-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
10698 // CHECK25-NEXT:    store i64 [[TMP4]], i64* [[TMP14]], align 8
10699 // CHECK25-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
10700 // CHECK25-NEXT:    store i64 [[TMP5]], i64* [[TMP15]], align 8
10701 // CHECK25-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
10702 // CHECK25-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 8
10703 // CHECK25-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
10704 // CHECK25-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 8
10705 // 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_]])
10706 // CHECK25-NEXT:    ret void
10707 //
10708 //
10709 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..3
10710 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
10711 // CHECK25-NEXT:  entry:
10712 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10713 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10714 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
10715 // CHECK25-NEXT:    [[A:%.*]] = alloca i32, align 4
10716 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10717 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10718 // CHECK25-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
10719 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
10720 // CHECK25-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
10721 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
10722 // CHECK25-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
10723 // CHECK25-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
10724 // CHECK25-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 8
10725 // CHECK25-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
10726 // CHECK25-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
10727 // CHECK25-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
10728 // CHECK25-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 8
10729 // CHECK25-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 4
10730 // CHECK25-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 8
10731 // CHECK25-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 5
10732 // CHECK25-NEXT:    [[TMP12:%.*]] = load i64, i64* [[TMP11]], align 8
10733 // CHECK25-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 6
10734 // CHECK25-NEXT:    [[TMP14:%.*]] = load i64, i64* [[TMP13]], align 8
10735 // CHECK25-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 7
10736 // CHECK25-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 8
10737 // CHECK25-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 8
10738 // CHECK25-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 8
10739 // CHECK25-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
10740 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
10741 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
10742 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i64 0, i64 2
10743 // CHECK25-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
10744 // CHECK25-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
10745 // CHECK25-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
10746 // CHECK25-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
10747 // CHECK25-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
10748 // CHECK25-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 3
10749 // CHECK25-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
10750 // CHECK25-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
10751 // CHECK25-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
10752 // CHECK25-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
10753 // CHECK25-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
10754 // CHECK25-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i64 0, i64 1
10755 // CHECK25-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i64 0, i64 2
10756 // CHECK25-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
10757 // CHECK25-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
10758 // CHECK25-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
10759 // CHECK25-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP14]]
10760 // CHECK25-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i64 [[TMP23]]
10761 // CHECK25-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i64 3
10762 // CHECK25-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
10763 // CHECK25-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
10764 // CHECK25-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
10765 // CHECK25-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
10766 // CHECK25-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8
10767 // CHECK25-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
10768 // CHECK25-NEXT:    store i64 [[ADD13]], i64* [[X]], align 8
10769 // CHECK25-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
10770 // CHECK25-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8
10771 // CHECK25-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
10772 // CHECK25-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
10773 // CHECK25-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
10774 // CHECK25-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 8
10775 // CHECK25-NEXT:    ret void
10776 //
10777 //
10778 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
10779 // CHECK25-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
10780 // CHECK25-NEXT:  entry:
10781 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10782 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10783 // CHECK25-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
10784 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
10785 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 8
10786 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10787 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10788 // CHECK25-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
10789 // CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
10790 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10791 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10792 // CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
10793 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
10794 // CHECK25-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
10795 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
10796 // CHECK25-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
10797 // CHECK25-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
10798 // CHECK25-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
10799 // CHECK25-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
10800 // CHECK25-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
10801 // CHECK25-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV2]], align 8
10802 // CHECK25-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
10803 // CHECK25-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
10804 // CHECK25-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 8
10805 // 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_]])
10806 // CHECK25-NEXT:    ret void
10807 //
10808 //
10809 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..4
10810 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
10811 // CHECK25-NEXT:  entry:
10812 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10813 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10814 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
10815 // CHECK25-NEXT:    [[A:%.*]] = alloca i32, align 4
10816 // CHECK25-NEXT:    [[AA:%.*]] = alloca i16, align 2
10817 // CHECK25-NEXT:    [[AAA:%.*]] = alloca i8, align 1
10818 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10819 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10820 // CHECK25-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
10821 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
10822 // CHECK25-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
10823 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
10824 // CHECK25-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
10825 // CHECK25-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
10826 // CHECK25-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
10827 // CHECK25-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
10828 // CHECK25-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
10829 // CHECK25-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
10830 // CHECK25-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
10831 // CHECK25-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
10832 // CHECK25-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 8
10833 // CHECK25-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
10834 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
10835 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
10836 // CHECK25-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
10837 // CHECK25-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
10838 // CHECK25-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
10839 // CHECK25-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
10840 // CHECK25-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
10841 // CHECK25-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
10842 // CHECK25-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
10843 // CHECK25-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
10844 // CHECK25-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
10845 // CHECK25-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
10846 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i64 0, i64 2
10847 // CHECK25-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
10848 // CHECK25-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
10849 // CHECK25-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
10850 // CHECK25-NEXT:    ret void
10851 //
10852 //
10853 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
10854 // CHECK25-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
10855 // CHECK25-NEXT:  entry:
10856 // CHECK25-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
10857 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
10858 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
10859 // CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
10860 // CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
10861 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
10862 // CHECK25-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
10863 // CHECK25-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
10864 // CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10865 // CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
10866 // CHECK25-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
10867 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
10868 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
10869 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10870 // CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
10871 // CHECK25-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
10872 // CHECK25-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
10873 // CHECK25-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 8
10874 // CHECK25-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
10875 // CHECK25-NEXT:    [[TMP6:%.*]] = load i32, i32* [[CONV]], align 8
10876 // CHECK25-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
10877 // CHECK25-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
10878 // CHECK25-NEXT:    store i64 [[TMP1]], i64* [[TMP7]], align 8
10879 // CHECK25-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
10880 // CHECK25-NEXT:    store i64 [[TMP2]], i64* [[TMP8]], align 8
10881 // CHECK25-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
10882 // CHECK25-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 8
10883 // 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_]])
10884 // CHECK25-NEXT:    ret void
10885 //
10886 //
10887 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..5
10888 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
10889 // CHECK25-NEXT:  entry:
10890 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10891 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10892 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
10893 // CHECK25-NEXT:    [[B:%.*]] = alloca i32, align 4
10894 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10895 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10896 // CHECK25-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
10897 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
10898 // CHECK25-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
10899 // CHECK25-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
10900 // CHECK25-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
10901 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
10902 // CHECK25-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
10903 // CHECK25-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
10904 // CHECK25-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
10905 // CHECK25-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
10906 // CHECK25-NEXT:    [[TMP8:%.*]] = load i64, i64* [[TMP7]], align 8
10907 // CHECK25-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
10908 // CHECK25-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 8
10909 // CHECK25-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
10910 // CHECK25-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
10911 // CHECK25-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
10912 // CHECK25-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
10913 // CHECK25-NEXT:    store double [[ADD]], double* [[A]], align 8
10914 // CHECK25-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
10915 // CHECK25-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 8
10916 // CHECK25-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
10917 // CHECK25-NEXT:    store double [[INC]], double* [[A1]], align 8
10918 // CHECK25-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
10919 // CHECK25-NEXT:    [[TMP13:%.*]] = mul nsw i64 1, [[TMP8]]
10920 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i64 [[TMP13]]
10921 // CHECK25-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
10922 // CHECK25-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
10923 // CHECK25-NEXT:    ret void
10924 //
10925 //
10926 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
10927 // CHECK25-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
