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 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple x86_64-apple-darwin13.4.0 -emit-pch -o %t.0 %s
4 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -std=c++11 -include-pch %t.0 -fsyntax-only -verify %s -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - | FileCheck %s --check-prefix=CHECK2
5 
6 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -fopenmp-enable-irbuilder -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK3
7 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -fopenmp-enable-irbuilder -x c++ -std=c++11 -triple x86_64-apple-darwin13.4.0 -emit-pch -o %t.1 %s
8 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -fopenmp-enable-irbuilder -std=c++11 -include-pch %t.1 -fsyntax-only -verify %s -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - | FileCheck %s --check-prefix=CHECK4
9 
10 // RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - %s | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
11 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple x86_64-apple-darwin13.4.0 -emit-pch -o %t.2 %s
12 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -std=c++11 -include-pch %t.2 -fsyntax-only -verify %s -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
13 
14 // RUN: %clang_cc1 -verify -fopenmp -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK7
15 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-apple-darwin13.4.0 -emit-pch -o %t.3 %s
16 // RUN: %clang_cc1 -fopenmp -std=c++11 -include-pch %t.3 -fsyntax-only -verify %s -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - | FileCheck %s --check-prefix=CHECK8
17 
18 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-enable-irbuilder -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK9
19 // RUN: %clang_cc1 -fopenmp -fopenmp-enable-irbuilder -x c++ -std=c++11 -triple x86_64-apple-darwin13.4.0 -emit-pch -o %t.4 %s
20 // RUN: %clang_cc1 -fopenmp -fopenmp-enable-irbuilder -std=c++11 -include-pch %t.4 -fsyntax-only -verify %s -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - | FileCheck %s --check-prefix=CHECK10
21 
22 // RUN: %clang_cc1 -verify -fopenmp-simd -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - %s | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
23 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-apple-darwin13.4.0 -emit-pch -o %t.5 %s
24 // RUN: %clang_cc1 -fopenmp-simd -std=c++11 -include-pch %t.5 -fsyntax-only -verify %s -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
25 
26 // expected-no-diagnostics
27 #ifndef HEADER
28 #define HEADER
29 
30 float flag;
31 int main (int argc, char **argv) {
32 #pragma omp parallel
33 {
34 #pragma omp cancel parallel if(flag)
35   argv[0][0] = argc;
36 #pragma omp barrier
37   argv[0][0] += argc;
38 }
39 #pragma omp sections
40 {
41 #pragma omp cancel sections
42 }
43 #pragma omp sections
44 {
45 #pragma omp cancel sections
46 #pragma omp section
47   {
48 #pragma omp cancel sections
49   }
50 }
51 #pragma omp for
52 for (int i = 0; i < argc; ++i) {
53 #pragma omp cancel for if(cancel: flag)
54 }
55 #pragma omp task
56 {
57 #pragma omp cancel taskgroup
58 }
59 #pragma omp parallel sections
60 {
61 #pragma omp cancel sections
62 }
63 #pragma omp parallel sections
64 {
65 #pragma omp cancel sections
66 #pragma omp section
67   {
68 #pragma omp cancel sections
69   }
70 }
71 int r = 0;
72 #pragma omp parallel for reduction(+: r)
73 for (int i = 0; i < argc; ++i) {
74 #pragma omp cancel for
75   r += i;
76 }
77   return argc;
78 }
79 
80 #endif
81 // CHECK1-LABEL: define {{[^@]+}}@main
82 // CHECK1-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
83 // CHECK1-NEXT:  entry:
84 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
85 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
86 // CHECK1-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
87 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
88 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
89 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
90 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
91 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
92 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_LB_1:%.*]] = alloca i32, align 4
93 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_UB_2:%.*]] = alloca i32, align 4
94 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_ST_3:%.*]] = alloca i32, align 4
95 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IL_4:%.*]] = alloca i32, align 4
96 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IV_5:%.*]] = alloca i32, align 4
97 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
98 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
99 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
100 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_21:%.*]] = alloca i32, align 4
101 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
102 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
103 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
104 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
105 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
106 // CHECK1-NEXT:    [[I24:%.*]] = alloca i32, align 4
107 // CHECK1-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
108 // CHECK1-NEXT:    [[R:%.*]] = alloca i32, align 4
109 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
110 // CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
111 // CHECK1-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
112 // CHECK1-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
113 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i8***, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i8*** [[ARGV_ADDR]], i32* [[ARGC_ADDR]])
114 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
115 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
116 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
117 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
118 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
119 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
120 // CHECK1-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
121 // CHECK1-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
122 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
123 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
124 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
125 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
126 // CHECK1:       omp.inner.for.cond:
127 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
128 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
129 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
130 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
131 // CHECK1:       omp.inner.for.body:
132 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
133 // CHECK1-NEXT:    switch i32 [[TMP7]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
134 // CHECK1-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
135 // CHECK1-NEXT:    ]
136 // CHECK1:       .omp.sections.case:
137 // CHECK1-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
138 // CHECK1-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
139 // CHECK1-NEXT:    br i1 [[TMP9]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
140 // CHECK1:       .cancel.exit:
141 // CHECK1-NEXT:    br label [[CANCEL_EXIT:%.*]]
142 // CHECK1:       .cancel.continue:
143 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
144 // CHECK1:       .omp.sections.exit:
145 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
146 // CHECK1:       omp.inner.for.inc:
147 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
148 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP10]], 1
149 // CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
150 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
151 // CHECK1:       omp.inner.for.end:
152 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
153 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
154 // CHECK1:       cancel.cont:
155 // CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
156 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
157 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
158 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_3]], align 4
159 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_4]], align 4
160 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_4]], i32* [[DOTOMP_SECTIONS_LB_1]], i32* [[DOTOMP_SECTIONS_UB_2]], i32* [[DOTOMP_SECTIONS_ST_3]], i32 1, i32 1)
161 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
162 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp slt i32 [[TMP11]], 1
163 // CHECK1-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 1
164 // CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_SECTIONS_UB_2]], align 4
165 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
166 // CHECK1-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
167 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND6:%.*]]
168 // CHECK1:       omp.inner.for.cond6:
169 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
170 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
171 // CHECK1-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
172 // CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY8:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
173 // CHECK1:       omp.inner.for.body8:
174 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
175 // CHECK1-NEXT:    switch i32 [[TMP17]], label [[DOTOMP_SECTIONS_EXIT15:%.*]] [
176 // CHECK1-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE9:%.*]]
177 // CHECK1-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE12:%.*]]
178 // CHECK1-NEXT:    ]
179 // CHECK1:       .omp.sections.case9:
180 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
181 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
182 // CHECK1-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT10:%.*]], label [[DOTCANCEL_CONTINUE11:%.*]]
183 // CHECK1:       .cancel.exit10:
184 // CHECK1-NEXT:    br label [[CANCEL_EXIT19:%.*]]
185 // CHECK1:       cancel.exit:
186 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
187 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
188 // CHECK1:       .cancel.continue11:
189 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
190 // CHECK1:       .omp.sections.case12:
191 // CHECK1-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
192 // CHECK1-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
193 // CHECK1-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT13:%.*]], label [[DOTCANCEL_CONTINUE14:%.*]]
194 // CHECK1:       .cancel.exit13:
195 // CHECK1-NEXT:    br label [[CANCEL_EXIT19]]
196 // CHECK1:       .cancel.continue14:
197 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
198 // CHECK1:       .omp.sections.exit15:
199 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
200 // CHECK1:       omp.inner.for.inc16:
201 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
202 // CHECK1-NEXT:    [[INC17:%.*]] = add nsw i32 [[TMP22]], 1
203 // CHECK1-NEXT:    store i32 [[INC17]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
204 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND6]]
205 // CHECK1:       omp.inner.for.end18:
206 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
207 // CHECK1-NEXT:    br label [[CANCEL_CONT20:%.*]]
208 // CHECK1:       cancel.cont20:
209 // CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
210 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
211 // CHECK1-NEXT:    store i32 [[TMP23]], i32* [[DOTCAPTURE_EXPR_]], align 4
212 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
213 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
214 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
215 // CHECK1-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[DIV]], 1
216 // CHECK1-NEXT:    store i32 [[SUB22]], i32* [[DOTCAPTURE_EXPR_21]], align 4
217 // CHECK1-NEXT:    store i32 0, i32* [[I]], align 4
218 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
219 // CHECK1-NEXT:    [[CMP23:%.*]] = icmp slt i32 0, [[TMP25]]
220 // CHECK1-NEXT:    br i1 [[CMP23]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
221 // CHECK1:       omp.precond.then:
222 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
223 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
224 // CHECK1-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_UB]], align 4
225 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
226 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
227 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
228 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
229 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
230 // CHECK1-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
231 // CHECK1-NEXT:    br i1 [[CMP25]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
232 // CHECK1:       cond.true:
233 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
234 // CHECK1-NEXT:    br label [[COND_END:%.*]]
235 // CHECK1:       cond.false:
236 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
237 // CHECK1-NEXT:    br label [[COND_END]]
238 // CHECK1:       cond.end:
239 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE]] ], [ [[TMP30]], [[COND_FALSE]] ]
240 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
241 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
242 // CHECK1-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
243 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND26:%.*]]
244 // CHECK1:       omp.inner.for.cond26:
245 // CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
246 // CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
247 // CHECK1-NEXT:    [[CMP27:%.*]] = icmp sle i32 [[TMP32]], [[TMP33]]
248 // CHECK1-NEXT:    br i1 [[CMP27]], label [[OMP_INNER_FOR_BODY28:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
249 // CHECK1:       omp.inner.for.body28:
250 // CHECK1-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
251 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP34]], 1
252 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
253 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I24]], align 4
254 // CHECK1-NEXT:    [[TMP35:%.*]] = load float, float* @flag, align 4
255 // CHECK1-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP35]], 0.000000e+00
256 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
257 // CHECK1:       omp_if.then:
258 // CHECK1-NEXT:    [[TMP36:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
259 // CHECK1-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
260 // CHECK1-NEXT:    br i1 [[TMP37]], label [[DOTCANCEL_EXIT29:%.*]], label [[DOTCANCEL_CONTINUE30:%.*]]
261 // CHECK1:       .cancel.exit29:
262 // CHECK1-NEXT:    br label [[CANCEL_EXIT34:%.*]]
263 // CHECK1:       cancel.exit19:
264 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
265 // CHECK1-NEXT:    br label [[CANCEL_CONT20]]
266 // CHECK1:       .cancel.continue30:
267 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
268 // CHECK1:       omp_if.else:
269 // CHECK1-NEXT:    br label [[OMP_IF_END]]
270 // CHECK1:       omp_if.end:
271 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
272 // CHECK1:       omp.body.continue:
273 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
274 // CHECK1:       omp.inner.for.inc31:
275 // CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
276 // CHECK1-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP38]], 1
277 // CHECK1-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV]], align 4
278 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND26]]
279 // CHECK1:       omp.inner.for.end33:
280 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
281 // CHECK1:       omp.loop.exit:
282 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
283 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
284 // CHECK1:       cancel.exit34:
285 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
286 // CHECK1-NEXT:    br label [[CANCEL_CONT35:%.*]]
287 // CHECK1:       omp.precond.end:
288 // CHECK1-NEXT:    br label [[CANCEL_CONT35]]
289 // CHECK1:       cancel.cont35:
290 // CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]])
291 // CHECK1-NEXT:    [[TMP39:%.*]] = call i8* @__kmpc_omp_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*)*))
292 // CHECK1-NEXT:    [[TMP40:%.*]] = bitcast i8* [[TMP39]] to %struct.kmp_task_t_with_privates*
293 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP40]], i32 0, i32 0
294 // CHECK1-NEXT:    [[TMP42:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP39]])
295 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
296 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*))
297 // CHECK1-NEXT:    store i32 0, i32* [[R]], align 4
298 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])
299 // CHECK1-NEXT:    [[TMP43:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
300 // CHECK1-NEXT:    ret i32 [[TMP43]]
301 //
302 //
303 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
304 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8*** nonnull align 8 dereferenceable(8) [[ARGV:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]]) #[[ATTR1:[0-9]+]] {
305 // CHECK1-NEXT:  entry:
306 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
307 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
308 // CHECK1-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8***, align 8
309 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
310 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
311 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
312 // CHECK1-NEXT:    store i8*** [[ARGV]], i8**** [[ARGV_ADDR]], align 8
313 // CHECK1-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
314 // CHECK1-NEXT:    [[TMP0:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
315 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
316 // CHECK1-NEXT:    [[TMP2:%.*]] = load float, float* @flag, align 4
317 // CHECK1-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP2]], 0.000000e+00
318 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
319 // CHECK1:       omp_if.then:
320 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
321 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
322 // CHECK1-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 1)
323 // CHECK1-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[TMP5]], 0
324 // CHECK1-NEXT:    br i1 [[TMP6]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
325 // CHECK1:       .cancel.exit:
326 // CHECK1-NEXT:    br label [[RETURN:%.*]]
327 // CHECK1:       .cancel.continue:
328 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
329 // CHECK1:       omp_if.else:
330 // CHECK1-NEXT:    br label [[OMP_IF_END]]
331 // CHECK1:       omp_if.end:
332 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP1]], align 4
333 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP7]] to i8
334 // CHECK1-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP0]], align 8
335 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 0
336 // CHECK1-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
337 // CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 0
338 // CHECK1-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX1]], align 1
339 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
340 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
341 // CHECK1-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP11]])
342 // CHECK1-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
343 // CHECK1-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
344 // CHECK1:       .cancel.exit2:
345 // CHECK1-NEXT:    br label [[RETURN]]
346 // CHECK1:       .cancel.continue3:
347 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP1]], align 4
348 // CHECK1-NEXT:    [[TMP15:%.*]] = load i8**, i8*** [[TMP0]], align 8
349 // CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP15]], i64 0
350 // CHECK1-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
351 // CHECK1-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP16]], i64 0
352 // CHECK1-NEXT:    [[TMP17:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
353 // CHECK1-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP17]] to i32
354 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP14]]
355 // CHECK1-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD]] to i8
356 // CHECK1-NEXT:    store i8 [[CONV7]], i8* [[ARRAYIDX5]], align 1
357 // CHECK1-NEXT:    br label [[RETURN]]
358 // CHECK1:       return:
359 // CHECK1-NEXT:    ret void
360 //
361 //
362 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
363 // CHECK1-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
364 // CHECK1-NEXT:  entry:
365 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
366 // CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
367 // CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
368 // CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
369 // CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
370 // CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
371 // CHECK1-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
372 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
373 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
374 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
375 // CHECK1-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
376 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
377 // CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
378 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
379 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
380 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
381 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
382 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
383 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
384 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META4:![0-9]+]])
385 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META7:![0-9]+]])
386 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META9:![0-9]+]])
387 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META11:![0-9]+]])
388 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !13
389 // CHECK1-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !13
390 // CHECK1-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !13
391 // CHECK1-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !13
392 // CHECK1-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !13
393 // CHECK1-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
394 // CHECK1-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
395 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !13
396 // CHECK1-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 4) #[[ATTR2:[0-9]+]]
397 // CHECK1-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
398 // CHECK1-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
399 // CHECK1:       .cancel.exit.i:
400 // CHECK1-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
401 // CHECK1-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
402 // CHECK1:       .cancel.continue.i:
403 // CHECK1-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
404 // CHECK1-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
405 // CHECK1:       .omp_outlined..1.exit:
406 // CHECK1-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
407 // CHECK1-NEXT:    ret i32 0
408 //
409 //
410 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
411 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
412 // CHECK1-NEXT:  entry:
413 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
414 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
415 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
416 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
417 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
418 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
419 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
420 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
421 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
422 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
423 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
424 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
425 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
426 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
427 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
428 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
429 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
430 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
431 // CHECK1-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
432 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
433 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
434 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
435 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
436 // CHECK1:       omp.inner.for.cond:
437 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
438 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
439 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
440 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
441 // CHECK1:       omp.inner.for.body:
442 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
443 // CHECK1-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
444 // CHECK1-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
445 // CHECK1-NEXT:    ]
446 // CHECK1:       .omp.sections.case:
447 // CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
448 // CHECK1-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
449 // CHECK1-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
450 // CHECK1:       .cancel.exit:
451 // CHECK1-NEXT:    br label [[CANCEL_EXIT:%.*]]
452 // CHECK1:       .cancel.continue:
453 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
454 // CHECK1:       .omp.sections.exit:
455 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
456 // CHECK1:       omp.inner.for.inc:
457 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
458 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
459 // CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
460 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
461 // CHECK1:       omp.inner.for.end:
462 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
463 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
464 // CHECK1:       cancel.cont:
465 // CHECK1-NEXT:    ret void
466 // CHECK1:       cancel.exit:
467 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
468 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
469 //
470 //
471 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
472 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
473 // CHECK1-NEXT:  entry:
474 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
475 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
476 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
477 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
478 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
479 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
480 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
481 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
482 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
483 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
484 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
485 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
486 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
487 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
488 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
489 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
490 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
491 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
492 // CHECK1-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
493 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
494 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
495 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
496 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
497 // CHECK1:       omp.inner.for.cond:
498 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
499 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
500 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
501 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
502 // CHECK1:       omp.inner.for.body:
503 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
504 // CHECK1-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
505 // CHECK1-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
506 // CHECK1-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE1:%.*]]
507 // CHECK1-NEXT:    ]
508 // CHECK1:       .omp.sections.case:
509 // CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
510 // CHECK1-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
511 // CHECK1-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
512 // CHECK1:       .cancel.exit:
513 // CHECK1-NEXT:    br label [[CANCEL_EXIT:%.*]]
514 // CHECK1:       .cancel.continue:
515 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
516 // CHECK1:       .omp.sections.case1:
517 // CHECK1-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
518 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
519 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
520 // CHECK1:       .cancel.exit2:
521 // CHECK1-NEXT:    br label [[CANCEL_EXIT]]
522 // CHECK1:       .cancel.continue3:
523 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
524 // CHECK1:       .omp.sections.exit:
525 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
526 // CHECK1:       omp.inner.for.inc:
527 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
528 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
529 // CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
530 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
531 // CHECK1:       omp.inner.for.end:
532 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
533 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
534 // CHECK1:       cancel.cont:
535 // CHECK1-NEXT:    ret void
536 // CHECK1:       cancel.exit:
537 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
538 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
539 //
540 //
541 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
542 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR1]] {
543 // CHECK1-NEXT:  entry:
544 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
545 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
546 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
547 // CHECK1-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
548 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
549 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
550 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
551 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
552 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
553 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
554 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
555 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