10928 // CHECK25-NEXT:  entry:
10929 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10930 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10931 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
10932 // CHECK25-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
10933 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10934 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10935 // CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
10936 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10937 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10938 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
10939 // CHECK25-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
10940 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
10941 // CHECK25-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
10942 // CHECK25-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
10943 // CHECK25-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
10944 // CHECK25-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
10945 // CHECK25-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
10946 // CHECK25-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 8
10947 // 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.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
10948 // CHECK25-NEXT:    ret void
10949 //
10950 //
10951 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..6
10952 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
10953 // CHECK25-NEXT:  entry:
10954 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10955 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10956 // CHECK25-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
10957 // CHECK25-NEXT:    [[A:%.*]] = alloca i32, align 4
10958 // CHECK25-NEXT:    [[AA:%.*]] = alloca i16, align 2
10959 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10960 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10961 // CHECK25-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
10962 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
10963 // CHECK25-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
10964 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
10965 // CHECK25-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
10966 // CHECK25-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
10967 // CHECK25-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
10968 // CHECK25-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
10969 // CHECK25-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
10970 // CHECK25-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 8
10971 // CHECK25-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
10972 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
10973 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
10974 // CHECK25-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
10975 // CHECK25-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
10976 // CHECK25-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
10977 // CHECK25-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
10978 // CHECK25-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
10979 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i64 0, i64 2
10980 // CHECK25-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
10981 // CHECK25-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
10982 // CHECK25-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
10983 // CHECK25-NEXT:    ret void
10984 //
10985 //
10986 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
10987 // CHECK26-SAME: () #[[ATTR0:[0-9]+]] {
10988 // CHECK26-NEXT:  entry:
10989 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
10990 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
10991 // CHECK26-NEXT:    ret void
10992 //
10993 //
10994 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined.
10995 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
10996 // CHECK26-NEXT:  entry:
10997 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10998 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10999 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
11000 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11001 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11002 // CHECK26-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
11003 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
11004 // CHECK26-NEXT:    ret void
11005 //
11006 //
11007 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
11008 // CHECK26-SAME: (i64 [[AA:%.*]]) #[[ATTR0]] {
11009 // CHECK26-NEXT:  entry:
11010 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
11011 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 2
11012 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11013 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
11014 // CHECK26-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11015 // CHECK26-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 8
11016 // CHECK26-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
11017 // 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_]])
11018 // CHECK26-NEXT:    ret void
11019 //
11020 //
11021 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..1
11022 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11023 // CHECK26-NEXT:  entry:
11024 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11025 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11026 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
11027 // CHECK26-NEXT:    [[AA:%.*]] = alloca i16, align 2
11028 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11029 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11030 // CHECK26-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
11031 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
11032 // CHECK26-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
11033 // CHECK26-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
11034 // CHECK26-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
11035 // CHECK26-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
11036 // CHECK26-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
11037 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
11038 // CHECK26-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
11039 // CHECK26-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
11040 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11041 // CHECK26-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
11042 // CHECK26-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
11043 // CHECK26-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
11044 // CHECK26-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
11045 // CHECK26:       .cancel.exit:
11046 // CHECK26-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
11047 // CHECK26-NEXT:    br label [[DOTCANCEL_CONTINUE]]
11048 // CHECK26:       .cancel.continue:
11049 // CHECK26-NEXT:    ret void
11050 //
11051 //
11052 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
11053 // CHECK26-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]]) #[[ATTR0]] {
11054 // CHECK26-NEXT:  entry:
11055 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11056 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
11057 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
11058 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11059 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11060 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11061 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
11062 // CHECK26-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11063 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
11064 // CHECK26-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
11065 // CHECK26-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
11066 // CHECK26-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 8
11067 // CHECK26-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
11068 // 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_]])
11069 // CHECK26-NEXT:    ret void
11070 //
11071 //
11072 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..2
11073 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11074 // CHECK26-NEXT:  entry:
11075 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11076 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11077 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
11078 // CHECK26-NEXT:    [[A:%.*]] = alloca i32, align 4
11079 // CHECK26-NEXT:    [[AA:%.*]] = alloca i16, align 2
11080 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11081 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11082 // CHECK26-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
11083 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
11084 // CHECK26-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
11085 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
11086 // CHECK26-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
11087 // CHECK26-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
11088 // CHECK26-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
11089 // CHECK26-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
11090 // CHECK26-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
11091 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
11092 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
11093 // CHECK26-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
11094 // CHECK26-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
11095 // CHECK26-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
11096 // CHECK26-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
11097 // CHECK26-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
11098 // CHECK26-NEXT:    ret void
11099 //
11100 //
11101 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
11102 // CHECK26-SAME: (i64 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i64 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 8 dereferenceable(400) [[C:%.*]], i64 [[VLA1:%.*]], i64 [[VLA3:%.*]], double* nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 8 dereferenceable(16) [[D:%.*]]) #[[ATTR0]] {
11103 // CHECK26-NEXT:  entry:
11104 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11105 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
11106 // CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
11107 // CHECK26-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
11108 // CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
11109 // CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
11110 // CHECK26-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
11111 // CHECK26-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
11112 // CHECK26-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
11113 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
11114 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11115 // CHECK26-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
11116 // CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
11117 // CHECK26-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
11118 // CHECK26-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
11119 // CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
11120 // CHECK26-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
11121 // CHECK26-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
11122 // CHECK26-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
11123 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11124 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
11125 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
11126 // CHECK26-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
11127 // CHECK26-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
11128 // CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
11129 // CHECK26-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
11130 // CHECK26-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
11131 // CHECK26-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
11132 // CHECK26-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11133 // CHECK26-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
11134 // CHECK26-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 8
11135 // CHECK26-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
11136 // CHECK26-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 8
11137 // CHECK26-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
11138 // CHECK26-NEXT:    store i64 [[TMP1]], i64* [[TMP11]], align 8
11139 // CHECK26-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
11140 // CHECK26-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 8
11141 // CHECK26-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
11142 // CHECK26-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 8
11143 // CHECK26-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
11144 // CHECK26-NEXT:    store i64 [[TMP4]], i64* [[TMP14]], align 8
11145 // CHECK26-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
11146 // CHECK26-NEXT:    store i64 [[TMP5]], i64* [[TMP15]], align 8
11147 // CHECK26-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
11148 // CHECK26-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 8
11149 // CHECK26-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
11150 // CHECK26-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 8
11151 // 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_]])
11152 // CHECK26-NEXT:    ret void
11153 //
11154 //
11155 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..3
11156 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11157 // CHECK26-NEXT:  entry:
11158 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11159 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11160 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
11161 // CHECK26-NEXT:    [[A:%.*]] = alloca i32, align 4
11162 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11163 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11164 // CHECK26-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
11165 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
11166 // CHECK26-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
11167 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
11168 // CHECK26-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
11169 // CHECK26-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
11170 // CHECK26-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 8
11171 // CHECK26-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