556 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
557 // CHECK1-NEXT:    [[R3:%.*]] = alloca i32, align 4
558 // CHECK1-NEXT:    [[I4:%.*]] = alloca i32, align 4
559 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
560 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
561 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
562 // CHECK1-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
563 // CHECK1-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
564 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
565 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
566 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
567 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
568 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
569 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
570 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
571 // CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
572 // CHECK1-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
573 // CHECK1-NEXT:    store i32 0, i32* [[I]], align 4
574 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
575 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
576 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
577 // CHECK1:       omp.precond.then:
578 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
579 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
580 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
581 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
582 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
583 // CHECK1-NEXT:    store i32 0, i32* [[R3]], align 4
584 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
585 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
586 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
587 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
588 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
589 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
590 // CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
591 // CHECK1:       cond.true:
592 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
593 // CHECK1-NEXT:    br label [[COND_END:%.*]]
594 // CHECK1:       cond.false:
595 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
596 // CHECK1-NEXT:    br label [[COND_END]]
597 // CHECK1:       cond.end:
598 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
599 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
600 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
601 // CHECK1-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
602 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
603 // CHECK1:       omp.inner.for.cond:
604 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
605 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
606 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
607 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
608 // CHECK1:       omp.inner.for.body:
609 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
610 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
611 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
612 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
613 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
614 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4
615 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP17]], i32 2)
616 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
617 // CHECK1-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
618 // CHECK1:       .cancel.exit:
619 // CHECK1-NEXT:    br label [[CANCEL_EXIT:%.*]]
620 // CHECK1:       .cancel.continue:
621 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4
622 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[R3]], align 4
623 // CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP20]]
624 // CHECK1-NEXT:    store i32 [[ADD7]], i32* [[R3]], align 4
625 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
626 // CHECK1:       omp.body.continue:
627 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
628 // CHECK1:       omp.inner.for.inc:
629 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
630 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
631 // CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
632 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
633 // CHECK1:       omp.inner.for.end:
634 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
635 // CHECK1:       omp.loop.exit:
636 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
637 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
638 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP24]])
639 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
640 // CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i32* [[R3]] to i8*
641 // CHECK1-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
642 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
643 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
644 // CHECK1-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
645 // CHECK1-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB7:[0-9]+]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
646 // CHECK1-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
647 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
648 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
649 // CHECK1-NEXT:    ]
650 // CHECK1:       .omp.reduction.case1:
651 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP1]], align 4
652 // CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[R3]], align 4
653 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
654 // CHECK1-NEXT:    store i32 [[ADD9]], i32* [[TMP1]], align 4
655 // CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB7]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
656 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
657 // CHECK1:       cancel.exit:
658 // CHECK1-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
659 // CHECK1-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
660 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP34]])
661 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
662 // CHECK1:       .omp.reduction.case2:
663 // CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[R3]], align 4
664 // CHECK1-NEXT:    [[TMP36:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP35]] monotonic, align 4
665 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
666 // CHECK1:       .omp.reduction.default:
667 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
668 // CHECK1:       omp.precond.end:
669 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
670 // CHECK1:       cancel.cont:
671 // CHECK1-NEXT:    ret void
672 //
673 //
674 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
675 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
676 // CHECK1-NEXT:  entry:
677 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
678 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
679 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
680 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
681 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
682 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
683 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
684 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
685 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
686 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
687 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
688 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
689 // CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
690 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
691 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
692 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
693 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
694 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
695 // CHECK1-NEXT:    ret void
696 //
697 //
698 // CHECK2-LABEL: define {{[^@]+}}@main
699 // CHECK2-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
700 // CHECK2-NEXT:  entry:
701 // CHECK2-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
702 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
703 // CHECK2-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
704 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
705 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
706 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
707 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
708 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
709 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_LB_1:%.*]] = alloca i32, align 4
710 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_UB_2:%.*]] = alloca i32, align 4
711 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_ST_3:%.*]] = alloca i32, align 4
712 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IL_4:%.*]] = alloca i32, align 4
713 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IV_5:%.*]] = alloca i32, align 4
714 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
715 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
716 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
717 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_21:%.*]] = alloca i32, align 4
718 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
719 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
720 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
721 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
722 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
723 // CHECK2-NEXT:    [[I24:%.*]] = alloca i32, align 4
724 // CHECK2-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
725 // CHECK2-NEXT:    [[R:%.*]] = alloca i32, align 4
726 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
727 // CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
728 // CHECK2-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
729 // CHECK2-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
730 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i8***, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i8*** [[ARGV_ADDR]], i32* [[ARGC_ADDR]])
731 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
732 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
733 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
734 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
735 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
736 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
737 // CHECK2-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
738 // CHECK2-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
739 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
740 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
741 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
742 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
743 // CHECK2:       omp.inner.for.cond:
744 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
745 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
746 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
747 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
748 // CHECK2:       omp.inner.for.body:
749 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
750 // CHECK2-NEXT:    switch i32 [[TMP7]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
751 // CHECK2-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
752 // CHECK2-NEXT:    ]
753 // CHECK2:       .omp.sections.case:
754 // CHECK2-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
755 // CHECK2-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
756 // CHECK2-NEXT:    br i1 [[TMP9]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
757 // CHECK2:       .cancel.exit:
758 // CHECK2-NEXT:    br label [[CANCEL_EXIT:%.*]]
759 // CHECK2:       .cancel.continue:
760 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
761 // CHECK2:       .omp.sections.exit:
762 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
763 // CHECK2:       omp.inner.for.inc:
764 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
765 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP10]], 1
766 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
767 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
768 // CHECK2:       omp.inner.for.end:
769 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
770 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
771 // CHECK2:       cancel.cont:
772 // CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
773 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
774 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
775 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_3]], align 4
776 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_4]], align 4
777 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_4]], i32* [[DOTOMP_SECTIONS_LB_1]], i32* [[DOTOMP_SECTIONS_UB_2]], i32* [[DOTOMP_SECTIONS_ST_3]], i32 1, i32 1)
778 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
779 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp slt i32 [[TMP11]], 1
780 // CHECK2-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 1
781 // CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_SECTIONS_UB_2]], align 4
782 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
783 // CHECK2-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
784 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND6:%.*]]
785 // CHECK2:       omp.inner.for.cond6:
786 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
787 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
788 // CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
789 // CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY8:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
790 // CHECK2:       omp.inner.for.body8:
791 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
792 // CHECK2-NEXT:    switch i32 [[TMP17]], label [[DOTOMP_SECTIONS_EXIT15:%.*]] [
793 // CHECK2-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE9:%.*]]
794 // CHECK2-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE12:%.*]]
795 // CHECK2-NEXT:    ]
796 // CHECK2:       .omp.sections.case9:
797 // CHECK2-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
798 // CHECK2-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
799 // CHECK2-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT10:%.*]], label [[DOTCANCEL_CONTINUE11:%.*]]
800 // CHECK2:       .cancel.exit10:
801 // CHECK2-NEXT:    br label [[CANCEL_EXIT19:%.*]]
802 // CHECK2:       cancel.exit:
803 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
804 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
805 // CHECK2:       .cancel.continue11:
806 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
807 // CHECK2:       .omp.sections.case12:
808 // CHECK2-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
809 // CHECK2-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
810 // CHECK2-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT13:%.*]], label [[DOTCANCEL_CONTINUE14:%.*]]
811 // CHECK2:       .cancel.exit13:
812 // CHECK2-NEXT:    br label [[CANCEL_EXIT19]]
813 // CHECK2:       .cancel.continue14:
814 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
815 // CHECK2:       .omp.sections.exit15:
816 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
817 // CHECK2:       omp.inner.for.inc16:
818 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
819 // CHECK2-NEXT:    [[INC17:%.*]] = add nsw i32 [[TMP22]], 1
820 // CHECK2-NEXT:    store i32 [[INC17]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
821 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND6]]
822 // CHECK2:       omp.inner.for.end18:
823 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
824 // CHECK2-NEXT:    br label [[CANCEL_CONT20:%.*]]
825 // CHECK2:       cancel.cont20:
826 // CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
827 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
828 // CHECK2-NEXT:    store i32 [[TMP23]], i32* [[DOTCAPTURE_EXPR_]], align 4
829 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
830 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
831 // CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
832 // CHECK2-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[DIV]], 1
833 // CHECK2-NEXT:    store i32 [[SUB22]], i32* [[DOTCAPTURE_EXPR_21]], align 4
834 // CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
835 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
836 // CHECK2-NEXT:    [[CMP23:%.*]] = icmp slt i32 0, [[TMP25]]
837 // CHECK2-NEXT:    br i1 [[CMP23]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
838 // CHECK2:       omp.precond.then:
839 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
840 // CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
841 // CHECK2-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_UB]], align 4
842 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
843 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
844 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
845 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
846 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
847 // CHECK2-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
848 // CHECK2-NEXT:    br i1 [[CMP25]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
849 // CHECK2:       cond.true:
850 // CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
851 // CHECK2-NEXT:    br label [[COND_END:%.*]]
852 // CHECK2:       cond.false:
853 // CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
854 // CHECK2-NEXT:    br label [[COND_END]]
855 // CHECK2:       cond.end:
856 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE]] ], [ [[TMP30]], [[COND_FALSE]] ]
857 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
858 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
859 // CHECK2-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
860 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND26:%.*]]
861 // CHECK2:       omp.inner.for.cond26:
862 // CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
863 // CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
864 // CHECK2-NEXT:    [[CMP27:%.*]] = icmp sle i32 [[TMP32]], [[TMP33]]
865 // CHECK2-NEXT:    br i1 [[CMP27]], label [[OMP_INNER_FOR_BODY28:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
866 // CHECK2:       omp.inner.for.body28:
867 // CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
868 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP34]], 1
869 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
870 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I24]], align 4
871 // CHECK2-NEXT:    [[TMP35:%.*]] = load float, float* @flag, align 4
872 // CHECK2-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP35]], 0.000000e+00
873 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
874 // CHECK2:       omp_if.then:
875 // CHECK2-NEXT:    [[TMP36:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
876 // CHECK2-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
877 // CHECK2-NEXT:    br i1 [[TMP37]], label [[DOTCANCEL_EXIT29:%.*]], label [[DOTCANCEL_CONTINUE30:%.*]]
878 // CHECK2:       .cancel.exit29:
879 // CHECK2-NEXT:    br label [[CANCEL_EXIT34:%.*]]
880 // CHECK2:       cancel.exit19:
881 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
882 // CHECK2-NEXT:    br label [[CANCEL_CONT20]]
883 // CHECK2:       .cancel.continue30:
884 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
885 // CHECK2:       omp_if.else:
886 // CHECK2-NEXT:    br label [[OMP_IF_END]]
887 // CHECK2:       omp_if.end:
888 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
889 // CHECK2:       omp.body.continue:
890 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
891 // CHECK2:       omp.inner.for.inc31:
892 // CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
893 // CHECK2-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP38]], 1
894 // CHECK2-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV]], align 4
895 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND26]]
896 // CHECK2:       omp.inner.for.end33:
897 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
898 // CHECK2:       omp.loop.exit:
899 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
900 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
901 // CHECK2:       cancel.exit34:
902 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
903 // CHECK2-NEXT:    br label [[CANCEL_CONT35:%.*]]
904 // CHECK2:       omp.precond.end:
905 // CHECK2-NEXT:    br label [[CANCEL_CONT35]]
906 // CHECK2:       cancel.cont35:
907 // CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]])
908 // CHECK2-NEXT:    [[TMP39:%.*]] = call i8* @__kmpc_omp_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*)*))
909 // CHECK2-NEXT:    [[TMP40:%.*]] = bitcast i8* [[TMP39]] to %struct.kmp_task_t_with_privates*
910 // CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP40]], i32 0, i32 0
911 // CHECK2-NEXT:    [[TMP42:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP39]])
912 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
913 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*))
914 // CHECK2-NEXT:    store i32 0, i32* [[R]], align 4
915 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])
916 // CHECK2-NEXT:    [[TMP43:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
917 // CHECK2-NEXT:    ret i32 [[TMP43]]
918 //
919 //
920 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
921 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8*** nonnull align 8 dereferenceable(8) [[ARGV:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]]) #[[ATTR1:[0-9]+]] {
922 // CHECK2-NEXT:  entry:
923 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
924 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
925 // CHECK2-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8***, align 8
926 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
927 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
928 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
929 // CHECK2-NEXT:    store i8*** [[ARGV]], i8**** [[ARGV_ADDR]], align 8
930 // CHECK2-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
931 // CHECK2-NEXT:    [[TMP0:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
932 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
933 // CHECK2-NEXT:    [[TMP2:%.*]] = load float, float* @flag, align 4
934 // CHECK2-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP2]], 0.000000e+00
935 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
936 // CHECK2:       omp_if.then:
937 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
938 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
939 // CHECK2-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 1)
940 // CHECK2-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[TMP5]], 0
941 // CHECK2-NEXT:    br i1 [[TMP6]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
942 // CHECK2:       .cancel.exit:
943 // CHECK2-NEXT:    br label [[RETURN:%.*]]
944 // CHECK2:       .cancel.continue:
945 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
946 // CHECK2:       omp_if.else:
947 // CHECK2-NEXT:    br label [[OMP_IF_END]]
948 // CHECK2:       omp_if.end:
949 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP1]], align 4
950 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP7]] to i8
951 // CHECK2-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP0]], align 8
952 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 0
953 // CHECK2-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
954 // CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 0
955 // CHECK2-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX1]], align 1
956 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
957 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
958 // CHECK2-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP11]])
959 // CHECK2-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
960 // CHECK2-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
961 // CHECK2:       .cancel.exit2:
962 // CHECK2-NEXT:    br label [[RETURN]]
963 // CHECK2:       .cancel.continue3:
964 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP1]], align 4
965 // CHECK2-NEXT:    [[TMP15:%.*]] = load i8**, i8*** [[TMP0]], align 8
966 // CHECK2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP15]], i64 0
967 // CHECK2-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
968 // CHECK2-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP16]], i64 0
969 // CHECK2-NEXT:    [[TMP17:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
970 // CHECK2-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP17]] to i32
971 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP14]]
972 // CHECK2-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD]] to i8
973 // CHECK2-NEXT:    store i8 [[CONV7]], i8* [[ARRAYIDX5]], align 1
974 // CHECK2-NEXT:    br label [[RETURN]]
975 // CHECK2:       return:
976 // CHECK2-NEXT:    ret void
977 //
978 //
979 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
980 // CHECK2-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
981 // CHECK2-NEXT:  entry:
982 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
983 // CHECK2-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
984 // CHECK2-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
985 // CHECK2-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
986 // CHECK2-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
987 // CHECK2-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
988 // CHECK2-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
989 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
990 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
991 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
992 // CHECK2-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
993 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
994 // CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
995 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
996 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
997 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
998 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
999 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
1000 // CHECK2-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
1001 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META4:![0-9]+]])
1002 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META7:![0-9]+]])
1003 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META9:![0-9]+]])
1004 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META11:![0-9]+]])
1005 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !13
1006 // CHECK2-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !13
1007 // CHECK2-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !13
1008 // CHECK2-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !13
1009 // CHECK2-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !13
1010 // CHECK2-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
1011 // CHECK2-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
1012 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !13
1013 // CHECK2-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 4) #[[ATTR2:[0-9]+]]
1014 // CHECK2-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
1015 // CHECK2-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
1016 // CHECK2:       .cancel.exit.i:
1017 // CHECK2-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
1018 // CHECK2-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
1019 // CHECK2:       .cancel.continue.i:
1020 // CHECK2-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
1021 // CHECK2-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
1022 // CHECK2:       .omp_outlined..1.exit:
1023 // CHECK2-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
1024 // CHECK2-NEXT:    ret i32 0
1025 //
1026 //
1027 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
1028 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1029 // CHECK2-NEXT:  entry:
1030 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1031 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1032 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
1033 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
1034 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
1035 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
1036 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
1037 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1038 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1039 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1040 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1041 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
1042 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
1043 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1044 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1045 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
1046 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1047 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
1048 // CHECK2-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
1049 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
1050 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1051 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1052 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1053 // CHECK2:       omp.inner.for.cond:
1054 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1055 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1056 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1057 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1058 // CHECK2:       omp.inner.for.body:
1059 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1060 // CHECK2-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
1061 // CHECK2-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
1062 // CHECK2-NEXT:    ]
1063 // CHECK2:       .omp.sections.case:
1064 // CHECK2-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
1065 // CHECK2-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
1066 // CHECK2-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
1067 // CHECK2:       .cancel.exit:
1068 // CHECK2-NEXT:    br label [[CANCEL_EXIT:%.*]]
1069 // CHECK2:       .cancel.continue:
1070 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1071 // CHECK2:       .omp.sections.exit:
1072 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1073 // CHECK2:       omp.inner.for.inc:
1074 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1075 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
1076 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1077 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
1078 // CHECK2:       omp.inner.for.end:
1079 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
1080 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
1081 // CHECK2:       cancel.cont:
1082 // CHECK2-NEXT:    ret void
1083 // CHECK2:       cancel.exit:
1084 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
1085 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
1086 //
1087 //
1088 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
1089 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1090 // CHECK2-NEXT:  entry:
1091 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1092 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1093 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
1094 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
1095 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
1096 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
1097 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
1098 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1099 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1100 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1101 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1102 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
1103 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
1104 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1105 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1106 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
1107 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1108 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
1109 // CHECK2-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
1110 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
1111 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1112 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1113 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1114 // CHECK2:       omp.inner.for.cond:
1115 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1116 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1117 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1118 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1119 // CHECK2:       omp.inner.for.body:
1120 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1121 // CHECK2-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
1122 // CHECK2-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
1123 // CHECK2-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE1:%.*]]
1124 // CHECK2-NEXT:    ]
1125 // CHECK2:       .omp.sections.case:
1126 // CHECK2-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
1127 // CHECK2-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
1128 // CHECK2-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
1129 // CHECK2:       .cancel.exit:
1130 // CHECK2-NEXT:    br label [[CANCEL_EXIT:%.*]]
1131 // CHECK2:       .cancel.continue:
1132 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1133 // CHECK2:       .omp.sections.case1:
1134 // CHECK2-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
1135 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1136 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
1137 // CHECK2:       .cancel.exit2:
1138 // CHECK2-NEXT:    br label [[CANCEL_EXIT]]
1139 // CHECK2:       .cancel.continue3:
1140 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1141 // CHECK2:       .omp.sections.exit:
1142 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1143 // CHECK2:       omp.inner.for.inc:
1144 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1145 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
1146 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1147 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
1148 // CHECK2:       omp.inner.for.end:
1149 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
1150 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
1151 // CHECK2:       cancel.cont:
1152 // CHECK2-NEXT:    ret void
1153 // CHECK2:       cancel.exit:
1154 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
1155 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
1156 //
1157 //
1158 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
1159 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR1]] {
1160 // CHECK2-NEXT:  entry:
1161 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1162 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1163 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
1164 // CHECK2-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
1165 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1166 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1167 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1168 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1169 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
1170 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1171 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1172 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1173 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1174 // CHECK2-NEXT:    [[R3:%.*]] = alloca i32, align 4
1175 // CHECK2-NEXT:    [[I4:%.*]] = alloca i32, align 4
1176 // CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
1177 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1178 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1179 // CHECK2-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
1180 // CHECK2-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
1181 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
1182 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
1183 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
1184 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
1185 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1186 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1187 // CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1188 // CHECK2-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1189 // CHECK2-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1190 // CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
1191 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1192 // CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1193 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1194 // CHECK2:       omp.precond.then:
1195 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1196 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1197 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
1198 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1199 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1200 // CHECK2-NEXT:    store i32 0, i32* [[R3]], align 4
1201 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1202 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
1203 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1204 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1205 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1206 // CHECK2-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
1207 // CHECK2-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1208 // CHECK2:       cond.true:
1209 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1210 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1211 // CHECK2:       cond.false:
1212 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1213 // CHECK2-NEXT:    br label [[COND_END]]
1214 // CHECK2:       cond.end:
1215 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1216 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1217 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1218 // CHECK2-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
1219 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1220 // CHECK2:       omp.inner.for.cond:
1221 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1222 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1223 // CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1224 // CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1225 // CHECK2:       omp.inner.for.body:
1226 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1227 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
1228 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1229 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
1230 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1231 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4
1232 // CHECK2-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP17]], i32 2)
1233 // CHECK2-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
1234 // CHECK2-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
1235 // CHECK2:       .cancel.exit:
1236 // CHECK2-NEXT:    br label [[CANCEL_EXIT:%.*]]
1237 // CHECK2:       .cancel.continue:
1238 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4
1239 // CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[R3]], align 4
1240 // CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP20]]
1241 // CHECK2-NEXT:    store i32 [[ADD7]], i32* [[R3]], align 4
1242 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1243 // CHECK2:       omp.body.continue:
1244 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1245 // CHECK2:       omp.inner.for.inc:
1246 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1247 // CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
1248 // CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
1249 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
1250 // CHECK2:       omp.inner.for.end:
1251 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1252 // CHECK2:       omp.loop.exit:
1253 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1254 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
1255 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP24]])
1256 // CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1257 // CHECK2-NEXT:    [[TMP26:%.*]] = bitcast i32* [[R3]] to i8*
1258 // CHECK2-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
1259 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1260 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
1261 // CHECK2-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
1262 // CHECK2-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB7:[0-9]+]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
1263 // CHECK2-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1264 // CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1265 // CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1266 // CHECK2-NEXT:    ]
1267 // CHECK2:       .omp.reduction.case1:
1268 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP1]], align 4
1269 // CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[R3]], align 4
1270 // CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
1271 // CHECK2-NEXT:    store i32 [[ADD9]], i32* [[TMP1]], align 4
1272 // CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB7]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
1273 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1274 // CHECK2:       cancel.exit:
1275 // CHECK2-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1276 // CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
1277 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP34]])
1278 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
1279 // CHECK2:       .omp.reduction.case2:
1280 // CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[R3]], align 4
1281 // CHECK2-NEXT:    [[TMP36:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP35]] monotonic, align 4
1282 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1283 // CHECK2:       .omp.reduction.default:
1284 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
1285 // CHECK2:       omp.precond.end:
1286 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
1287 // CHECK2:       cancel.cont:
1288 // CHECK2-NEXT:    ret void
1289 //
1290 //
1291 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
1292 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
1293 // CHECK2-NEXT:  entry:
1294 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
1295 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
1296 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1297 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1298 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1299 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
1300 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1301 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
1302 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
1303 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1304 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1305 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
1306 // CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1307 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1308 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
1309 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
1310 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1311 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
1312 // CHECK2-NEXT:    ret void
1313 //
1314 //
1315 // CHECK3-LABEL: define {{[^@]+}}@main
1316 // CHECK3-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1317 // CHECK3-NEXT:  entry:
1318 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1319 // CHECK3-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1320 // CHECK3-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
1321 // CHECK3-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
1322 // CHECK3-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
1323 // CHECK3-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
1324 // CHECK3-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
1325 // CHECK3-NEXT:    [[P_LASTITER26:%.*]] = alloca i32, align 4
1326 // CHECK3-NEXT:    [[P_LOWERBOUND27:%.*]] = alloca i32, align 4
1327 // CHECK3-NEXT:    [[P_UPPERBOUND28:%.*]] = alloca i32, align 4
1328 // CHECK3-NEXT:    [[P_STRIDE29:%.*]] = alloca i32, align 4
1329 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1330 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1331 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1332 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_33:%.*]] = alloca i32, align 4
1333 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1334 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1335 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1336 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1337 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1338 // CHECK3-NEXT:    [[I35:%.*]] = alloca i32, align 4
1339 // CHECK3-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
1340 // CHECK3-NEXT:    [[R:%.*]] = alloca i32, align 4
1341 // CHECK3-NEXT:    store i32 0, i32* [[RETVAL]], align 4
1342 // CHECK3-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1343 // CHECK3-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
1344 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
1345 // CHECK3-NEXT:    br label [[OMP_PARALLEL:%.*]]
1346 // CHECK3:       omp_parallel:
1347 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8***)* @main..omp_par to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i8*** [[ARGV_ADDR]])
1348 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
1349 // CHECK3:       omp.par.outlined.exit:
1350 // CHECK3-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
1351 // CHECK3:       omp.par.exit.split:
1352 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
1353 // CHECK3:       omp_section_loop.preheader:
1354 // CHECK3-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
1355 // CHECK3-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
1356 // CHECK3-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
1357 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1358 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
1359 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
1360 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
1361 // CHECK3-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
1362 // CHECK3-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
1363 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
1364 // CHECK3:       omp_section_loop.header:
1365 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
1366 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
1367 // CHECK3:       omp_section_loop.cond:
1368 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
1369 // CHECK3-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
1370 // CHECK3:       omp_section_loop.body:
1371 // CHECK3-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
1372 // CHECK3-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
1373 // CHECK3-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
1374 // CHECK3-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
1375 // CHECK3-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
1376 // CHECK3-NEXT:    ]
1377 // CHECK3:       omp_section_loop.inc:
1378 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
1379 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
1380 // CHECK3:       omp_section_loop.exit:
1381 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
1382 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1383 // CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
1384 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
1385 // CHECK3:       omp_section_loop.after:
1386 // CHECK3-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
1387 // CHECK3:       omp_sections.end:
1388 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER12:%.*]]
1389 // CHECK3:       omp_section_loop.preheader12:
1390 // CHECK3-NEXT:    store i32 0, i32* [[P_LOWERBOUND27]], align 4
1391 // CHECK3-NEXT:    store i32 1, i32* [[P_UPPERBOUND28]], align 4
1392 // CHECK3-NEXT:    store i32 1, i32* [[P_STRIDE29]], align 4
1393 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM30:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1394 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]], i32 34, i32* [[P_LASTITER26]], i32* [[P_LOWERBOUND27]], i32* [[P_UPPERBOUND28]], i32* [[P_STRIDE29]], i32 1, i32 1)
1395 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND27]], align 4
1396 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND28]], align 4
1397 // CHECK3-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
1398 // CHECK3-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
1399 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13:%.*]]
1400 // CHECK3:       omp_section_loop.header13:
1401 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_IV19:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER12]] ], [ [[OMP_SECTION_LOOP_NEXT21:%.*]], [[OMP_SECTION_LOOP_INC16:%.*]] ]
1402 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_COND14:%.*]]
1403 // CHECK3:       omp_section_loop.cond14:
1404 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_CMP20:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV19]], [[TMP10]]
1405 // CHECK3-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP20]], label [[OMP_SECTION_LOOP_BODY15:%.*]], label [[OMP_SECTION_LOOP_EXIT17:%.*]]
1406 // CHECK3:       omp_section_loop.body15:
1407 // CHECK3-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV19]], [[TMP7]]
1408 // CHECK3-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
1409 // CHECK3-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
1410 // CHECK3-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC16]] [
1411 // CHECK3-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE22:%.*]]
1412 // CHECK3-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE24:%.*]]
1413 // CHECK3-NEXT:    ]
1414 // CHECK3:       omp_section_loop.inc16:
1415 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_NEXT21]] = add nuw i32 [[OMP_SECTION_LOOP_IV19]], 1
1416 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13]]
1417 // CHECK3:       omp_section_loop.exit17:
1418 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]])
1419 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1420 // CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
1421 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER18:%.*]]
1422 // CHECK3:       omp_section_loop.after18:
1423 // CHECK3-NEXT:    br label [[OMP_SECTIONS_END32:%.*]]
1424 // CHECK3:       omp_sections.end32:
1425 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1426 // CHECK3-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
1427 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1428 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
1429 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1430 // CHECK3-NEXT:    [[SUB34:%.*]] = sub nsw i32 [[DIV]], 1
1431 // CHECK3-NEXT:    store i32 [[SUB34]], i32* [[DOTCAPTURE_EXPR_33]], align 4
1432 // CHECK3-NEXT:    store i32 0, i32* [[I]], align 4
1433 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1434 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
1435 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1436 // CHECK3:       omp.precond.then:
1437 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1438 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
1439 // CHECK3-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
1440 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1441 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1442 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM36:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
1443 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM36]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1444 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1445 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
1446 // CHECK3-NEXT:    [[CMP37:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
1447 // CHECK3-NEXT:    br i1 [[CMP37]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1448 // CHECK3:       cond.true:
1449 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
1450 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1451 // CHECK3:       cond.false:
1452 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1453 // CHECK3-NEXT:    br label [[COND_END]]
1454 // CHECK3:       cond.end:
1455 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
1456 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1457 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1458 // CHECK3-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
1459 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1460 // CHECK3:       omp.inner.for.cond:
1461 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1462 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1463 // CHECK3-NEXT:    [[CMP38:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
1464 // CHECK3-NEXT:    br i1 [[CMP38]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1465 // CHECK3:       omp.inner.for.body:
1466 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1467 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
1468 // CHECK3-NEXT:    [[ADD39:%.*]] = add nsw i32 0, [[MUL]]
1469 // CHECK3-NEXT:    store i32 [[ADD39]], i32* [[I35]], align 4
1470 // CHECK3-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
1471 // CHECK3-NEXT:    [[TOBOOL40:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
1472 // CHECK3-NEXT:    br i1 [[TOBOOL40]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1473 // CHECK3:       omp_if.then:
1474 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM41:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
1475 // CHECK3-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM41]], i32 2)
1476 // CHECK3-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
1477 // CHECK3-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
1478 // CHECK3:       .cancel.exit:
1479 // CHECK3-NEXT:    br label [[CANCEL_EXIT:%.*]]
1480 // CHECK3:       omp_section_loop.body.case:
1481 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM9:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1482 // CHECK3-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM9]], i32 3)
1483 // CHECK3-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
1484 // CHECK3-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
1485 // CHECK3:       omp_section_loop.body.case.split:
1486 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
1487 // CHECK3:       omp_section_loop.body.case.cncl:
1488 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
1489 // CHECK3:       omp_section_loop.body.case22:
1490 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM23:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1491 // CHECK3-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]], i32 3)
1492 // CHECK3-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
1493 // CHECK3-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE22_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE22_CNCL:%.*]]
1494 // CHECK3:       omp_section_loop.body.case22.split:
1495 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
1496 // CHECK3:       omp_section_loop.body.case22.cncl:
1497 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
1498 // CHECK3:       omp_section_loop.body.case24:
1499 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM25:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1500 // CHECK3-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM25]], i32 3)
1501 // CHECK3-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
1502 // CHECK3-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE24_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE24_CNCL:%.*]]
1503 // CHECK3:       omp_section_loop.body.case24.split:
1504 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
1505 // CHECK3:       omp_section_loop.body.case24.cncl:
1506 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
1507 // CHECK3:       .cancel.continue:
1508 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
1509 // CHECK3:       omp_if.else:
1510 // CHECK3-NEXT:    br label [[OMP_IF_END]]
1511 // CHECK3:       omp_if.end:
1512 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1513 // CHECK3:       omp.body.continue:
1514 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1515 // CHECK3:       omp.inner.for.inc:
1516 // CHECK3-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1517 // CHECK3-NEXT:    [[ADD42:%.*]] = add nsw i32 [[TMP35]], 1
1518 // CHECK3-NEXT:    store i32 [[ADD42]], i32* [[DOTOMP_IV]], align 4
1519 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1520 // CHECK3:       omp.inner.for.end:
1521 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1522 // CHECK3:       omp.loop.exit:
1523 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
1524 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
1525 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
1526 // CHECK3:       cancel.exit:
1527 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM43:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
1528 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM43]])
1529 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
1530 // CHECK3:       omp.precond.end:
1531 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1532 // CHECK3:       cancel.cont:
1533 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1534 // CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
1535 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
1536 // CHECK3-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM46]], 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*)*))
1537 // CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
1538 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
1539 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
1540 // CHECK3-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], i8* [[TMP36]])
1541 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
1542 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
1543 // CHECK3-NEXT:    store i32 0, i32* [[R]], align 4
1544 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])
1545 // CHECK3-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1546 // CHECK3-NEXT:    ret i32 [[TMP40]]
1547 //
1548 //
1549 // CHECK3-LABEL: define {{[^@]+}}@main..omp_par
1550 // CHECK3-SAME: (i32* noalias [[TID_ADDR:%.*]], i32* noalias [[ZERO_ADDR:%.*]], i32* [[ARGC_ADDR:%.*]], i8*** [[ARGV_ADDR:%.*]]) #[[ATTR1:[0-9]+]] {
1551 // CHECK3-NEXT:  omp.par.entry:
1552 // CHECK3-NEXT:    [[TID_ADDR_LOCAL:%.*]] = alloca i32, align 4
1553 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[TID_ADDR]], align 4
1554 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[TID_ADDR_LOCAL]], align 4
1555 // CHECK3-NEXT:    [[TID:%.*]] = load i32, i32* [[TID_ADDR_LOCAL]], align 4
1556 // CHECK3-NEXT:    br label [[OMP_PAR_REGION:%.*]]
1557 // CHECK3:       omp.par.outlined.exit.exitStub:
1558 // CHECK3-NEXT:    ret void
1559 // CHECK3:       omp.par.region:
1560 // CHECK3-NEXT:    [[TMP1:%.*]] = load float, float* @flag, align 4
1561 // CHECK3-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP1]], 0.000000e+00
1562 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[TMP13:%.*]], label [[TMP2:%.*]]
1563 // CHECK3:       2:
1564 // CHECK3-NEXT:    br label [[TMP3:%.*]]
1565 // CHECK3:       3:
1566 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1567 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP4]] to i8
1568 // CHECK3-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1569 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 0
1570 // CHECK3-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
1571 // CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
1572 // CHECK3-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX2]], align 1
1573 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1574 // CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
1575 // CHECK3-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
1576 // CHECK3-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL4:%.*]]
1577 // CHECK3:       .cncl4:
1578 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
1579 // CHECK3:       .cont:
1580 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1581 // CHECK3-NEXT:    [[TMP10:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1582 // CHECK3-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
1583 // CHECK3-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX5]], align 8
1584 // CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
1585 // CHECK3-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX6]], align 1
1586 // CHECK3-NEXT:    [[CONV7:%.*]] = sext i8 [[TMP12]] to i32
1587 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV7]], [[TMP9]]
1588 // CHECK3-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD]] to i8
1589 // CHECK3-NEXT:    store i8 [[CONV8]], i8* [[ARRAYIDX6]], align 1
1590 // CHECK3-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
1591 // CHECK3:       omp.par.pre_finalize:
1592 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
1593 // CHECK3:       13:
1594 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1595 // CHECK3-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
1596 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 0
1597 // CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
1598 // CHECK3:       .cncl:
1599 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
1600 // CHECK3:       .split:
1601 // CHECK3-NEXT:    br label [[TMP3]]
1602 //
1603 //
1604 // CHECK3-LABEL: define {{[^@]+}}@.omp_task_entry.