11172 // CHECK26-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
11173 // CHECK26-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
11174 // CHECK26-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 8
11175 // CHECK26-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 4
11176 // CHECK26-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 8
11177 // CHECK26-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 5
11178 // CHECK26-NEXT:    [[TMP12:%.*]] = load i64, i64* [[TMP11]], align 8
11179 // CHECK26-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 6
11180 // CHECK26-NEXT:    [[TMP14:%.*]] = load i64, i64* [[TMP13]], align 8
11181 // CHECK26-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 7
11182 // CHECK26-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 8
11183 // CHECK26-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 8
11184 // CHECK26-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 8
11185 // CHECK26-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
11186 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
11187 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
11188 // CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i64 0, i64 2
11189 // CHECK26-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
11190 // CHECK26-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
11191 // CHECK26-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
11192 // CHECK26-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
11193 // CHECK26-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
11194 // CHECK26-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 3
11195 // CHECK26-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
11196 // CHECK26-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
11197 // CHECK26-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
11198 // CHECK26-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
11199 // CHECK26-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
11200 // CHECK26-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i64 0, i64 1
11201 // CHECK26-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i64 0, i64 2
11202 // CHECK26-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
11203 // CHECK26-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
11204 // CHECK26-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
11205 // CHECK26-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP14]]
11206 // CHECK26-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i64 [[TMP23]]
11207 // CHECK26-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i64 3
11208 // CHECK26-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
11209 // CHECK26-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
11210 // CHECK26-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
11211 // CHECK26-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
11212 // CHECK26-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8
11213 // CHECK26-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
11214 // CHECK26-NEXT:    store i64 [[ADD13]], i64* [[X]], align 8
11215 // CHECK26-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
11216 // CHECK26-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8
11217 // CHECK26-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
11218 // CHECK26-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
11219 // CHECK26-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
11220 // CHECK26-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 8
11221 // CHECK26-NEXT:    ret void
11222 //
11223 //
11224 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
11225 // CHECK26-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], i64 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
11226 // CHECK26-NEXT:  entry:
11227 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11228 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
11229 // CHECK26-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
11230 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
11231 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 8
11232 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11233 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11234 // CHECK26-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
11235 // CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
11236 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11237 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
11238 // CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
11239 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
11240 // CHECK26-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11241 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
11242 // CHECK26-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
11243 // CHECK26-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
11244 // CHECK26-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
11245 // CHECK26-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
11246 // CHECK26-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
11247 // CHECK26-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV2]], align 8
11248 // CHECK26-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
11249 // CHECK26-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
11250 // CHECK26-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 8
11251 // 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_]])
11252 // CHECK26-NEXT:    ret void
11253 //
11254 //
11255 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..4
11256 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11257 // CHECK26-NEXT:  entry:
11258 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11259 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11260 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
11261 // CHECK26-NEXT:    [[A:%.*]] = alloca i32, align 4
11262 // CHECK26-NEXT:    [[AA:%.*]] = alloca i16, align 2
11263 // CHECK26-NEXT:    [[AAA:%.*]] = alloca i8, align 1
11264 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11265 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11266 // CHECK26-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
11267 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
11268 // CHECK26-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
11269 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
11270 // CHECK26-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
11271 // CHECK26-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
11272 // CHECK26-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
11273 // CHECK26-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
11274 // CHECK26-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
11275 // CHECK26-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
11276 // CHECK26-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
11277 // CHECK26-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
11278 // CHECK26-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 8
11279 // CHECK26-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
11280 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
11281 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
11282 // CHECK26-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
11283 // CHECK26-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
11284 // CHECK26-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
11285 // CHECK26-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
11286 // CHECK26-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
11287 // CHECK26-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
11288 // CHECK26-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
11289 // CHECK26-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
11290 // CHECK26-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
11291 // CHECK26-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
11292 // CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i64 0, i64 2
11293 // CHECK26-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
11294 // CHECK26-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
11295 // CHECK26-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
11296 // CHECK26-NEXT:    ret void
11297 //
11298 //
11299 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
11300 // CHECK26-SAME: (%struct.S1* [[THIS:%.*]], i64 [[B:%.*]], i64 [[VLA:%.*]], i64 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
11301 // CHECK26-NEXT:  entry:
11302 // CHECK26-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
11303 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
11304 // CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
11305 // CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
11306 // CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
11307 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
11308 // CHECK26-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
11309 // CHECK26-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
11310 // CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
11311 // CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
11312 // CHECK26-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
11313 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
11314 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
11315 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
11316 // CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
11317 // CHECK26-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
11318 // CHECK26-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11319 // CHECK26-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 8
11320 // CHECK26-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
11321 // CHECK26-NEXT:    [[TMP6:%.*]] = load i32, i32* [[CONV]], align 8
11322 // CHECK26-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
11323 // CHECK26-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
11324 // CHECK26-NEXT:    store i64 [[TMP1]], i64* [[TMP7]], align 8
11325 // CHECK26-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
11326 // CHECK26-NEXT:    store i64 [[TMP2]], i64* [[TMP8]], align 8
11327 // CHECK26-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
11328 // CHECK26-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 8
11329 // 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_]])
11330 // CHECK26-NEXT:    ret void
11331 //
11332 //
11333 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..5
11334 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11335 // CHECK26-NEXT:  entry:
11336 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11337 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11338 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
11339 // CHECK26-NEXT:    [[B:%.*]] = alloca i32, align 4
11340 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11341 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11342 // CHECK26-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
11343 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
11344 // CHECK26-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
11345 // CHECK26-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 8
11346 // CHECK26-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
11347 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 8
11348 // CHECK26-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
11349 // CHECK26-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
11350 // CHECK26-NEXT:    [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
11351 // CHECK26-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
11352 // CHECK26-NEXT:    [[TMP8:%.*]] = load i64, i64* [[TMP7]], align 8
11353 // CHECK26-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
11354 // CHECK26-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 8
11355 // CHECK26-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
11356 // CHECK26-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
11357 // CHECK26-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
11358 // CHECK26-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
11359 // CHECK26-NEXT:    store double [[ADD]], double* [[A]], align 8
11360 // CHECK26-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
11361 // CHECK26-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 8
11362 // CHECK26-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
11363 // CHECK26-NEXT:    store double [[INC]], double* [[A1]], align 8
11364 // CHECK26-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
11365 // CHECK26-NEXT:    [[TMP13:%.*]] = mul nsw i64 1, [[TMP8]]
11366 // CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i64 [[TMP13]]
11367 // CHECK26-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
11368 // CHECK26-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
11369 // CHECK26-NEXT:    ret void
11370 //
11371 //
11372 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
11373 // CHECK26-SAME: (i64 [[A:%.*]], i64 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
11374 // CHECK26-NEXT:  entry:
11375 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11376 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
11377 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
11378 // CHECK26-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
11379 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11380 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11381 // CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
11382 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11383 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
11384 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
11385 // CHECK26-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11386 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
11387 // CHECK26-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 8
11388 // CHECK26-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
11389 // CHECK26-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV1]], align 8
11390 // CHECK26-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
11391 // CHECK26-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
11392 // CHECK26-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 8
11393 // 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.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
11394 // CHECK26-NEXT:    ret void
11395 //
11396 //
11397 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..6
11398 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11399 // CHECK26-NEXT:  entry:
11400 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11401 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11402 // CHECK26-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
11403 // CHECK26-NEXT:    [[A:%.*]] = alloca i32, align 4
11404 // CHECK26-NEXT:    [[AA:%.*]] = alloca i16, align 2
11405 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11406 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11407 // CHECK26-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
11408 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
11409 // CHECK26-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
11410 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 8
11411 // CHECK26-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
11412 // CHECK26-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
11413 // CHECK26-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
11414 // CHECK26-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
11415 // CHECK26-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
11416 // CHECK26-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 8
11417 // CHECK26-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
11418 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
11419 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
11420 // CHECK26-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
11421 // CHECK26-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
11422 // CHECK26-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
11423 // CHECK26-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
11424 // CHECK26-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
11425 // CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i64 0, i64 2
11426 // CHECK26-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
11427 // CHECK26-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
11428 // CHECK26-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
11429 // CHECK26-NEXT:    ret void
11430 //
11431 //
11432 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
11433 // CHECK27-SAME: () #[[ATTR0:[0-9]+]] {
11434 // CHECK27-NEXT:  entry:
11435 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
11436 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
11437 // CHECK27-NEXT:    ret void
11438 //
11439 //
11440 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined.
11441 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11442 // CHECK27-NEXT:  entry:
11443 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11444 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11445 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
11446 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11447 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11448 // CHECK27-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
11449 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
11450 // CHECK27-NEXT:    ret void
11451 //
11452 //
11453 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
11454 // CHECK27-SAME: (i32 [[AA:%.*]]) #[[ATTR0]] {
11455 // CHECK27-NEXT:  entry:
11456 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11457 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 2
11458 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11459 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11460 // CHECK27-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11461 // CHECK27-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
11462 // CHECK27-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
11463 // 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_]])
11464 // CHECK27-NEXT:    ret void
11465 //
11466 //
11467 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..1
11468 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11469 // CHECK27-NEXT:  entry:
11470 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11471 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11472 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4
11473 // CHECK27-NEXT:    [[AA:%.*]] = alloca i16, align 2
11474 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11475 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11476 // CHECK27-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4
11477 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4
11478 // CHECK27-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
11479 // CHECK27-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
11480 // CHECK27-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
11481 // CHECK27-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
11482 // CHECK27-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
11483 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
11484 // CHECK27-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
11485 // CHECK27-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
11486 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
11487 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
11488 // CHECK27-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
11489 // CHECK27-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
11490 // CHECK27-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
11491 // CHECK27:       .cancel.exit:
11492 // CHECK27-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
11493 // CHECK27-NEXT:    br label [[DOTCANCEL_CONTINUE]]
11494 // CHECK27:       .cancel.continue:
11495 // CHECK27-NEXT:    ret void
11496 //
11497 //
11498 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
11499 // CHECK27-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
11500 // CHECK27-NEXT:  entry:
11501 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11502 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11503 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
11504 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11505 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11506 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11507 // CHECK27-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11508 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
11509 // CHECK27-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
11510 // CHECK27-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
11511 // CHECK27-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
11512 // CHECK27-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
11513 // 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_]])
11514 // CHECK27-NEXT:    ret void
11515 //
11516 //
11517 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..2
11518 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11519 // CHECK27-NEXT:  entry:
11520 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11521 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11522 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
11523 // CHECK27-NEXT:    [[A:%.*]] = alloca i32, align 4
11524 // CHECK27-NEXT:    [[AA:%.*]] = alloca i16, align 2
11525 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11526 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11527 // CHECK27-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
11528 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
11529 // CHECK27-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
11530 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
11531 // CHECK27-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
11532 // CHECK27-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
11533 // CHECK27-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
11534 // CHECK27-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
11535 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
11536 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
11537 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
11538 // CHECK27-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
11539 // CHECK27-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
11540 // CHECK27-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
11541 // CHECK27-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
11542 // CHECK27-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
11543 // CHECK27-NEXT:    ret void
11544 //
11545 //
11546 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
11547 // CHECK27-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR0]] {
11548 // CHECK27-NEXT:  entry:
11549 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11550 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
11551 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11552 // CHECK27-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
11553 // CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
11554 // CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
11555 // CHECK27-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
11556 // CHECK27-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
11557 // CHECK27-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
11558 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 4
11559 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11560 // CHECK27-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
11561 // CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
11562 // CHECK27-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
11563 // CHECK27-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
11564 // CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
11565 // CHECK27-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
11566 // CHECK27-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
11567 // CHECK27-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
11568 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
11569 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
11570 // CHECK27-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
11571 // CHECK27-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
11572 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
11573 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
11574 // CHECK27-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
11575 // CHECK27-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
11576 // CHECK27-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11577 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
11578 // CHECK27-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 4
11579 // CHECK27-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
11580 // CHECK27-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 4
11581 // CHECK27-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
11582 // CHECK27-NEXT:    store i32 [[TMP1]], i32* [[TMP11]], align 4
11583 // CHECK27-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
11584 // CHECK27-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 4
11585 // CHECK27-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
11586 // CHECK27-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 4
11587 // CHECK27-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
11588 // CHECK27-NEXT:    store i32 [[TMP4]], i32* [[TMP14]], align 4
11589 // CHECK27-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
11590 // CHECK27-NEXT:    store i32 [[TMP5]], i32* [[TMP15]], align 4
11591 // CHECK27-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
11592 // CHECK27-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 4
11593 // CHECK27-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
11594 // CHECK27-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 4
11595 // 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_]])
11596 // CHECK27-NEXT:    ret void
11597 //
11598 //
11599 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..3
11600 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11601 // CHECK27-NEXT:  entry:
11602 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11603 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11604 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
11605 // CHECK27-NEXT:    [[A:%.*]] = alloca i32, align 4
11606 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11607 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11608 // CHECK27-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
11609 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
11610 // CHECK27-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
11611 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
11612 // CHECK27-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
11613 // CHECK27-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
11614 // CHECK27-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 4
11615 // CHECK27-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
11616 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
11617 // CHECK27-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
11618 // CHECK27-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 4
11619 // CHECK27-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 4
11620 // CHECK27-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 4
11621 // CHECK27-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 5
11622 // CHECK27-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
11623 // CHECK27-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 6
11624 // CHECK27-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
11625 // CHECK27-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 7
11626 // CHECK27-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 4
11627 // CHECK27-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 8
11628 // CHECK27-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 4
11629 // CHECK27-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
11630 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
11631 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
11632 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i32 0, i32 2
11633 // CHECK27-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
11634 // CHECK27-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
11635 // CHECK27-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
11636 // CHECK27-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
11637 // CHECK27-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
11638 // CHECK27-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i32 3
11639 // CHECK27-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
11640 // CHECK27-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
11641 // CHECK27-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
11642 // CHECK27-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