1605 // CHECK3-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
1606 // CHECK3-NEXT:  entry:
1607 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
1608 // CHECK3-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
1609 // CHECK3-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
1610 // CHECK3-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
1611 // CHECK3-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
1612 // CHECK3-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
1613 // CHECK3-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
1614 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
1615 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
1616 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
1617 // CHECK3-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
1618 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
1619 // CHECK3-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
1620 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
1621 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
1622 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
1623 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1624 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
1625 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
1626 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META4:![0-9]+]])
1627 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META7:![0-9]+]])
1628 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META9:![0-9]+]])
1629 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META11:![0-9]+]])
1630 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !13
1631 // CHECK3-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !13
1632 // CHECK3-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !13
1633 // CHECK3-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !13
1634 // CHECK3-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !13
1635 // CHECK3-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
1636 // CHECK3-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
1637 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM_I:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]]) #[[ATTR2:[0-9]+]]
1638 // CHECK3-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM_I]], i32 4) #[[ATTR2]]
1639 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1640 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
1641 // CHECK3:       .cancel.exit.i:
1642 // CHECK3-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
1643 // CHECK3-NEXT:    br label [[DOTOMP_OUTLINED__EXIT:%.*]]
1644 // CHECK3:       .cancel.continue.i:
1645 // CHECK3-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
1646 // CHECK3-NEXT:    br label [[DOTOMP_OUTLINED__EXIT]]
1647 // CHECK3:       .omp_outlined..exit:
1648 // CHECK3-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
1649 // CHECK3-NEXT:    ret i32 0
1650 //
1651 //
1652 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
1653 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5:[0-9]+]] {
1654 // CHECK3-NEXT:  entry:
1655 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1656 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1657 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
1658 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
1659 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
1660 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
1661 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
1662 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1663 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1664 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1665 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1666 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
1667 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
1668 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB17:[0-9]+]])
1669 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB15:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
1670 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1671 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 0
1672 // CHECK3-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 0
1673 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
1674 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1675 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1676 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1677 // CHECK3:       omp.inner.for.cond:
1678 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1679 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1680 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
1681 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1682 // CHECK3:       omp.inner.for.body:
1683 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1684 // CHECK3-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
1685 // CHECK3-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
1686 // CHECK3-NEXT:    ]
1687 // CHECK3:       .omp.sections.case:
1688 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1689 // CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
1690 // CHECK3-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
1691 // CHECK3-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
1692 // CHECK3:       .omp.sections.case.split:
1693 // CHECK3-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1694 // CHECK3:       .omp.sections.case.cncl:
1695 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
1696 // CHECK3:       .omp.sections.exit:
1697 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1698 // CHECK3:       omp.inner.for.inc:
1699 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1700 // CHECK3-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
1701 // CHECK3-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1702 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1703 // CHECK3:       omp.inner.for.end:
1704 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19:[0-9]+]])
1705 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
1706 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1707 // CHECK3:       cancel.cont:
1708 // CHECK3-NEXT:    ret void
1709 // CHECK3:       cancel.exit:
1710 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19]])
1711 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
1712 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1713 //
1714 //
1715 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
1716 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5]] {
1717 // CHECK3-NEXT:  entry:
1718 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1719 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1720 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
1721 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
1722 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
1723 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
1724 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
1725 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1726 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1727 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1728 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1729 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
1730 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
1731 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB21:[0-9]+]])
1732 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
1733 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1734 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 1
1735 // CHECK3-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 1
1736 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
1737 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1738 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1739 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1740 // CHECK3:       omp.inner.for.cond:
1741 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1742 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1743 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
1744 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1745 // CHECK3:       omp.inner.for.body:
1746 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1747 // CHECK3-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
1748 // CHECK3-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
1749 // CHECK3-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE2:%.*]]
1750 // CHECK3-NEXT:    ]
1751 // CHECK3:       .omp.sections.case:
1752 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1753 // CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
1754 // CHECK3-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
1755 // CHECK3-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
1756 // CHECK3:       .omp.sections.case.split:
1757 // CHECK3-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1758 // CHECK3:       .omp.sections.case.cncl:
1759 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
1760 // CHECK3:       .omp.sections.case2:
1761 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1762 // CHECK3-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
1763 // CHECK3-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
1764 // CHECK3-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
1765 // CHECK3:       .omp.sections.case2.split:
1766 // CHECK3-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1767 // CHECK3:       .omp.sections.case2.cncl:
1768 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END]]
1769 // CHECK3:       .omp.sections.exit:
1770 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1771 // CHECK3:       omp.inner.for.inc:
1772 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1773 // CHECK3-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
1774 // CHECK3-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1775 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1776 // CHECK3:       omp.inner.for.end:
1777 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM5:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23:[0-9]+]])
1778 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM5]])
1779 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1780 // CHECK3:       cancel.cont:
1781 // CHECK3-NEXT:    ret void
1782 // CHECK3:       cancel.exit:
1783 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23]])
1784 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
1785 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1786 //
1787 //
1788 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
1789 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR5]] {
1790 // CHECK3-NEXT:  entry:
1791 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1792 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1793 // CHECK3-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
1794 // CHECK3-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
1795 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1796 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1797 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1798 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1799 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1800 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1801 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1802 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1803 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1804 // CHECK3-NEXT:    [[R3:%.*]] = alloca i32, align 4
1805 // CHECK3-NEXT:    [[I4:%.*]] = alloca i32, align 4
1806 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
1807 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1808 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1809 // CHECK3-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
1810 // CHECK3-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
1811 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
1812 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
1813 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
1814 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
1815 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1816 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1817 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1818 // CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1819 // CHECK3-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1820 // CHECK3-NEXT:    store i32 0, i32* [[I]], align 4
1821 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1822 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1823 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1824 // CHECK3:       omp.precond.then:
1825 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1826 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1827 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
1828 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1829 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1830 // CHECK3-NEXT:    store i32 0, i32* [[R3]], align 4
1831 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB25:[0-9]+]])
1832 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1833 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1834 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1835 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
1836 // CHECK3-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1837 // CHECK3:       cond.true:
1838 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1839 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1840 // CHECK3:       cond.false:
1841 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1842 // CHECK3-NEXT:    br label [[COND_END]]
1843 // CHECK3:       cond.end:
1844 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
1845 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1846 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1847 // CHECK3-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
1848 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1849 // CHECK3:       omp.inner.for.cond:
1850 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1851 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1852 // CHECK3-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
1853 // CHECK3-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1854 // CHECK3:       omp.inner.for.body:
1855 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1856 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1
1857 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1858 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
1859 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM7:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB27:[0-9]+]])
1860 // CHECK3-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM7]], i32 2)
1861 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
1862 // CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
1863 // CHECK3:       .cancel.exit:
1864 // CHECK3-NEXT:    br label [[CANCEL_EXIT:%.*]]
1865 // CHECK3:       .cancel.continue:
1866 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I4]], align 4
1867 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[R3]], align 4
1868 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP16]]
1869 // CHECK3-NEXT:    store i32 [[ADD8]], i32* [[R3]], align 4
1870 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1871 // CHECK3:       omp.body.continue:
1872 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1873 // CHECK3:       omp.inner.for.inc:
1874 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1875 // CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], 1
1876 // CHECK3-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
1877 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1878 // CHECK3:       omp.inner.for.end:
1879 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1880 // CHECK3:       omp.loop.exit:
1881 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29:[0-9]+]])
1882 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
1883 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1884 // CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i32* [[R3]] to i8*
1885 // CHECK3-NEXT:    store i8* [[TMP20]], i8** [[TMP19]], align 8
1886 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
1887 // CHECK3-NEXT:    [[TMP21:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
1888 // CHECK3-NEXT:    [[TMP22:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB30:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]], i32 1, i64 8, i8* [[TMP21]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
1889 // CHECK3-NEXT:    switch i32 [[TMP22]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1890 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1891 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1892 // CHECK3-NEXT:    ]
1893 // CHECK3:       .omp.reduction.case1:
1894 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP1]], align 4
1895 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[R3]], align 4
1896 // CHECK3-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
1897 // CHECK3-NEXT:    store i32 [[ADD13]], i32* [[TMP1]], align 4
1898 // CHECK3-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB30]], i32 [[OMP_GLOBAL_THREAD_NUM12]], [8 x i32]* @.gomp_critical_user_.reduction.var)
1899 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1900 // CHECK3:       cancel.exit:
1901 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
1902 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
1903 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
1904 // CHECK3:       .omp.reduction.case2:
1905 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[R3]], align 4
1906 // CHECK3-NEXT:    [[TMP26:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP25]] monotonic, align 4
1907 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1908 // CHECK3:       .omp.reduction.default:
1909 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
1910 // CHECK3:       omp.precond.end:
1911 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1912 // CHECK3:       cancel.cont:
1913 // CHECK3-NEXT:    ret void
1914 //
1915 //
1916 // CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
1917 // CHECK3-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
1918 // CHECK3-NEXT:  entry:
1919 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
1920 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
1921 // CHECK3-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1922 // CHECK3-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1923 // CHECK3-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1924 // CHECK3-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
1925 // CHECK3-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1926 // CHECK3-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
1927 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
1928 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1929 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1930 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
1931 // CHECK3-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1932 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1933 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
1934 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
1935 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1936 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
1937 // CHECK3-NEXT:    ret void
1938 //
1939 //
1940 // CHECK4-LABEL: define {{[^@]+}}@main
1941 // CHECK4-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1942 // CHECK4-NEXT:  entry:
1943 // CHECK4-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1944 // CHECK4-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1945 // CHECK4-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
1946 // CHECK4-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
1947 // CHECK4-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
1948 // CHECK4-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
1949 // CHECK4-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
1950 // CHECK4-NEXT:    [[P_LASTITER26:%.*]] = alloca i32, align 4
1951 // CHECK4-NEXT:    [[P_LOWERBOUND27:%.*]] = alloca i32, align 4
1952 // CHECK4-NEXT:    [[P_UPPERBOUND28:%.*]] = alloca i32, align 4
1953 // CHECK4-NEXT:    [[P_STRIDE29:%.*]] = alloca i32, align 4
1954 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1955 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1956 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1957 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_33:%.*]] = alloca i32, align 4
1958 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
1959 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1960 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1961 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1962 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1963 // CHECK4-NEXT:    [[I35:%.*]] = alloca i32, align 4
1964 // CHECK4-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
1965 // CHECK4-NEXT:    [[R:%.*]] = alloca i32, align 4
1966 // CHECK4-NEXT:    store i32 0, i32* [[RETVAL]], align 4
1967 // CHECK4-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1968 // CHECK4-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
1969 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
1970 // CHECK4-NEXT:    br label [[OMP_PARALLEL:%.*]]
1971 // CHECK4:       omp_parallel:
1972 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8***)* @main..omp_par to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i8*** [[ARGV_ADDR]])
1973 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
1974 // CHECK4:       omp.par.outlined.exit:
1975 // CHECK4-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
1976 // CHECK4:       omp.par.exit.split:
1977 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
1978 // CHECK4:       omp_section_loop.preheader:
1979 // CHECK4-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
1980 // CHECK4-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
1981 // CHECK4-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
1982 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1983 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
1984 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
1985 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
1986 // CHECK4-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
1987 // CHECK4-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
1988 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
1989 // CHECK4:       omp_section_loop.header:
1990 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
1991 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
1992 // CHECK4:       omp_section_loop.cond:
1993 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
1994 // CHECK4-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
1995 // CHECK4:       omp_section_loop.body:
1996 // CHECK4-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
1997 // CHECK4-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
1998 // CHECK4-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
1999 // CHECK4-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
2000 // CHECK4-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
2001 // CHECK4-NEXT:    ]
2002 // CHECK4:       omp_section_loop.inc:
2003 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
2004 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
2005 // CHECK4:       omp_section_loop.exit:
2006 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
2007 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2008 // CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
2009 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
2010 // CHECK4:       omp_section_loop.after:
2011 // CHECK4-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
2012 // CHECK4:       omp_sections.end:
2013 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER12:%.*]]
2014 // CHECK4:       omp_section_loop.preheader12:
2015 // CHECK4-NEXT:    store i32 0, i32* [[P_LOWERBOUND27]], align 4
2016 // CHECK4-NEXT:    store i32 1, i32* [[P_UPPERBOUND28]], align 4
2017 // CHECK4-NEXT:    store i32 1, i32* [[P_STRIDE29]], align 4
2018 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM30:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2019 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]], i32 34, i32* [[P_LASTITER26]], i32* [[P_LOWERBOUND27]], i32* [[P_UPPERBOUND28]], i32* [[P_STRIDE29]], i32 1, i32 1)
2020 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND27]], align 4
2021 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND28]], align 4
2022 // CHECK4-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
2023 // CHECK4-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
2024 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13:%.*]]
2025 // CHECK4:       omp_section_loop.header13:
2026 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_IV19:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER12]] ], [ [[OMP_SECTION_LOOP_NEXT21:%.*]], [[OMP_SECTION_LOOP_INC16:%.*]] ]