11643 // CHECK27-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
11644 // CHECK27-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i32 0, i32 1
11645 // CHECK27-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i32 0, i32 2
11646 // CHECK27-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
11647 // CHECK27-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
11648 // CHECK27-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
11649 // CHECK27-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP14]]
11650 // CHECK27-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i32 [[TMP23]]
11651 // CHECK27-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i32 3
11652 // CHECK27-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
11653 // CHECK27-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
11654 // CHECK27-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
11655 // CHECK27-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
11656 // CHECK27-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4
11657 // CHECK27-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
11658 // CHECK27-NEXT:    store i64 [[ADD13]], i64* [[X]], align 4
11659 // CHECK27-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
11660 // CHECK27-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4
11661 // CHECK27-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
11662 // CHECK27-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
11663 // CHECK27-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
11664 // CHECK27-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 4
11665 // CHECK27-NEXT:    ret void
11666 //
11667 //
11668 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
11669 // CHECK27-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
11670 // CHECK27-NEXT:  entry:
11671 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11672 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11673 // CHECK27-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
11674 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
11675 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 4
11676 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11677 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11678 // CHECK27-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
11679 // CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
11680 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11681 // CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
11682 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
11683 // CHECK27-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11684 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
11685 // CHECK27-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
11686 // CHECK27-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
11687 // CHECK27-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
11688 // CHECK27-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
11689 // CHECK27-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
11690 // CHECK27-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV1]], align 4
11691 // CHECK27-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
11692 // CHECK27-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
11693 // CHECK27-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 4
11694 // 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_]])
11695 // CHECK27-NEXT:    ret void
11696 //
11697 //
11698 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..4
11699 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11700 // CHECK27-NEXT:  entry:
11701 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11702 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11703 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
11704 // CHECK27-NEXT:    [[A:%.*]] = alloca i32, align 4
11705 // CHECK27-NEXT:    [[AA:%.*]] = alloca i16, align 2
11706 // CHECK27-NEXT:    [[AAA:%.*]] = alloca i8, align 1
11707 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11708 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11709 // CHECK27-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
11710 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
11711 // CHECK27-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
11712 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
11713 // CHECK27-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
11714 // CHECK27-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
11715 // CHECK27-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
11716 // CHECK27-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
11717 // CHECK27-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
11718 // CHECK27-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
11719 // CHECK27-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
11720 // CHECK27-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
11721 // CHECK27-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 4
11722 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
11723 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
11724 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
11725 // CHECK27-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
11726 // CHECK27-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
11727 // CHECK27-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
11728 // CHECK27-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
11729 // CHECK27-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
11730 // CHECK27-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
11731 // CHECK27-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
11732 // CHECK27-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
11733 // CHECK27-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
11734 // CHECK27-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
11735 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i32 0, i32 2
11736 // CHECK27-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
11737 // CHECK27-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
11738 // CHECK27-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
11739 // CHECK27-NEXT:    ret void
11740 //
11741 //
11742 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
11743 // CHECK27-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
11744 // CHECK27-NEXT:  entry:
11745 // CHECK27-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
11746 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
11747 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11748 // CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
11749 // CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
11750 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
11751 // CHECK27-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
11752 // CHECK27-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
11753 // CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
11754 // CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
11755 // CHECK27-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
11756 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
11757 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
11758 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
11759 // CHECK27-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
11760 // CHECK27-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11761 // CHECK27-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 4
11762 // CHECK27-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
11763 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_ADDR]], align 4
11764 // CHECK27-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 4
11765 // CHECK27-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
11766 // CHECK27-NEXT:    store i32 [[TMP1]], i32* [[TMP7]], align 4
11767 // CHECK27-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
11768 // CHECK27-NEXT:    store i32 [[TMP2]], i32* [[TMP8]], align 4
11769 // CHECK27-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
11770 // CHECK27-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 4
11771 // 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_]])
11772 // CHECK27-NEXT:    ret void
11773 //
11774 //
11775 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..5
11776 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11777 // CHECK27-NEXT:  entry:
11778 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11779 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11780 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
11781 // CHECK27-NEXT:    [[B:%.*]] = alloca i32, align 4
11782 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11783 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11784 // CHECK27-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
11785 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
11786 // CHECK27-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
11787 // CHECK27-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
11788 // CHECK27-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
11789 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
11790 // CHECK27-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
11791 // CHECK27-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
11792 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
11793 // CHECK27-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
11794 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
11795 // CHECK27-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
11796 // CHECK27-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 4
11797 // CHECK27-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
11798 // CHECK27-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
11799 // CHECK27-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
11800 // CHECK27-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
11801 // CHECK27-NEXT:    store double [[ADD]], double* [[A]], align 4
11802 // CHECK27-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
11803 // CHECK27-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 4
11804 // CHECK27-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
11805 // CHECK27-NEXT:    store double [[INC]], double* [[A1]], align 4
11806 // CHECK27-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
11807 // CHECK27-NEXT:    [[TMP13:%.*]] = mul nsw i32 1, [[TMP8]]
11808 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i32 [[TMP13]]
11809 // CHECK27-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
11810 // CHECK27-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
11811 // CHECK27-NEXT:    ret void
11812 //
11813 //
11814 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
11815 // CHECK27-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
11816 // CHECK27-NEXT:  entry:
11817 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11818 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11819 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
11820 // CHECK27-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 4
11821 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11822 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11823 // CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
11824 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11825 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
11826 // CHECK27-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11827 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
11828 // CHECK27-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
11829 // CHECK27-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
11830 // CHECK27-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
11831 // CHECK27-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
11832 // CHECK27-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
11833 // CHECK27-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 4
11834 // 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.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
11835 // CHECK27-NEXT:    ret void
11836 //
11837 //
11838 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..6
11839 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11840 // CHECK27-NEXT:  entry:
11841 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11842 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11843 // CHECK27-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 4
11844 // CHECK27-NEXT:    [[A:%.*]] = alloca i32, align 4
11845 // CHECK27-NEXT:    [[AA:%.*]] = alloca i16, align 2
11846 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11847 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11848 // CHECK27-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 4
11849 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 4
11850 // CHECK27-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
11851 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
11852 // CHECK27-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
11853 // CHECK27-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
11854 // CHECK27-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
11855 // CHECK27-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
11856 // CHECK27-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
11857 // CHECK27-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 4
11858 // CHECK27-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
11859 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
11860 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
11861 // CHECK27-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
11862 // CHECK27-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
11863 // CHECK27-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
11864 // CHECK27-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
11865 // CHECK27-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
11866 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i32 0, i32 2
11867 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
11868 // CHECK27-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
11869 // CHECK27-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
11870 // CHECK27-NEXT:    ret void
11871 //
11872 //
11873 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l100
11874 // CHECK28-SAME: () #[[ATTR0:[0-9]+]] {
11875 // CHECK28-NEXT:  entry:
11876 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
11877 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
11878 // CHECK28-NEXT:    ret void
11879 //
11880 //
11881 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined.
11882 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11883 // CHECK28-NEXT:  entry:
11884 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11885 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11886 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4
11887 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11888 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11889 // CHECK28-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4
11890 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4
11891 // CHECK28-NEXT:    ret void
11892 //
11893 //
11894 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l110
11895 // CHECK28-SAME: (i32 [[AA:%.*]]) #[[ATTR0]] {
11896 // CHECK28-NEXT:  entry:
11897 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11898 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 2
11899 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11900 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11901 // CHECK28-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11902 // CHECK28-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 4
11903 // CHECK28-NEXT:    store i16 [[TMP1]], i16* [[TMP0]], align 2
11904 // 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_]])
11905 // CHECK28-NEXT:    ret void
11906 //
11907 //
11908 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..1
11909 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11910 // CHECK28-NEXT:  entry:
11911 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11912 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11913 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4
11914 // CHECK28-NEXT:    [[AA:%.*]] = alloca i16, align 2
11915 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11916 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11917 // CHECK28-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4
11918 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4
11919 // CHECK28-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
11920 // CHECK28-NEXT:    [[TMP2:%.*]] = load i16, i16* [[TMP1]], align 2
11921 // CHECK28-NEXT:    store i16 [[TMP2]], i16* [[AA]], align 2
11922 // CHECK28-NEXT:    [[TMP3:%.*]] = load i16, i16* [[AA]], align 2
11923 // CHECK28-NEXT:    [[CONV:%.*]] = sext i16 [[TMP3]] to i32
11924 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 1
11925 // CHECK28-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
11926 // CHECK28-NEXT:    store i16 [[CONV1]], i16* [[AA]], align 2
11927 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
11928 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
11929 // CHECK28-NEXT:    [[TMP6:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 1)
11930 // CHECK28-NEXT:    [[TMP7:%.*]] = icmp ne i32 [[TMP6]], 0
11931 // CHECK28-NEXT:    br i1 [[TMP7]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
11932 // CHECK28:       .cancel.exit:
11933 // CHECK28-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]])
11934 // CHECK28-NEXT:    br label [[DOTCANCEL_CONTINUE]]
11935 // CHECK28:       .cancel.continue:
11936 // CHECK28-NEXT:    ret void
11937 //
11938 //
11939 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l119
11940 // CHECK28-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]]) #[[ATTR0]] {
11941 // CHECK28-NEXT:  entry:
11942 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11943 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11944 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4
11945 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11946 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11947 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11948 // CHECK28-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
11949 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
11950 // CHECK28-NEXT:    store i32 [[TMP1]], i32* [[TMP0]], align 4
11951 // CHECK28-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
11952 // CHECK28-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 4
11953 // CHECK28-NEXT:    store i16 [[TMP3]], i16* [[TMP2]], align 4
11954 // 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_]])
11955 // CHECK28-NEXT:    ret void
11956 //
11957 //
11958 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..2
11959 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
11960 // CHECK28-NEXT:  entry:
11961 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11962 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11963 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4
11964 // CHECK28-NEXT:    [[A:%.*]] = alloca i32, align 4
11965 // CHECK28-NEXT:    [[AA:%.*]] = alloca i16, align 2
11966 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11967 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11968 // CHECK28-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4
11969 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4
11970 // CHECK28-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
11971 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
11972 // CHECK28-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
11973 // CHECK28-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
11974 // CHECK28-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
11975 // CHECK28-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
11976 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
11977 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP5]], 1
11978 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
11979 // CHECK28-NEXT:    [[TMP6:%.*]] = load i16, i16* [[AA]], align 2
11980 // CHECK28-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
11981 // CHECK28-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
11982 // CHECK28-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
11983 // CHECK28-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
11984 // CHECK28-NEXT:    ret void
11985 //
11986 //
11987 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l144
11988 // CHECK28-SAME: (i32 [[A:%.*]], [10 x float]* nonnull align 4 dereferenceable(40) [[B:%.*]], i32 [[VLA:%.*]], float* nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* nonnull align 4 dereferenceable(400) [[C:%.*]], i32 [[VLA1:%.*]], i32 [[VLA3:%.*]], double* nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* nonnull align 4 dereferenceable(12) [[D:%.*]]) #[[ATTR0]] {
11989 // CHECK28-NEXT:  entry:
11990 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11991 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
11992 // CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11993 // CHECK28-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
11994 // CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
11995 // CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
11996 // CHECK28-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
11997 // CHECK28-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
11998 // CHECK28-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
11999 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 4
12000 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
12001 // CHECK28-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
12002 // CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
12003 // CHECK28-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
12004 // CHECK28-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
12005 // CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
12006 // CHECK28-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
12007 // CHECK28-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
12008 // CHECK28-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
12009 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
12010 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
12011 // CHECK28-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
12012 // CHECK28-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
12013 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
12014 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
12015 // CHECK28-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
12016 // CHECK28-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
12017 // CHECK28-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
12018 // CHECK28-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
12019 // CHECK28-NEXT:    store i32 [[TMP9]], i32* [[TMP8]], align 4
12020 // CHECK28-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
12021 // CHECK28-NEXT:    store [10 x float]* [[TMP0]], [10 x float]** [[TMP10]], align 4
12022 // CHECK28-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
12023 // CHECK28-NEXT:    store i32 [[TMP1]], i32* [[TMP11]], align 4
12024 // CHECK28-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
12025 // CHECK28-NEXT:    store float* [[TMP2]], float** [[TMP12]], align 4
12026 // CHECK28-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
12027 // CHECK28-NEXT:    store [5 x [10 x double]]* [[TMP3]], [5 x [10 x double]]** [[TMP13]], align 4
12028 // CHECK28-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 5
12029 // CHECK28-NEXT:    store i32 [[TMP4]], i32* [[TMP14]], align 4
12030 // CHECK28-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 6
12031 // CHECK28-NEXT:    store i32 [[TMP5]], i32* [[TMP15]], align 4