2027 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_COND14:%.*]]
2028 // CHECK4:       omp_section_loop.cond14:
2029 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_CMP20:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV19]], [[TMP10]]
2030 // CHECK4-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP20]], label [[OMP_SECTION_LOOP_BODY15:%.*]], label [[OMP_SECTION_LOOP_EXIT17:%.*]]
2031 // CHECK4:       omp_section_loop.body15:
2032 // CHECK4-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV19]], [[TMP7]]
2033 // CHECK4-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
2034 // CHECK4-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
2035 // CHECK4-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC16]] [
2036 // CHECK4-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE22:%.*]]
2037 // CHECK4-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE24:%.*]]
2038 // CHECK4-NEXT:    ]
2039 // CHECK4:       omp_section_loop.inc16:
2040 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_NEXT21]] = add nuw i32 [[OMP_SECTION_LOOP_IV19]], 1
2041 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13]]
2042 // CHECK4:       omp_section_loop.exit17:
2043 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]])
2044 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2045 // CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
2046 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_AFTER18:%.*]]
2047 // CHECK4:       omp_section_loop.after18:
2048 // CHECK4-NEXT:    br label [[OMP_SECTIONS_END32:%.*]]
2049 // CHECK4:       omp_sections.end32:
2050 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2051 // CHECK4-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
2052 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2053 // CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
2054 // CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2055 // CHECK4-NEXT:    [[SUB34:%.*]] = sub nsw i32 [[DIV]], 1
2056 // CHECK4-NEXT:    store i32 [[SUB34]], i32* [[DOTCAPTURE_EXPR_33]], align 4
2057 // CHECK4-NEXT:    store i32 0, i32* [[I]], align 4
2058 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2059 // CHECK4-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
2060 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2061 // CHECK4:       omp.precond.then:
2062 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2063 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
2064 // CHECK4-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
2065 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2066 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2067 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM36:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
2068 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM36]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2069 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2070 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
2071 // CHECK4-NEXT:    [[CMP37:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
2072 // CHECK4-NEXT:    br i1 [[CMP37]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2073 // CHECK4:       cond.true:
2074 // CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
2075 // CHECK4-NEXT:    br label [[COND_END:%.*]]
2076 // CHECK4:       cond.false:
2077 // CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2078 // CHECK4-NEXT:    br label [[COND_END]]
2079 // CHECK4:       cond.end:
2080 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
2081 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2082 // CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2083 // CHECK4-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
2084 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2085 // CHECK4:       omp.inner.for.cond:
2086 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2087 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2088 // CHECK4-NEXT:    [[CMP38:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
2089 // CHECK4-NEXT:    br i1 [[CMP38]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2090 // CHECK4:       omp.inner.for.body:
2091 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2092 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
2093 // CHECK4-NEXT:    [[ADD39:%.*]] = add nsw i32 0, [[MUL]]
2094 // CHECK4-NEXT:    store i32 [[ADD39]], i32* [[I35]], align 4
2095 // CHECK4-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
2096 // CHECK4-NEXT:    [[TOBOOL40:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
2097 // CHECK4-NEXT:    br i1 [[TOBOOL40]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2098 // CHECK4:       omp_if.then:
2099 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM41:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
2100 // CHECK4-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM41]], i32 2)
2101 // CHECK4-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
2102 // CHECK4-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
2103 // CHECK4:       .cancel.exit:
2104 // CHECK4-NEXT:    br label [[CANCEL_EXIT:%.*]]
2105 // CHECK4:       omp_section_loop.body.case:
2106 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM9:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2107 // CHECK4-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM9]], i32 3)
2108 // CHECK4-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
2109 // CHECK4-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
2110 // CHECK4:       omp_section_loop.body.case.split:
2111 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
2112 // CHECK4:       omp_section_loop.body.case.cncl:
2113 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
2114 // CHECK4:       omp_section_loop.body.case22:
2115 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM23:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2116 // CHECK4-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]], i32 3)
2117 // CHECK4-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
2118 // CHECK4-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE22_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE22_CNCL:%.*]]
2119 // CHECK4:       omp_section_loop.body.case22.split:
2120 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
2121 // CHECK4:       omp_section_loop.body.case22.cncl:
2122 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
2123 // CHECK4:       omp_section_loop.body.case24:
2124 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM25:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2125 // CHECK4-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM25]], i32 3)
2126 // CHECK4-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
2127 // CHECK4-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE24_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE24_CNCL:%.*]]
2128 // CHECK4:       omp_section_loop.body.case24.split:
2129 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
2130 // CHECK4:       omp_section_loop.body.case24.cncl:
2131 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
2132 // CHECK4:       .cancel.continue:
2133 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
2134 // CHECK4:       omp_if.else:
2135 // CHECK4-NEXT:    br label [[OMP_IF_END]]
2136 // CHECK4:       omp_if.end:
2137 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2138 // CHECK4:       omp.body.continue:
2139 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2140 // CHECK4:       omp.inner.for.inc:
2141 // CHECK4-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2142 // CHECK4-NEXT:    [[ADD42:%.*]] = add nsw i32 [[TMP35]], 1
2143 // CHECK4-NEXT:    store i32 [[ADD42]], i32* [[DOTOMP_IV]], align 4
2144 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
2145 // CHECK4:       omp.inner.for.end:
2146 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2147 // CHECK4:       omp.loop.exit:
2148 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
2149 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
2150 // CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
2151 // CHECK4:       cancel.exit:
2152 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM43:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
2153 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM43]])
2154 // CHECK4-NEXT:    br label [[CANCEL_CONT:%.*]]
2155 // CHECK4:       omp.precond.end:
2156 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
2157 // CHECK4:       cancel.cont:
2158 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2159 // CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
2160 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
2161 // CHECK4-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM46]], 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*)*))
2162 // CHECK4-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
2163 // CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
2164 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
2165 // CHECK4-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], i8* [[TMP36]])
2166 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
2167 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
2168 // CHECK4-NEXT:    store i32 0, i32* [[R]], align 4
2169 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])
2170 // CHECK4-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2171 // CHECK4-NEXT:    ret i32 [[TMP40]]
2172 //
2173 //
2174 // CHECK4-LABEL: define {{[^@]+}}@main..omp_par
2175 // CHECK4-SAME: (i32* noalias [[TID_ADDR:%.*]], i32* noalias [[ZERO_ADDR:%.*]], i32* [[ARGC_ADDR:%.*]], i8*** [[ARGV_ADDR:%.*]]) #[[ATTR1:[0-9]+]] {
2176 // CHECK4-NEXT:  omp.par.entry:
2177 // CHECK4-NEXT:    [[TID_ADDR_LOCAL:%.*]] = alloca i32, align 4
2178 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[TID_ADDR]], align 4
2179 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[TID_ADDR_LOCAL]], align 4
2180 // CHECK4-NEXT:    [[TID:%.*]] = load i32, i32* [[TID_ADDR_LOCAL]], align 4
2181 // CHECK4-NEXT:    br label [[OMP_PAR_REGION:%.*]]
2182 // CHECK4:       omp.par.outlined.exit.exitStub:
2183 // CHECK4-NEXT:    ret void
2184 // CHECK4:       omp.par.region:
2185 // CHECK4-NEXT:    [[TMP1:%.*]] = load float, float* @flag, align 4
2186 // CHECK4-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP1]], 0.000000e+00
2187 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[TMP13:%.*]], label [[TMP2:%.*]]
2188 // CHECK4:       2:
2189 // CHECK4-NEXT:    br label [[TMP3:%.*]]
2190 // CHECK4:       3:
2191 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2192 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP4]] to i8
2193 // CHECK4-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
2194 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 0
2195 // CHECK4-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
2196 // CHECK4-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
2197 // CHECK4-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX2]], align 1
2198 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2199 // CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
2200 // CHECK4-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
2201 // CHECK4-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL4:%.*]]
2202 // CHECK4:       .cncl4:
2203 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
2204 // CHECK4:       .cont:
2205 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2206 // CHECK4-NEXT:    [[TMP10:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
2207 // CHECK4-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
2208 // CHECK4-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX5]], align 8
2209 // CHECK4-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
2210 // CHECK4-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX6]], align 1
2211 // CHECK4-NEXT:    [[CONV7:%.*]] = sext i8 [[TMP12]] to i32
2212 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV7]], [[TMP9]]
2213 // CHECK4-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD]] to i8
2214 // CHECK4-NEXT:    store i8 [[CONV8]], i8* [[ARRAYIDX6]], align 1
2215 // CHECK4-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
2216 // CHECK4:       omp.par.pre_finalize:
2217 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
2218 // CHECK4:       13:
2219 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2220 // CHECK4-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
2221 // CHECK4-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 0
2222 // CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
2223 // CHECK4:       .cncl:
2224 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
2225 // CHECK4:       .split:
2226 // CHECK4-NEXT:    br label [[TMP3]]
2227 //
2228 //
2229 // CHECK4-LABEL: define {{[^@]+}}@.omp_task_entry.
2230 // CHECK4-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
2231 // CHECK4-NEXT:  entry:
2232 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
2233 // CHECK4-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
2234 // CHECK4-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
2235 // CHECK4-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
2236 // CHECK4-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
2237 // CHECK4-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
2238 // CHECK4-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
2239 // CHECK4-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
2240 // CHECK4-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
2241 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
2242 // CHECK4-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
2243 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
2244 // CHECK4-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
2245 // CHECK4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
2246 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
2247 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
2248 // CHECK4-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2249 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
2250 // CHECK4-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
2251 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META4:![0-9]+]])
2252 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META7:![0-9]+]])
2253 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META9:![0-9]+]])
2254 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META11:![0-9]+]])
2255 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !13
2256 // CHECK4-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !13
2257 // CHECK4-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !13
2258 // CHECK4-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !13
2259 // CHECK4-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !13
2260 // CHECK4-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
2261 // CHECK4-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
2262 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM_I:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]]) #[[ATTR2:[0-9]+]]
2263 // CHECK4-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM_I]], i32 4) #[[ATTR2]]
2264 // CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2265 // CHECK4-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
2266 // CHECK4:       .cancel.exit.i:
2267 // CHECK4-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
2268 // CHECK4-NEXT:    br label [[DOTOMP_OUTLINED__EXIT:%.*]]
2269 // CHECK4:       .cancel.continue.i:
2270 // CHECK4-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
2271 // CHECK4-NEXT:    br label [[DOTOMP_OUTLINED__EXIT]]
2272 // CHECK4:       .omp_outlined..exit:
2273 // CHECK4-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
2274 // CHECK4-NEXT:    ret i32 0
2275 //
2276 //
2277 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
2278 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5:[0-9]+]] {
2279 // CHECK4-NEXT:  entry:
2280 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2281 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2282 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
2283 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
2284 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
2285 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
2286 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
2287 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2288 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2289 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2290 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2291 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
2292 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
2293 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB17:[0-9]+]])
2294 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB15:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
2295 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2296 // CHECK4-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 0
2297 // CHECK4-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 0
2298 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
2299 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2300 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2301 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2302 // CHECK4:       omp.inner.for.cond:
2303 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2304 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2305 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
2306 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2307 // CHECK4:       omp.inner.for.body:
2308 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2309 // CHECK4-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
2310 // CHECK4-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
2311 // CHECK4-NEXT:    ]
2312 // CHECK4:       .omp.sections.case:
2313 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2314 // CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
2315 // CHECK4-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
2316 // CHECK4-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
2317 // CHECK4:       .omp.sections.case.split:
2318 // CHECK4-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
2319 // CHECK4:       .omp.sections.case.cncl:
2320 // CHECK4-NEXT:    br label [[CANCEL_CONT:%.*]]
2321 // CHECK4:       .omp.sections.exit:
2322 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2323 // CHECK4:       omp.inner.for.inc:
2324 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2325 // CHECK4-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
2326 // CHECK4-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2327 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
2328 // CHECK4:       omp.inner.for.end:
2329 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19:[0-9]+]])
2330 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
2331 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
2332 // CHECK4:       cancel.cont:
2333 // CHECK4-NEXT:    ret void
2334 // CHECK4:       cancel.exit:
2335 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19]])
2336 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
2337 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
2338 //
2339 //
2340 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
2341 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5]] {
2342 // CHECK4-NEXT:  entry:
2343 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2344 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2345 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
2346 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
2347 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
2348 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
2349 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
2350 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2351 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2352 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2353 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2354 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
2355 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
2356 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB21:[0-9]+]])
2357 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
2358 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2359 // CHECK4-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 1
2360 // CHECK4-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 1
2361 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
2362 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2363 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2364 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2365 // CHECK4:       omp.inner.for.cond:
2366 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2367 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2368 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
2369 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2370 // CHECK4:       omp.inner.for.body:
2371 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2372 // CHECK4-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
2373 // CHECK4-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
2374 // CHECK4-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE2:%.*]]
2375 // CHECK4-NEXT:    ]
2376 // CHECK4:       .omp.sections.case:
2377 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2378 // CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
2379 // CHECK4-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
2380 // CHECK4-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
2381 // CHECK4:       .omp.sections.case.split:
2382 // CHECK4-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
2383 // CHECK4:       .omp.sections.case.cncl:
2384 // CHECK4-NEXT:    br label [[CANCEL_CONT:%.*]]
2385 // CHECK4:       .omp.sections.case2:
2386 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2387 // CHECK4-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
2388 // CHECK4-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
2389 // CHECK4-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
2390 // CHECK4:       .omp.sections.case2.split:
2391 // CHECK4-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
2392 // CHECK4:       .omp.sections.case2.cncl:
2393 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_END]]
2394 // CHECK4:       .omp.sections.exit:
2395 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2396 // CHECK4:       omp.inner.for.inc:
2397 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2398 // CHECK4-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
2399 // CHECK4-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2400 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
2401 // CHECK4:       omp.inner.for.end:
2402 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM5:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23:[0-9]+]])
2403 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM5]])
2404 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
2405 // CHECK4:       cancel.cont:
2406 // CHECK4-NEXT:    ret void
2407 // CHECK4:       cancel.exit:
2408 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23]])
2409 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
2410 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
2411 //
2412 //
2413 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3
2414 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR5]] {
2415 // CHECK4-NEXT:  entry:
2416 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2417 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2418 // CHECK4-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
2419 // CHECK4-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
2420 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2421 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2422 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2423 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2424 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
2425 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2426 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2427 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2428 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2429 // CHECK4-NEXT:    [[R3:%.*]] = alloca i32, align 4
2430 // CHECK4-NEXT:    [[I4:%.*]] = alloca i32, align 4
2431 // CHECK4-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
2432 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2433 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2434 // CHECK4-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
2435 // CHECK4-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
2436 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
2437 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
2438 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
2439 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
2440 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2441 // CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