12032 // CHECK28-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 7
12033 // CHECK28-NEXT:    store double* [[TMP6]], double** [[TMP16]], align 4
12034 // CHECK28-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 8
12035 // CHECK28-NEXT:    store %struct.TT* [[TMP7]], %struct.TT** [[TMP17]], align 4
12036 // 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_]])
12037 // CHECK28-NEXT:    ret void
12038 //
12039 //
12040 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..3
12041 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
12042 // CHECK28-NEXT:  entry:
12043 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12044 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12045 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4
12046 // CHECK28-NEXT:    [[A:%.*]] = alloca i32, align 4
12047 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12048 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12049 // CHECK28-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4
12050 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4
12051 // CHECK28-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
12052 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
12053 // CHECK28-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
12054 // CHECK28-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
12055 // CHECK28-NEXT:    [[TMP4:%.*]] = load [10 x float]*, [10 x float]** [[TMP3]], align 4
12056 // CHECK28-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
12057 // CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
12058 // CHECK28-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
12059 // CHECK28-NEXT:    [[TMP8:%.*]] = load float*, float** [[TMP7]], align 4
12060 // CHECK28-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 4
12061 // CHECK28-NEXT:    [[TMP10:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[TMP9]], align 4
12062 // CHECK28-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 5
12063 // CHECK28-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
12064 // CHECK28-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 6
12065 // CHECK28-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
12066 // CHECK28-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 7
12067 // CHECK28-NEXT:    [[TMP16:%.*]] = load double*, double** [[TMP15]], align 4
12068 // CHECK28-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 8
12069 // CHECK28-NEXT:    [[TMP18:%.*]] = load %struct.TT*, %struct.TT** [[TMP17]], align 4
12070 // CHECK28-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
12071 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
12072 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
12073 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP4]], i32 0, i32 2
12074 // CHECK28-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4
12075 // CHECK28-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
12076 // CHECK28-NEXT:    [[ADD1:%.*]] = fadd double [[CONV]], 1.000000e+00
12077 // CHECK28-NEXT:    [[CONV2:%.*]] = fptrunc double [[ADD1]] to float
12078 // CHECK28-NEXT:    store float [[CONV2]], float* [[ARRAYIDX]], align 4
12079 // CHECK28-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP8]], i32 3
12080 // CHECK28-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX3]], align 4
12081 // CHECK28-NEXT:    [[CONV4:%.*]] = fpext float [[TMP21]] to double
12082 // CHECK28-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.000000e+00
12083 // CHECK28-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
12084 // CHECK28-NEXT:    store float [[CONV6]], float* [[ARRAYIDX3]], align 4
12085 // CHECK28-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP10]], i32 0, i32 1
12086 // CHECK28-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX7]], i32 0, i32 2
12087 // CHECK28-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX8]], align 8
12088 // CHECK28-NEXT:    [[ADD9:%.*]] = fadd double [[TMP22]], 1.000000e+00
12089 // CHECK28-NEXT:    store double [[ADD9]], double* [[ARRAYIDX8]], align 8
12090 // CHECK28-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP14]]
12091 // CHECK28-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds double, double* [[TMP16]], i32 [[TMP23]]
12092 // CHECK28-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX10]], i32 3
12093 // CHECK28-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX11]], align 8
12094 // CHECK28-NEXT:    [[ADD12:%.*]] = fadd double [[TMP24]], 1.000000e+00
12095 // CHECK28-NEXT:    store double [[ADD12]], double* [[ARRAYIDX11]], align 8
12096 // CHECK28-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP18]], i32 0, i32 0
12097 // CHECK28-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4
12098 // CHECK28-NEXT:    [[ADD13:%.*]] = add nsw i64 [[TMP25]], 1
12099 // CHECK28-NEXT:    store i64 [[ADD13]], i64* [[X]], align 4
12100 // CHECK28-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP18]], i32 0, i32 1
12101 // CHECK28-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4
12102 // CHECK28-NEXT:    [[CONV14:%.*]] = sext i8 [[TMP26]] to i32
12103 // CHECK28-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
12104 // CHECK28-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i8
12105 // CHECK28-NEXT:    store i8 [[CONV16]], i8* [[Y]], align 4
12106 // CHECK28-NEXT:    ret void
12107 //
12108 //
12109 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l198
12110 // CHECK28-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], i32 [[AAA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
12111 // CHECK28-NEXT:  entry:
12112 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
12113 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
12114 // CHECK28-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
12115 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
12116 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 4
12117 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
12118 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
12119 // CHECK28-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
12120 // CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
12121 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
12122 // CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
12123 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
12124 // CHECK28-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
12125 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
12126 // CHECK28-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
12127 // CHECK28-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
12128 // CHECK28-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
12129 // CHECK28-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
12130 // CHECK28-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
12131 // CHECK28-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CONV1]], align 4
12132 // CHECK28-NEXT:    store i8 [[TMP6]], i8* [[TMP5]], align 2
12133 // CHECK28-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
12134 // CHECK28-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP7]], align 4
12135 // 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_]])
12136 // CHECK28-NEXT:    ret void
12137 //
12138 //
12139 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..4
12140 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
12141 // CHECK28-NEXT:  entry:
12142 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12143 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12144 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 4
12145 // CHECK28-NEXT:    [[A:%.*]] = alloca i32, align 4
12146 // CHECK28-NEXT:    [[AA:%.*]] = alloca i16, align 2
12147 // CHECK28-NEXT:    [[AAA:%.*]] = alloca i8, align 1
12148 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12149 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12150 // CHECK28-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 4
12151 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 4
12152 // CHECK28-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
12153 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
12154 // CHECK28-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
12155 // CHECK28-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
12156 // CHECK28-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
12157 // CHECK28-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
12158 // CHECK28-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
12159 // CHECK28-NEXT:    [[TMP6:%.*]] = load i8, i8* [[TMP5]], align 2
12160 // CHECK28-NEXT:    store i8 [[TMP6]], i8* [[AAA]], align 1
12161 // CHECK28-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
12162 // CHECK28-NEXT:    [[TMP8:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP7]], align 4
12163 // CHECK28-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
12164 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], 1
12165 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
12166 // CHECK28-NEXT:    [[TMP10:%.*]] = load i16, i16* [[AA]], align 2
12167 // CHECK28-NEXT:    [[CONV:%.*]] = sext i16 [[TMP10]] to i32
12168 // CHECK28-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