2442 // CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2443 // CHECK4-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2444 // CHECK4-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2445 // CHECK4-NEXT:    store i32 0, i32* [[I]], align 4
2446 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2447 // CHECK4-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
2448 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2449 // CHECK4:       omp.precond.then:
2450 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2451 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2452 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
2453 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2454 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2455 // CHECK4-NEXT:    store i32 0, i32* [[R3]], align 4
2456 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB25:[0-9]+]])
2457 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2458 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2459 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2460 // CHECK4-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
2461 // CHECK4-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2462 // CHECK4:       cond.true:
2463 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2464 // CHECK4-NEXT:    br label [[COND_END:%.*]]
2465 // CHECK4:       cond.false:
2466 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2467 // CHECK4-NEXT:    br label [[COND_END]]
2468 // CHECK4:       cond.end:
2469 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
2470 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2471 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2472 // CHECK4-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
2473 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2474 // CHECK4:       omp.inner.for.cond:
2475 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2476 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2477 // CHECK4-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
2478 // CHECK4-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2479 // CHECK4:       omp.inner.for.body:
2480 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2481 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1
2482 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2483 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
2484 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM7:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB27:[0-9]+]])
2485 // CHECK4-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM7]], i32 2)
2486 // CHECK4-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
2487 // CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
2488 // CHECK4:       .cancel.exit:
2489 // CHECK4-NEXT:    br label [[CANCEL_EXIT:%.*]]
2490 // CHECK4:       .cancel.continue:
2491 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I4]], align 4
2492 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[R3]], align 4
2493 // CHECK4-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP16]]
2494 // CHECK4-NEXT:    store i32 [[ADD8]], i32* [[R3]], align 4
2495 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2496 // CHECK4:       omp.body.continue:
2497 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2498 // CHECK4:       omp.inner.for.inc:
2499 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2500 // CHECK4-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], 1
2501 // CHECK4-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
2502 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
2503 // CHECK4:       omp.inner.for.end:
2504 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2505 // CHECK4:       omp.loop.exit:
2506 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29:[0-9]+]])
2507 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
2508 // CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
2509 // CHECK4-NEXT:    [[TMP20:%.*]] = bitcast i32* [[R3]] to i8*
2510 // CHECK4-NEXT:    store i8* [[TMP20]], i8** [[TMP19]], align 8
2511 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
2512 // CHECK4-NEXT:    [[TMP21:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
2513 // CHECK4-NEXT:    [[TMP22:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB30:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]], i32 1, i64 8, i8* [[TMP21]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
2514 // CHECK4-NEXT:    switch i32 [[TMP22]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2515 // CHECK4-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2516 // CHECK4-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2517 // CHECK4-NEXT:    ]
2518 // CHECK4:       .omp.reduction.case1:
2519 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP1]], align 4
2520 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[R3]], align 4
2521 // CHECK4-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
2522 // CHECK4-NEXT:    store i32 [[ADD13]], i32* [[TMP1]], align 4
2523 // CHECK4-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB30]], i32 [[OMP_GLOBAL_THREAD_NUM12]], [8 x i32]* @.gomp_critical_user_.reduction.var)
2524 // CHECK4-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
2525 // CHECK4:       cancel.exit:
2526 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
2527 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
2528 // CHECK4-NEXT:    br label [[CANCEL_CONT:%.*]]
2529 // CHECK4:       .omp.reduction.case2:
2530 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[R3]], align 4
2531 // CHECK4-NEXT:    [[TMP26:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP25]] monotonic, align 4
2532 // CHECK4-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
2533 // CHECK4:       .omp.reduction.default:
2534 // CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
2535 // CHECK4:       omp.precond.end:
2536 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
2537 // CHECK4:       cancel.cont:
2538 // CHECK4-NEXT:    ret void
2539 //
2540 //
2541 // CHECK4-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
2542 // CHECK4-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
2543 // CHECK4-NEXT:  entry:
2544 // CHECK4-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
2545 // CHECK4-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
2546 // CHECK4-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
2547 // CHECK4-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
2548 // CHECK4-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
2549 // CHECK4-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
2550 // CHECK4-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
2551 // CHECK4-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
2552 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
2553 // CHECK4-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2554 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
2555 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
2556 // CHECK4-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
2557 // CHECK4-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
2558 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
2559 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
2560 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2561 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
2562 // CHECK4-NEXT:    ret void
2563 //
2564 //
2565 // CHECK7-LABEL: define {{[^@]+}}@main
2566 // CHECK7-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
2567 // CHECK7-NEXT:  entry:
2568 // CHECK7-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
2569 // CHECK7-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
2570 // CHECK7-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
2571 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
2572 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
2573 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
2574 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
2575 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
2576 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_LB_1:%.*]] = alloca i32, align 4
2577 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_UB_2:%.*]] = alloca i32, align 4
2578 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_ST_3:%.*]] = alloca i32, align 4
2579 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IL_4:%.*]] = alloca i32, align 4
2580 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IV_5:%.*]] = alloca i32, align 4
2581 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2582 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2583 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2584 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_21:%.*]] = alloca i32, align 4
2585 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
2586 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2587 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2588 // CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2589 // CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2590 // CHECK7-NEXT:    [[I24:%.*]] = alloca i32, align 4
2591 // CHECK7-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
2592 // CHECK7-NEXT:    [[R:%.*]] = alloca i32, align 4
2593 // CHECK7-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
2594 // CHECK7-NEXT:    store i32 0, i32* [[RETVAL]], align 4
2595 // CHECK7-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
2596 // CHECK7-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
2597 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i8***, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i8*** [[ARGV_ADDR]], i32* [[ARGC_ADDR]])
2598 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2599 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2600 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
2601 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
2602 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
2603 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2604 // CHECK7-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
2605 // CHECK7-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
2606 // CHECK7-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
2607 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2608 // CHECK7-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2609 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2610 // CHECK7:       omp.inner.for.cond:
2611 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2612 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2613 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2614 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2615 // CHECK7:       omp.inner.for.body:
2616 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2617 // CHECK7-NEXT:    switch i32 [[TMP7]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
2618 // CHECK7-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
2619 // CHECK7-NEXT:    ]
2620 // CHECK7:       .omp.sections.case:
2621 // CHECK7-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
2622 // CHECK7-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
2623 // CHECK7-NEXT:    br i1 [[TMP9]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
2624 // CHECK7:       .cancel.exit:
2625 // CHECK7-NEXT:    br label [[CANCEL_EXIT:%.*]]
2626 // CHECK7:       .cancel.continue:
2627 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
2628 // CHECK7:       .omp.sections.exit:
2629 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2630 // CHECK7:       omp.inner.for.inc:
2631 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2632 // CHECK7-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP10]], 1
2633 // CHECK7-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2634 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
2635 // CHECK7:       omp.inner.for.end:
2636 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
2637 // CHECK7-NEXT:    br label [[CANCEL_CONT:%.*]]
2638 // CHECK7:       cancel.cont:
2639 // CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
2640 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
2641 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
2642 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_3]], align 4
2643 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_4]], align 4
2644 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_4]], i32* [[DOTOMP_SECTIONS_LB_1]], i32* [[DOTOMP_SECTIONS_UB_2]], i32* [[DOTOMP_SECTIONS_ST_3]], i32 1, i32 1)
2645 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
2646 // CHECK7-NEXT:    [[TMP12:%.*]] = icmp slt i32 [[TMP11]], 1
2647 // CHECK7-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 1
2648 // CHECK7-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_SECTIONS_UB_2]], align 4
2649 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
2650 // CHECK7-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
2651 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND6:%.*]]
2652 // CHECK7:       omp.inner.for.cond6:
2653 // CHECK7-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
2654 // CHECK7-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
2655 // CHECK7-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
2656 // CHECK7-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY8:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
2657 // CHECK7:       omp.inner.for.body8:
2658 // CHECK7-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
2659 // CHECK7-NEXT:    switch i32 [[TMP17]], label [[DOTOMP_SECTIONS_EXIT15:%.*]] [
2660 // CHECK7-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE9:%.*]]
2661 // CHECK7-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE12:%.*]]
2662 // CHECK7-NEXT:    ]
2663 // CHECK7:       .omp.sections.case9:
2664 // CHECK7-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
2665 // CHECK7-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
2666 // CHECK7-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT10:%.*]], label [[DOTCANCEL_CONTINUE11:%.*]]
2667 // CHECK7:       .cancel.exit10:
2668 // CHECK7-NEXT:    br label [[CANCEL_EXIT19:%.*]]
2669 // CHECK7:       cancel.exit:
2670 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
2671 // CHECK7-NEXT:    br label [[CANCEL_CONT]]
2672 // CHECK7:       .cancel.continue11:
2673 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
2674 // CHECK7:       .omp.sections.case12:
2675 // CHECK7-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
2676 // CHECK7-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
2677 // CHECK7-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT13:%.*]], label [[DOTCANCEL_CONTINUE14:%.*]]
2678 // CHECK7:       .cancel.exit13:
2679 // CHECK7-NEXT:    br label [[CANCEL_EXIT19]]
2680 // CHECK7:       .cancel.continue14:
2681 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
2682 // CHECK7:       .omp.sections.exit15:
2683 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
2684 // CHECK7:       omp.inner.for.inc16:
2685 // CHECK7-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
2686 // CHECK7-NEXT:    [[INC17:%.*]] = add nsw i32 [[TMP22]], 1
2687 // CHECK7-NEXT:    store i32 [[INC17]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
2688 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND6]]
2689 // CHECK7:       omp.inner.for.end18:
2690 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
2691 // CHECK7-NEXT:    br label [[CANCEL_CONT20:%.*]]
2692 // CHECK7:       cancel.cont20:
2693 // CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
2694 // CHECK7-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2695 // CHECK7-NEXT:    store i32 [[TMP23]], i32* [[DOTCAPTURE_EXPR_]], align 4
2696 // CHECK7-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2697 // CHECK7-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
2698 // CHECK7-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2699 // CHECK7-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[DIV]], 1
2700 // CHECK7-NEXT:    store i32 [[SUB22]], i32* [[DOTCAPTURE_EXPR_21]], align 4
2701 // CHECK7-NEXT:    store i32 0, i32* [[I]], align 4
2702 // CHECK7-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2703 // CHECK7-NEXT:    [[CMP23:%.*]] = icmp slt i32 0, [[TMP25]]
2704 // CHECK7-NEXT:    br i1 [[CMP23]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2705 // CHECK7:       omp.precond.then:
2706 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2707 // CHECK7-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
2708 // CHECK7-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_UB]], align 4
2709 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2710 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2711 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2712 // CHECK7-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2713 // CHECK7-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
2714 // CHECK7-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
2715 // CHECK7-NEXT:    br i1 [[CMP25]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2716 // CHECK7:       cond.true:
2717 // CHECK7-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
2718 // CHECK7-NEXT:    br label [[COND_END:%.*]]
2719 // CHECK7:       cond.false:
2720 // CHECK7-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2721 // CHECK7-NEXT:    br label [[COND_END]]
2722 // CHECK7:       cond.end:
2723 // CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE]] ], [ [[TMP30]], [[COND_FALSE]] ]
2724 // CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2725 // CHECK7-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2726 // CHECK7-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
2727 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND26:%.*]]
2728 // CHECK7:       omp.inner.for.cond26:
2729 // CHECK7-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2730 // CHECK7-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2731 // CHECK7-NEXT:    [[CMP27:%.*]] = icmp sle i32 [[TMP32]], [[TMP33]]
2732 // CHECK7-NEXT:    br i1 [[CMP27]], label [[OMP_INNER_FOR_BODY28:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
2733 // CHECK7:       omp.inner.for.body28:
2734 // CHECK7-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2735 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP34]], 1
2736 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2737 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[I24]], align 4
2738 // CHECK7-NEXT:    [[TMP35:%.*]] = load float, float* @flag, align 4
2739 // CHECK7-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP35]], 0.000000e+00
2740 // CHECK7-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2741 // CHECK7:       omp_if.then:
2742 // CHECK7-NEXT:    [[TMP36:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
2743 // CHECK7-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
2744 // CHECK7-NEXT:    br i1 [[TMP37]], label [[DOTCANCEL_EXIT29:%.*]], label [[DOTCANCEL_CONTINUE30:%.*]]
2745 // CHECK7:       .cancel.exit29:
2746 // CHECK7-NEXT:    br label [[CANCEL_EXIT34:%.*]]
2747 // CHECK7:       cancel.exit19:
2748 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
2749 // CHECK7-NEXT:    br label [[CANCEL_CONT20]]
2750 // CHECK7:       .cancel.continue30:
2751 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
2752 // CHECK7:       omp_if.else:
2753 // CHECK7-NEXT:    br label [[OMP_IF_END]]
2754 // CHECK7:       omp_if.end:
2755 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2756 // CHECK7:       omp.body.continue:
2757 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
2758 // CHECK7:       omp.inner.for.inc31:
2759 // CHECK7-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2760 // CHECK7-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP38]], 1
2761 // CHECK7-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV]], align 4
2762 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND26]]
2763 // CHECK7:       omp.inner.for.end33:
2764 // CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2765 // CHECK7:       omp.loop.exit:
2766 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
2767 // CHECK7-NEXT:    br label [[OMP_PRECOND_END]]
2768 // CHECK7:       cancel.exit34:
2769 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
2770 // CHECK7-NEXT:    br label [[CANCEL_CONT35:%.*]]
2771 // CHECK7:       omp.precond.end:
2772 // CHECK7-NEXT:    br label [[CANCEL_CONT35]]
2773 // CHECK7:       cancel.cont35:
2774 // CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]])
2775 // CHECK7-NEXT:    [[TMP39:%.*]] = call i8* @__kmpc_omp_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*)*))
2776 // CHECK7-NEXT:    [[TMP40:%.*]] = bitcast i8* [[TMP39]] to %struct.kmp_task_t_with_privates*
2777 // CHECK7-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP40]], i32 0, i32 0
2778 // CHECK7-NEXT:    [[TMP42:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP39]])
2779 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
2780 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*))
2781 // CHECK7-NEXT:    store i32 0, i32* [[R]], align 4
2782 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])
2783 // CHECK7-NEXT:    [[TMP43:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2784 // CHECK7-NEXT:    ret i32 [[TMP43]]
2785 //
2786 //
2787 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined.
2788 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8*** nonnull align 8 dereferenceable(8) [[ARGV:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]]) #[[ATTR1:[0-9]+]] {
2789 // CHECK7-NEXT:  entry:
2790 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2791 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2792 // CHECK7-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8***, align 8
2793 // CHECK7-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
2794 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2795 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2796 // CHECK7-NEXT:    store i8*** [[ARGV]], i8**** [[ARGV_ADDR]], align 8
2797 // CHECK7-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
2798 // CHECK7-NEXT:    [[TMP0:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
2799 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
2800 // CHECK7-NEXT:    [[TMP2:%.*]] = load float, float* @flag, align 4
2801 // CHECK7-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP2]], 0.000000e+00
2802 // CHECK7-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2803 // CHECK7:       omp_if.then:
2804 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2805 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2806 // CHECK7-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 1)
2807 // CHECK7-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[TMP5]], 0
2808 // CHECK7-NEXT:    br i1 [[TMP6]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
2809 // CHECK7:       .cancel.exit:
2810 // CHECK7-NEXT:    br label [[RETURN:%.*]]
2811 // CHECK7:       .cancel.continue:
2812 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
2813 // CHECK7:       omp_if.else:
2814 // CHECK7-NEXT:    br label [[OMP_IF_END]]
2815 // CHECK7:       omp_if.end:
2816 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP1]], align 4
2817 // CHECK7-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP7]] to i8
2818 // CHECK7-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP0]], align 8
2819 // CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 0
2820 // CHECK7-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
2821 // CHECK7-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 0
2822 // CHECK7-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX1]], align 1
2823 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2824 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
2825 // CHECK7-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP11]])
2826 // CHECK7-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
2827 // CHECK7-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
2828 // CHECK7:       .cancel.exit2:
2829 // CHECK7-NEXT:    br label [[RETURN]]
2830 // CHECK7:       .cancel.continue3:
2831 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP1]], align 4
2832 // CHECK7-NEXT:    [[TMP15:%.*]] = load i8**, i8*** [[TMP0]], align 8
2833 // CHECK7-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP15]], i64 0
2834 // CHECK7-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
2835 // CHECK7-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP16]], i64 0
2836 // CHECK7-NEXT:    [[TMP17:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
2837 // CHECK7-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP17]] to i32
2838 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP14]]
2839 // CHECK7-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD]] to i8
2840 // CHECK7-NEXT:    store i8 [[CONV7]], i8* [[ARRAYIDX5]], align 1
2841 // CHECK7-NEXT:    br label [[RETURN]]
2842 // CHECK7:       return:
2843 // CHECK7-NEXT:    ret void
2844 //
2845 //
2846 // CHECK7-LABEL: define {{[^@]+}}@.omp_task_entry.
2847 // CHECK7-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
2848 // CHECK7-NEXT:  entry:
2849 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
2850 // CHECK7-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
2851 // CHECK7-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
2852 // CHECK7-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
2853 // CHECK7-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
2854 // CHECK7-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
2855 // CHECK7-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
2856 // CHECK7-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
2857 // CHECK7-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
2858 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
2859 // CHECK7-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
2860 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
2861 // CHECK7-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
2862 // CHECK7-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
2863 // CHECK7-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
2864 // CHECK7-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
2865 // CHECK7-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2866 // CHECK7-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
2867 // CHECK7-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
2868 // CHECK7-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META4:![0-9]+]])
2869 // CHECK7-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META7:![0-9]+]])
2870 // CHECK7-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META9:![0-9]+]])
2871 // CHECK7-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META11:![0-9]+]])
2872 // CHECK7-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !13
2873 // CHECK7-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !13
2874 // CHECK7-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !13
2875 // CHECK7-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !13
2876 // CHECK7-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !13
2877 // CHECK7-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
2878 // CHECK7-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
2879 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !13
2880 // CHECK7-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 4) #[[ATTR2:[0-9]+]]
2881 // CHECK7-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
2882 // CHECK7-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
2883 // CHECK7:       .cancel.exit.i:
2884 // CHECK7-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
2885 // CHECK7-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
2886 // CHECK7:       .cancel.continue.i:
2887 // CHECK7-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
2888 // CHECK7-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
2889 // CHECK7:       .omp_outlined..1.exit:
2890 // CHECK7-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
2891 // CHECK7-NEXT:    ret i32 0
2892 //
2893 //
2894 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..2
2895 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2896 // CHECK7-NEXT:  entry:
2897 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2898 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2899 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
2900 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
2901 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
2902 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
2903 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
2904 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2905 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2906 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2907 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2908 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
2909 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
2910 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2911 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2912 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
2913 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2914 // CHECK7-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
2915 // CHECK7-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
2916 // CHECK7-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
2917 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2918 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2919 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2920 // CHECK7:       omp.inner.for.cond:
2921 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2922 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2923 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2924 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2925 // CHECK7:       omp.inner.for.body:
2926 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2927 // CHECK7-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
2928 // CHECK7-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
2929 // CHECK7-NEXT:    ]
2930 // CHECK7:       .omp.sections.case:
2931 // CHECK7-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
2932 // CHECK7-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
2933 // CHECK7-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
2934 // CHECK7:       .cancel.exit:
2935 // CHECK7-NEXT:    br label [[CANCEL_EXIT:%.*]]
2936 // CHECK7:       .cancel.continue:
2937 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
2938 // CHECK7:       .omp.sections.exit:
2939 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2940 // CHECK7:       omp.inner.for.inc:
2941 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2942 // CHECK7-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
2943 // CHECK7-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2944 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
2945 // CHECK7:       omp.inner.for.end:
2946 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
2947 // CHECK7-NEXT:    br label [[CANCEL_CONT:%.*]]
2948 // CHECK7:       cancel.cont:
2949 // CHECK7-NEXT:    ret void
2950 // CHECK7:       cancel.exit:
2951 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
2952 // CHECK7-NEXT:    br label [[CANCEL_CONT]]
2953 //
2954 //
2955 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..3
2956 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2957 // CHECK7-NEXT:  entry:
2958 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2959 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2960 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
2961 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
2962 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
2963 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
2964 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
2965 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2966 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2967 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2968 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2969 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
2970 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
2971 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2972 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2973 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
2974 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2975 // CHECK7-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
2976 // CHECK7-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
2977 // CHECK7-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
2978 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2979 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2980 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2981 // CHECK7:       omp.inner.for.cond:
2982 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2983 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2984 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2985 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2986 // CHECK7:       omp.inner.for.body:
2987 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2988 // CHECK7-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
2989 // CHECK7-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
2990 // CHECK7-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE1:%.*]]
2991 // CHECK7-NEXT:    ]
2992 // CHECK7:       .omp.sections.case:
2993 // CHECK7-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
2994 // CHECK7-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
2995 // CHECK7-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
2996 // CHECK7:       .cancel.exit:
2997 // CHECK7-NEXT:    br label [[CANCEL_EXIT:%.*]]
2998 // CHECK7:       .cancel.continue:
2999 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
3000 // CHECK7:       .omp.sections.case1:
3001 // CHECK7-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
3002 // CHECK7-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3003 // CHECK7-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
3004 // CHECK7:       .cancel.exit2:
3005 // CHECK7-NEXT:    br label [[CANCEL_EXIT]]
3006 // CHECK7:       .cancel.continue3:
3007 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
3008 // CHECK7:       .omp.sections.exit:
3009 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3010 // CHECK7:       omp.inner.for.inc:
3011 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3012 // CHECK7-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
3013 // CHECK7-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3014 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
3015 // CHECK7:       omp.inner.for.end:
3016 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
3017 // CHECK7-NEXT:    br label [[CANCEL_CONT:%.*]]
3018 // CHECK7:       cancel.cont:
3019 // CHECK7-NEXT:    ret void
3020 // CHECK7:       cancel.exit:
3021 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
3022 // CHECK7-NEXT:    br label [[CANCEL_CONT]]
3023 //
3024 //
3025 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..4
3026 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR1]] {
3027 // CHECK7-NEXT:  entry:
3028 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3029 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3030 // CHECK7-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
3031 // CHECK7-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
3032 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3033 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3034 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3035 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3036 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
3037 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3038 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3039 // CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3040 // CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3041 // CHECK7-NEXT:    [[R3:%.*]] = alloca i32, align 4
3042 // CHECK7-NEXT:    [[I4:%.*]] = alloca i32, align 4
3043 // CHECK7-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
3044 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3045 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3046 // CHECK7-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
3047 // CHECK7-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
3048 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
3049 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
3050 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
3051 // CHECK7-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
3052 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3053 // CHECK7-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
3054 // CHECK7-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3055 // CHECK7-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3056 // CHECK7-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3057 // CHECK7-NEXT:    store i32 0, i32* [[I]], align 4
3058 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3059 // CHECK7-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
3060 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3061 // CHECK7:       omp.precond.then:
3062 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3063 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3064 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
3065 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3066 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3067 // CHECK7-NEXT:    store i32 0, i32* [[R3]], align 4
3068 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3069 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
3070 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3071 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3072 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3073 // CHECK7-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
3074 // CHECK7-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3075 // CHECK7:       cond.true:
3076 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3077 // CHECK7-NEXT:    br label [[COND_END:%.*]]
3078 // CHECK7:       cond.false:
3079 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3080 // CHECK7-NEXT:    br label [[COND_END]]
3081 // CHECK7:       cond.end:
3082 // CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
3083 // CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3084 // CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3085 // CHECK7-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
3086 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3087 // CHECK7:       omp.inner.for.cond:
3088 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3089 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3090 // CHECK7-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
3091 // CHECK7-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3092 // CHECK7:       omp.inner.for.body:
3093 // CHECK7-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3094 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
3095 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3096 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
3097 // CHECK7-NEXT:    [[TMP16:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3098 // CHECK7-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4
3099 // CHECK7-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP17]], i32 2)
3100 // CHECK7-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
3101 // CHECK7-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3102 // CHECK7:       .cancel.exit:
3103 // CHECK7-NEXT:    br label [[CANCEL_EXIT:%.*]]
3104 // CHECK7:       .cancel.continue:
3105 // CHECK7-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4
3106 // CHECK7-NEXT:    [[TMP21:%.*]] = load i32, i32* [[R3]], align 4
3107 // CHECK7-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP20]]
3108 // CHECK7-NEXT:    store i32 [[ADD7]], i32* [[R3]], align 4
3109 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3110 // CHECK7:       omp.body.continue:
3111 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3112 // CHECK7:       omp.inner.for.inc:
3113 // CHECK7-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3114 // CHECK7-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
3115 // CHECK7-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
3116 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
3117 // CHECK7:       omp.inner.for.end:
3118 // CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3119 // CHECK7:       omp.loop.exit:
3120 // CHECK7-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3121 // CHECK7-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
3122 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP24]])
3123 // CHECK7-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
3124 // CHECK7-NEXT:    [[TMP26:%.*]] = bitcast i32* [[R3]] to i8*
3125 // CHECK7-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
3126 // CHECK7-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3127 // CHECK7-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
3128 // CHECK7-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
3129 // CHECK7-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB7:[0-9]+]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
3130 // CHECK7-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3131 // CHECK7-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3132 // CHECK7-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3133 // CHECK7-NEXT:    ]
3134 // CHECK7:       .omp.reduction.case1:
3135 // CHECK7-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP1]], align 4
3136 // CHECK7-NEXT:    [[TMP32:%.*]] = load i32, i32* [[R3]], align 4
3137 // CHECK7-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
3138 // CHECK7-NEXT:    store i32 [[ADD9]], i32* [[TMP1]], align 4
3139 // CHECK7-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB7]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
3140 // CHECK7-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3141 // CHECK7:       cancel.exit:
3142 // CHECK7-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3143 // CHECK7-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
3144 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP34]])
3145 // CHECK7-NEXT:    br label [[CANCEL_CONT:%.*]]
3146 // CHECK7:       .omp.reduction.case2:
3147 // CHECK7-NEXT:    [[TMP35:%.*]] = load i32, i32* [[R3]], align 4
3148 // CHECK7-NEXT:    [[TMP36:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP35]] monotonic, align 4
3149 // CHECK7-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3150 // CHECK7:       .omp.reduction.default:
3151 // CHECK7-NEXT:    br label [[OMP_PRECOND_END]]
3152 // CHECK7:       omp.precond.end:
3153 // CHECK7-NEXT:    br label [[CANCEL_CONT]]
3154 // CHECK7:       cancel.cont:
3155 // CHECK7-NEXT:    ret void
3156 //
3157 //
3158 // CHECK7-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
3159 // CHECK7-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
3160 // CHECK7-NEXT:  entry:
3161 // CHECK7-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
3162 // CHECK7-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
3163 // CHECK7-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
3164 // CHECK7-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
3165 // CHECK7-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
3166 // CHECK7-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
3167 // CHECK7-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
3168 // CHECK7-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
3169 // CHECK7-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
3170 // CHECK7-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
3171 // CHECK7-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
3172 // CHECK7-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
3173 // CHECK7-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
3174 // CHECK7-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
3175 // CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
3176 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
3177 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3178 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
3179 // CHECK7-NEXT:    ret void
3180 //
3181 //
3182 // CHECK8-LABEL: define {{[^@]+}}@main
3183 // CHECK8-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
3184 // CHECK8-NEXT:  entry:
3185 // CHECK8-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
3186 // CHECK8-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
3187 // CHECK8-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
3188 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
3189 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
3190 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
3191 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
3192 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
3193 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_LB_1:%.*]] = alloca i32, align 4
3194 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_UB_2:%.*]] = alloca i32, align 4
3195 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_ST_3:%.*]] = alloca i32, align 4
3196 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IL_4:%.*]] = alloca i32, align 4
3197 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IV_5:%.*]] = alloca i32, align 4
3198 // CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3199 // CHECK8-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3200 // CHECK8-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3201 // CHECK8-NEXT:    [[DOTCAPTURE_EXPR_21:%.*]] = alloca i32, align 4
3202 // CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
3203 // CHECK8-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3204 // CHECK8-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3205 // CHECK8-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3206 // CHECK8-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3207 // CHECK8-NEXT:    [[I24:%.*]] = alloca i32, align 4
3208 // CHECK8-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
3209 // CHECK8-NEXT:    [[R:%.*]] = alloca i32, align 4
3210 // CHECK8-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
3211 // CHECK8-NEXT:    store i32 0, i32* [[RETVAL]], align 4
3212 // CHECK8-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3213 // CHECK8-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
3214 // CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i8***, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i8*** [[ARGV_ADDR]], i32* [[ARGC_ADDR]])
3215 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
3216 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3217 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
3218 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
3219 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
3220 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3221 // CHECK8-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
3222 // CHECK8-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
3223 // CHECK8-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
3224 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
3225 // CHECK8-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3226 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3227 // CHECK8:       omp.inner.for.cond:
3228 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3229 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3230 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3231 // CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3232 // CHECK8:       omp.inner.for.body:
3233 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3234 // CHECK8-NEXT:    switch i32 [[TMP7]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
3235 // CHECK8-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
3236 // CHECK8-NEXT:    ]
3237 // CHECK8:       .omp.sections.case:
3238 // CHECK8-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
3239 // CHECK8-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
3240 // CHECK8-NEXT:    br i1 [[TMP9]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3241 // CHECK8:       .cancel.exit:
3242 // CHECK8-NEXT:    br label [[CANCEL_EXIT:%.*]]
3243 // CHECK8:       .cancel.continue:
3244 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
3245 // CHECK8:       .omp.sections.exit:
3246 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3247 // CHECK8:       omp.inner.for.inc:
3248 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3249 // CHECK8-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP10]], 1
3250 // CHECK8-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3251 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
3252 // CHECK8:       omp.inner.for.end:
3253 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
3254 // CHECK8-NEXT:    br label [[CANCEL_CONT:%.*]]
3255 // CHECK8:       cancel.cont:
3256 // CHECK8-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP0]])
3257 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
3258 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
3259 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_3]], align 4
3260 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_4]], align 4
3261 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_SECTIONS_IL_4]], i32* [[DOTOMP_SECTIONS_LB_1]], i32* [[DOTOMP_SECTIONS_UB_2]], i32* [[DOTOMP_SECTIONS_ST_3]], i32 1, i32 1)
3262 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
3263 // CHECK8-NEXT:    [[TMP12:%.*]] = icmp slt i32 [[TMP11]], 1
3264 // CHECK8-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 1
3265 // CHECK8-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_SECTIONS_UB_2]], align 4
3266 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
3267 // CHECK8-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
3268 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND6:%.*]]
3269 // CHECK8:       omp.inner.for.cond6:
3270 // CHECK8-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
3271 // CHECK8-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
3272 // CHECK8-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
3273 // CHECK8-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY8:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
3274 // CHECK8:       omp.inner.for.body8:
3275 // CHECK8-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
3276 // CHECK8-NEXT:    switch i32 [[TMP17]], label [[DOTOMP_SECTIONS_EXIT15:%.*]] [
3277 // CHECK8-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE9:%.*]]
3278 // CHECK8-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE12:%.*]]
3279 // CHECK8-NEXT:    ]
3280 // CHECK8:       .omp.sections.case9:
3281 // CHECK8-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
3282 // CHECK8-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
3283 // CHECK8-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT10:%.*]], label [[DOTCANCEL_CONTINUE11:%.*]]
3284 // CHECK8:       .cancel.exit10:
3285 // CHECK8-NEXT:    br label [[CANCEL_EXIT19:%.*]]
3286 // CHECK8:       cancel.exit:
3287 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
3288 // CHECK8-NEXT:    br label [[CANCEL_CONT]]
3289 // CHECK8:       .cancel.continue11:
3290 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
3291 // CHECK8:       .omp.sections.case12:
3292 // CHECK8-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
3293 // CHECK8-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
3294 // CHECK8-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT13:%.*]], label [[DOTCANCEL_CONTINUE14:%.*]]
3295 // CHECK8:       .cancel.exit13:
3296 // CHECK8-NEXT:    br label [[CANCEL_EXIT19]]
3297 // CHECK8:       .cancel.continue14:
3298 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
3299 // CHECK8:       .omp.sections.exit15:
3300 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
3301 // CHECK8:       omp.inner.for.inc16:
3302 // CHECK8-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
3303 // CHECK8-NEXT:    [[INC17:%.*]] = add nsw i32 [[TMP22]], 1
3304 // CHECK8-NEXT:    store i32 [[INC17]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
3305 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND6]]
3306 // CHECK8:       omp.inner.for.end18:
3307 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
3308 // CHECK8-NEXT:    br label [[CANCEL_CONT20:%.*]]
3309 // CHECK8:       cancel.cont20:
3310 // CHECK8-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
3311 // CHECK8-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3312 // CHECK8-NEXT:    store i32 [[TMP23]], i32* [[DOTCAPTURE_EXPR_]], align 4
3313 // CHECK8-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3314 // CHECK8-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
3315 // CHECK8-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3316 // CHECK8-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[DIV]], 1
3317 // CHECK8-NEXT:    store i32 [[SUB22]], i32* [[DOTCAPTURE_EXPR_21]], align 4
3318 // CHECK8-NEXT:    store i32 0, i32* [[I]], align 4
3319 // CHECK8-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3320 // CHECK8-NEXT:    [[CMP23:%.*]] = icmp slt i32 0, [[TMP25]]
3321 // CHECK8-NEXT:    br i1 [[CMP23]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3322 // CHECK8:       omp.precond.then:
3323 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3324 // CHECK8-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
3325 // CHECK8-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_UB]], align 4
3326 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3327 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3328 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3329 // CHECK8-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3330 // CHECK8-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
3331 // CHECK8-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
3332 // CHECK8-NEXT:    br i1 [[CMP25]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3333 // CHECK8:       cond.true:
3334 // CHECK8-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
3335 // CHECK8-NEXT:    br label [[COND_END:%.*]]
3336 // CHECK8:       cond.false:
3337 // CHECK8-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3338 // CHECK8-NEXT:    br label [[COND_END]]
3339 // CHECK8:       cond.end:
3340 // CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE]] ], [ [[TMP30]], [[COND_FALSE]] ]
3341 // CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3342 // CHECK8-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3343 // CHECK8-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
3344 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND26:%.*]]
3345 // CHECK8:       omp.inner.for.cond26:
3346 // CHECK8-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3347 // CHECK8-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3348 // CHECK8-NEXT:    [[CMP27:%.*]] = icmp sle i32 [[TMP32]], [[TMP33]]
3349 // CHECK8-NEXT:    br i1 [[CMP27]], label [[OMP_INNER_FOR_BODY28:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
3350 // CHECK8:       omp.inner.for.body28:
3351 // CHECK8-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3352 // CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP34]], 1
3353 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3354 // CHECK8-NEXT:    store i32 [[ADD]], i32* [[I24]], align 4
3355 // CHECK8-NEXT:    [[TMP35:%.*]] = load float, float* @flag, align 4
3356 // CHECK8-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP35]], 0.000000e+00
3357 // CHECK8-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3358 // CHECK8:       omp_if.then:
3359 // CHECK8-NEXT:    [[TMP36:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
3360 // CHECK8-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
3361 // CHECK8-NEXT:    br i1 [[TMP37]], label [[DOTCANCEL_EXIT29:%.*]], label [[DOTCANCEL_CONTINUE30:%.*]]
3362 // CHECK8:       .cancel.exit29:
3363 // CHECK8-NEXT:    br label [[CANCEL_EXIT34:%.*]]
3364 // CHECK8:       cancel.exit19:
3365 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]])
3366 // CHECK8-NEXT:    br label [[CANCEL_CONT20]]
3367 // CHECK8:       .cancel.continue30:
3368 // CHECK8-NEXT:    br label [[OMP_IF_END:%.*]]
3369 // CHECK8:       omp_if.else:
3370 // CHECK8-NEXT:    br label [[OMP_IF_END]]
3371 // CHECK8:       omp_if.end:
3372 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3373 // CHECK8:       omp.body.continue:
3374 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
3375 // CHECK8:       omp.inner.for.inc31:
3376 // CHECK8-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3377 // CHECK8-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP38]], 1
3378 // CHECK8-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV]], align 4
3379 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND26]]
3380 // CHECK8:       omp.inner.for.end33:
3381 // CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3382 // CHECK8:       omp.loop.exit:
3383 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
3384 // CHECK8-NEXT:    br label [[OMP_PRECOND_END]]
3385 // CHECK8:       cancel.exit34:
3386 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
3387 // CHECK8-NEXT:    br label [[CANCEL_CONT35:%.*]]
3388 // CHECK8:       omp.precond.end:
3389 // CHECK8-NEXT:    br label [[CANCEL_CONT35]]
3390 // CHECK8:       cancel.cont35:
3391 // CHECK8-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]])
3392 // CHECK8-NEXT:    [[TMP39:%.*]] = call i8* @__kmpc_omp_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*)*))
3393 // CHECK8-NEXT:    [[TMP40:%.*]] = bitcast i8* [[TMP39]] to %struct.kmp_task_t_with_privates*
3394 // CHECK8-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP40]], i32 0, i32 0
3395 // CHECK8-NEXT:    [[TMP42:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP39]])
3396 // CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
3397 // CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*))
3398 // CHECK8-NEXT:    store i32 0, i32* [[R]], align 4
3399 // CHECK8-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])
3400 // CHECK8-NEXT:    [[TMP43:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3401 // CHECK8-NEXT:    ret i32 [[TMP43]]
3402 //
3403 //
3404 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined.