12169 // CHECK28-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
12170 // CHECK28-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
12171 // CHECK28-NEXT:    [[TMP11:%.*]] = load i8, i8* [[AAA]], align 1
12172 // CHECK28-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP11]] to i32
12173 // CHECK28-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
12174 // CHECK28-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i8
12175 // CHECK28-NEXT:    store i8 [[CONV5]], i8* [[AAA]], align 1
12176 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP8]], i32 0, i32 2
12177 // CHECK28-NEXT:    [[TMP12:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
12178 // CHECK28-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP12]], 1
12179 // CHECK28-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
12180 // CHECK28-NEXT:    ret void
12181 //
12182 //
12183 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
12184 // CHECK28-SAME: (%struct.S1* [[THIS:%.*]], i32 [[B:%.*]], i32 [[VLA:%.*]], i32 [[VLA1:%.*]], i16* nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
12185 // CHECK28-NEXT:  entry:
12186 // CHECK28-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
12187 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
12188 // CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
12189 // CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
12190 // CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
12191 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 4
12192 // CHECK28-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
12193 // CHECK28-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
12194 // CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
12195 // CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
12196 // CHECK28-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
12197 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
12198 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
12199 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
12200 // CHECK28-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
12201 // CHECK28-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
12202 // CHECK28-NEXT:    store %struct.S1* [[TMP0]], %struct.S1** [[TMP4]], align 4
12203 // CHECK28-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
12204 // CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_ADDR]], align 4
12205 // CHECK28-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 4
12206 // CHECK28-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
12207 // CHECK28-NEXT:    store i32 [[TMP1]], i32* [[TMP7]], align 4
12208 // CHECK28-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
12209 // CHECK28-NEXT:    store i32 [[TMP2]], i32* [[TMP8]], align 4
12210 // CHECK28-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
12211 // CHECK28-NEXT:    store i16* [[TMP3]], i16** [[TMP9]], align 4
12212 // 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_]])
12213 // CHECK28-NEXT:    ret void
12214 //
12215 //
12216 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..5
12217 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
12218 // CHECK28-NEXT:  entry:
12219 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12220 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12221 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 4
12222 // CHECK28-NEXT:    [[B:%.*]] = alloca i32, align 4
12223 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12224 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12225 // CHECK28-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 4
12226 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 4
12227 // CHECK28-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
12228 // CHECK28-NEXT:    [[TMP2:%.*]] = load %struct.S1*, %struct.S1** [[TMP1]], align 4
12229 // CHECK28-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
12230 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
12231 // CHECK28-NEXT:    store i32 [[TMP4]], i32* [[B]], align 4
12232 // CHECK28-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
12233 // CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
12234 // CHECK28-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
12235 // CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
12236 // CHECK28-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 4
12237 // CHECK28-NEXT:    [[TMP10:%.*]] = load i16*, i16** [[TMP9]], align 4
12238 // CHECK28-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4
12239 // CHECK28-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
12240 // CHECK28-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
12241 // CHECK28-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP2]], i32 0, i32 0
12242 // CHECK28-NEXT:    store double [[ADD]], double* [[A]], align 4
12243 // CHECK28-NEXT:    [[A1:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP2]], i32 0, i32 0
12244 // CHECK28-NEXT:    [[TMP12:%.*]] = load double, double* [[A1]], align 4
12245 // CHECK28-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
12246 // CHECK28-NEXT:    store double [[INC]], double* [[A1]], align 4
12247 // CHECK28-NEXT:    [[CONV2:%.*]] = fptosi double [[INC]] to i16
12248 // CHECK28-NEXT:    [[TMP13:%.*]] = mul nsw i32 1, [[TMP8]]
12249 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP10]], i32 [[TMP13]]
12250 // CHECK28-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
12251 // CHECK28-NEXT:    store i16 [[CONV2]], i16* [[ARRAYIDX3]], align 2
12252 // CHECK28-NEXT:    ret void
12253 //
12254 //
12255 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l181
12256 // CHECK28-SAME: (i32 [[A:%.*]], i32 [[AA:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
12257 // CHECK28-NEXT:  entry:
12258 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
12259 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
12260 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
12261 // CHECK28-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 4
12262 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
12263 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
12264 // CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
12265 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
12266 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
12267 // CHECK28-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
12268 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_ADDR]], align 4
12269 // CHECK28-NEXT:    store i32 [[TMP2]], i32* [[TMP1]], align 4
12270 // CHECK28-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
12271 // CHECK28-NEXT:    [[TMP4:%.*]] = load i16, i16* [[CONV]], align 4
12272 // CHECK28-NEXT:    store i16 [[TMP4]], i16* [[TMP3]], align 4
12273 // CHECK28-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
12274 // CHECK28-NEXT:    store [10 x i32]* [[TMP0]], [10 x i32]** [[TMP5]], align 4
12275 // 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.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
12276 // CHECK28-NEXT:    ret void
12277 //
12278 //
12279 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..6
12280 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR0]] {
12281 // CHECK28-NEXT:  entry:
12282 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12283 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12284 // CHECK28-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 4
12285 // CHECK28-NEXT:    [[A:%.*]] = alloca i32, align 4
12286 // CHECK28-NEXT:    [[AA:%.*]] = alloca i16, align 2
12287 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12288 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12289 // CHECK28-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 4
12290 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 4
12291 // CHECK28-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
12292 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
12293 // CHECK28-NEXT:    store i32 [[TMP2]], i32* [[A]], align 4
12294 // CHECK28-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
12295 // CHECK28-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP3]], align 4
12296 // CHECK28-NEXT:    store i16 [[TMP4]], i16* [[AA]], align 2
12297 // CHECK28-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
12298 // CHECK28-NEXT:    [[TMP6:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP5]], align 4
12299 // CHECK28-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
12300 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
12301 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
12302 // CHECK28-NEXT:    [[TMP8:%.*]] = load i16, i16* [[AA]], align 2
12303 // CHECK28-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
12304 // CHECK28-NEXT:    [[ADD1:%.*]] = add nsw i32 [[CONV]], 1
12305 // CHECK28-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD1]] to i16
12306 // CHECK28-NEXT:    store i16 [[CONV2]], i16* [[AA]], align 2
12307 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP6]], i32 0, i32 2
12308 // CHECK28-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
12309 // CHECK28-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
12310 // CHECK28-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX]], align 4
12311 // CHECK28-NEXT:    ret void
12312 //
12313