3405 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i8*** nonnull align 8 dereferenceable(8) [[ARGV:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]]) #[[ATTR1:[0-9]+]] {
3406 // CHECK8-NEXT:  entry:
3407 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3408 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3409 // CHECK8-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8***, align 8
3410 // CHECK8-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
3411 // CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3412 // CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3413 // CHECK8-NEXT:    store i8*** [[ARGV]], i8**** [[ARGV_ADDR]], align 8
3414 // CHECK8-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
3415 // CHECK8-NEXT:    [[TMP0:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
3416 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
3417 // CHECK8-NEXT:    [[TMP2:%.*]] = load float, float* @flag, align 4
3418 // CHECK8-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP2]], 0.000000e+00
3419 // CHECK8-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3420 // CHECK8:       omp_if.then:
3421 // CHECK8-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3422 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
3423 // CHECK8-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 1)
3424 // CHECK8-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[TMP5]], 0
3425 // CHECK8-NEXT:    br i1 [[TMP6]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3426 // CHECK8:       .cancel.exit:
3427 // CHECK8-NEXT:    br label [[RETURN:%.*]]
3428 // CHECK8:       .cancel.continue:
3429 // CHECK8-NEXT:    br label [[OMP_IF_END:%.*]]
3430 // CHECK8:       omp_if.else:
3431 // CHECK8-NEXT:    br label [[OMP_IF_END]]
3432 // CHECK8:       omp_if.end:
3433 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP1]], align 4
3434 // CHECK8-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP7]] to i8
3435 // CHECK8-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP0]], align 8
3436 // CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 0
3437 // CHECK8-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
3438 // CHECK8-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 0
3439 // CHECK8-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX1]], align 1
3440 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3441 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
3442 // CHECK8-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP11]])
3443 // CHECK8-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
3444 // CHECK8-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
3445 // CHECK8:       .cancel.exit2:
3446 // CHECK8-NEXT:    br label [[RETURN]]
3447 // CHECK8:       .cancel.continue3:
3448 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP1]], align 4
3449 // CHECK8-NEXT:    [[TMP15:%.*]] = load i8**, i8*** [[TMP0]], align 8
3450 // CHECK8-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP15]], i64 0
3451 // CHECK8-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
3452 // CHECK8-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP16]], i64 0
3453 // CHECK8-NEXT:    [[TMP17:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
3454 // CHECK8-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP17]] to i32
3455 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP14]]
3456 // CHECK8-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD]] to i8
3457 // CHECK8-NEXT:    store i8 [[CONV7]], i8* [[ARRAYIDX5]], align 1
3458 // CHECK8-NEXT:    br label [[RETURN]]
3459 // CHECK8:       return:
3460 // CHECK8-NEXT:    ret void
3461 //
3462 //
3463 // CHECK8-LABEL: define {{[^@]+}}@.omp_task_entry.
3464 // CHECK8-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
3465 // CHECK8-NEXT:  entry:
3466 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
3467 // CHECK8-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
3468 // CHECK8-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
3469 // CHECK8-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
3470 // CHECK8-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
3471 // CHECK8-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
3472 // CHECK8-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
3473 // CHECK8-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
3474 // CHECK8-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
3475 // CHECK8-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
3476 // CHECK8-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
3477 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
3478 // CHECK8-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
3479 // CHECK8-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
3480 // CHECK8-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
3481 // CHECK8-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
3482 // CHECK8-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
3483 // CHECK8-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
3484 // CHECK8-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
3485 // CHECK8-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META4:![0-9]+]])
3486 // CHECK8-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META7:![0-9]+]])
3487 // CHECK8-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META9:![0-9]+]])
3488 // CHECK8-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META11:![0-9]+]])
3489 // CHECK8-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !13
3490 // CHECK8-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !13
3491 // CHECK8-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !13
3492 // CHECK8-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !13
3493 // CHECK8-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !13
3494 // CHECK8-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
3495 // CHECK8-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
3496 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !13
3497 // CHECK8-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 4) #[[ATTR2:[0-9]+]]
3498 // CHECK8-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
3499 // CHECK8-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
3500 // CHECK8:       .cancel.exit.i:
3501 // CHECK8-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
3502 // CHECK8-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
3503 // CHECK8:       .cancel.continue.i:
3504 // CHECK8-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
3505 // CHECK8-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
3506 // CHECK8:       .omp_outlined..1.exit:
3507 // CHECK8-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
3508 // CHECK8-NEXT:    ret i32 0
3509 //
3510 //
3511 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..2
3512 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3513 // CHECK8-NEXT:  entry:
3514 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3515 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3516 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
3517 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
3518 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
3519 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
3520 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
3521 // CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3522 // CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3523 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
3524 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3525 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
3526 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
3527 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3528 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3529 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
3530 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3531 // CHECK8-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
3532 // CHECK8-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
3533 // CHECK8-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
3534 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
3535 // CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3536 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3537 // CHECK8:       omp.inner.for.cond:
3538 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3539 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3540 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3541 // CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3542 // CHECK8:       omp.inner.for.body:
3543 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3544 // CHECK8-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
3545 // CHECK8-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
3546 // CHECK8-NEXT:    ]
3547 // CHECK8:       .omp.sections.case:
3548 // CHECK8-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
3549 // CHECK8-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
3550 // CHECK8-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3551 // CHECK8:       .cancel.exit:
3552 // CHECK8-NEXT:    br label [[CANCEL_EXIT:%.*]]
3553 // CHECK8:       .cancel.continue:
3554 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
3555 // CHECK8:       .omp.sections.exit:
3556 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3557 // CHECK8:       omp.inner.for.inc:
3558 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3559 // CHECK8-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
3560 // CHECK8-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3561 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
3562 // CHECK8:       omp.inner.for.end:
3563 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
3564 // CHECK8-NEXT:    br label [[CANCEL_CONT:%.*]]
3565 // CHECK8:       cancel.cont:
3566 // CHECK8-NEXT:    ret void
3567 // CHECK8:       cancel.exit:
3568 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
3569 // CHECK8-NEXT:    br label [[CANCEL_CONT]]
3570 //
3571 //
3572 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..3
3573 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3574 // CHECK8-NEXT:  entry:
3575 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3576 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3577 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
3578 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
3579 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
3580 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
3581 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
3582 // CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3583 // CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3584 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
3585 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3586 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
3587 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
3588 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3589 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3590 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
3591 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3592 // CHECK8-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
3593 // CHECK8-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
3594 // CHECK8-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
3595 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
3596 // CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3597 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3598 // CHECK8:       omp.inner.for.cond:
3599 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3600 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3601 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3602 // CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3603 // CHECK8:       omp.inner.for.body:
3604 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3605 // CHECK8-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
3606 // CHECK8-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
3607 // CHECK8-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE1:%.*]]
3608 // CHECK8-NEXT:    ]
3609 // CHECK8:       .omp.sections.case:
3610 // CHECK8-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
3611 // CHECK8-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
3612 // CHECK8-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3613 // CHECK8:       .cancel.exit:
3614 // CHECK8-NEXT:    br label [[CANCEL_EXIT:%.*]]
3615 // CHECK8:       .cancel.continue:
3616 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
3617 // CHECK8:       .omp.sections.case1:
3618 // CHECK8-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
3619 // CHECK8-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3620 // CHECK8-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
3621 // CHECK8:       .cancel.exit2:
3622 // CHECK8-NEXT:    br label [[CANCEL_EXIT]]
3623 // CHECK8:       .cancel.continue3:
3624 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
3625 // CHECK8:       .omp.sections.exit:
3626 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3627 // CHECK8:       omp.inner.for.inc:
3628 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3629 // CHECK8-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
3630 // CHECK8-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3631 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
3632 // CHECK8:       omp.inner.for.end:
3633 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
3634 // CHECK8-NEXT:    br label [[CANCEL_CONT:%.*]]
3635 // CHECK8:       cancel.cont:
3636 // CHECK8-NEXT:    ret void
3637 // CHECK8:       cancel.exit:
3638 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
3639 // CHECK8-NEXT:    br label [[CANCEL_CONT]]
3640 //
3641 //
3642 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..4
3643 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR1]] {
3644 // CHECK8-NEXT:  entry:
3645 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3646 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3647 // CHECK8-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
3648 // CHECK8-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
3649 // CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3650 // CHECK8-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3651 // CHECK8-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3652 // CHECK8-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3653 // CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
3654 // CHECK8-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3655 // CHECK8-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3656 // CHECK8-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3657 // CHECK8-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3658 // CHECK8-NEXT:    [[R3:%.*]] = alloca i32, align 4
3659 // CHECK8-NEXT:    [[I4:%.*]] = alloca i32, align 4
3660 // CHECK8-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
3661 // CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3662 // CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3663 // CHECK8-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
3664 // CHECK8-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
3665 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
3666 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
3667 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
3668 // CHECK8-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
3669 // CHECK8-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3670 // CHECK8-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
3671 // CHECK8-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3672 // CHECK8-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3673 // CHECK8-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3674 // CHECK8-NEXT:    store i32 0, i32* [[I]], align 4
3675 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3676 // CHECK8-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
3677 // CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3678 // CHECK8:       omp.precond.then:
3679 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3680 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3681 // CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
3682 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3683 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3684 // CHECK8-NEXT:    store i32 0, i32* [[R3]], align 4
3685 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3686 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
3687 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB5]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3688 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3689 // CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3690 // CHECK8-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
3691 // CHECK8-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3692 // CHECK8:       cond.true:
3693 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3694 // CHECK8-NEXT:    br label [[COND_END:%.*]]
3695 // CHECK8:       cond.false:
3696 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3697 // CHECK8-NEXT:    br label [[COND_END]]
3698 // CHECK8:       cond.end:
3699 // CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
3700 // CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3701 // CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3702 // CHECK8-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
3703 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3704 // CHECK8:       omp.inner.for.cond:
3705 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3706 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3707 // CHECK8-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
3708 // CHECK8-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3709 // CHECK8:       omp.inner.for.body:
3710 // CHECK8-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3711 // CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
3712 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3713 // CHECK8-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
3714 // CHECK8-NEXT:    [[TMP16:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3715 // CHECK8-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4
3716 // CHECK8-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP17]], i32 2)
3717 // CHECK8-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
3718 // CHECK8-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3719 // CHECK8:       .cancel.exit:
3720 // CHECK8-NEXT:    br label [[CANCEL_EXIT:%.*]]
3721 // CHECK8:       .cancel.continue:
3722 // CHECK8-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4
3723 // CHECK8-NEXT:    [[TMP21:%.*]] = load i32, i32* [[R3]], align 4
3724 // CHECK8-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP20]]
3725 // CHECK8-NEXT:    store i32 [[ADD7]], i32* [[R3]], align 4
3726 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3727 // CHECK8:       omp.body.continue:
3728 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3729 // CHECK8:       omp.inner.for.inc:
3730 // CHECK8-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3731 // CHECK8-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
3732 // CHECK8-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
3733 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
3734 // CHECK8:       omp.inner.for.end:
3735 // CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3736 // CHECK8:       omp.loop.exit:
3737 // CHECK8-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3738 // CHECK8-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
3739 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP24]])
3740 // CHECK8-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
3741 // CHECK8-NEXT:    [[TMP26:%.*]] = bitcast i32* [[R3]] to i8*
3742 // CHECK8-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
3743 // CHECK8-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3744 // CHECK8-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
3745 // CHECK8-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
3746 // CHECK8-NEXT:    [[TMP30:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB7:[0-9]+]], i32 [[TMP28]], i32 1, i64 8, i8* [[TMP29]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
3747 // CHECK8-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3748 // CHECK8-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3749 // CHECK8-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3750 // CHECK8-NEXT:    ]
3751 // CHECK8:       .omp.reduction.case1:
3752 // CHECK8-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP1]], align 4
3753 // CHECK8-NEXT:    [[TMP32:%.*]] = load i32, i32* [[R3]], align 4
3754 // CHECK8-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
3755 // CHECK8-NEXT:    store i32 [[ADD9]], i32* [[TMP1]], align 4
3756 // CHECK8-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB7]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
3757 // CHECK8-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3758 // CHECK8:       cancel.exit:
3759 // CHECK8-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3760 // CHECK8-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
3761 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB5]], i32 [[TMP34]])
3762 // CHECK8-NEXT:    br label [[CANCEL_CONT:%.*]]
3763 // CHECK8:       .omp.reduction.case2:
3764 // CHECK8-NEXT:    [[TMP35:%.*]] = load i32, i32* [[R3]], align 4
3765 // CHECK8-NEXT:    [[TMP36:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP35]] monotonic, align 4
3766 // CHECK8-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3767 // CHECK8:       .omp.reduction.default:
3768 // CHECK8-NEXT:    br label [[OMP_PRECOND_END]]
3769 // CHECK8:       omp.precond.end:
3770 // CHECK8-NEXT:    br label [[CANCEL_CONT]]
3771 // CHECK8:       cancel.cont:
3772 // CHECK8-NEXT:    ret void
3773 //
3774 //
3775 // CHECK8-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
3776 // CHECK8-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
3777 // CHECK8-NEXT:  entry:
3778 // CHECK8-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
3779 // CHECK8-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
3780 // CHECK8-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
3781 // CHECK8-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
3782 // CHECK8-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
3783 // CHECK8-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
3784 // CHECK8-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
3785 // CHECK8-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
3786 // CHECK8-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
3787 // CHECK8-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
3788 // CHECK8-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
3789 // CHECK8-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
3790 // CHECK8-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
3791 // CHECK8-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
3792 // CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
3793 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
3794 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3795 // CHECK8-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
3796 // CHECK8-NEXT:    ret void
3797 //
3798 //
3799 // CHECK9-LABEL: define {{[^@]+}}@main
3800 // CHECK9-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
3801 // CHECK9-NEXT:  entry:
3802 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
3803 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
3804 // CHECK9-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
3805 // CHECK9-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
3806 // CHECK9-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
3807 // CHECK9-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
3808 // CHECK9-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
3809 // CHECK9-NEXT:    [[P_LASTITER26:%.*]] = alloca i32, align 4
3810 // CHECK9-NEXT:    [[P_LOWERBOUND27:%.*]] = alloca i32, align 4
3811 // CHECK9-NEXT:    [[P_UPPERBOUND28:%.*]] = alloca i32, align 4
3812 // CHECK9-NEXT:    [[P_STRIDE29:%.*]] = alloca i32, align 4
3813 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3814 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3815 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3816 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_33:%.*]] = alloca i32, align 4
3817 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
3818 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3819 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3820 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3821 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3822 // CHECK9-NEXT:    [[I35:%.*]] = alloca i32, align 4
3823 // CHECK9-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
3824 // CHECK9-NEXT:    [[R:%.*]] = alloca i32, align 4
3825 // CHECK9-NEXT:    store i32 0, i32* [[RETVAL]], align 4
3826 // CHECK9-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3827 // CHECK9-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
3828 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
3829 // CHECK9-NEXT:    br label [[OMP_PARALLEL:%.*]]
3830 // CHECK9:       omp_parallel:
3831 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8***)* @main..omp_par to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i8*** [[ARGV_ADDR]])
3832 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
3833 // CHECK9:       omp.par.outlined.exit:
3834 // CHECK9-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
3835 // CHECK9:       omp.par.exit.split:
3836 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
3837 // CHECK9:       omp_section_loop.preheader:
3838 // CHECK9-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
3839 // CHECK9-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
3840 // CHECK9-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
3841 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3842 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
3843 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
3844 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
3845 // CHECK9-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
3846 // CHECK9-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
3847 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
3848 // CHECK9:       omp_section_loop.header:
3849 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
3850 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
3851 // CHECK9:       omp_section_loop.cond:
3852 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
3853 // CHECK9-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
3854 // CHECK9:       omp_section_loop.body:
3855 // CHECK9-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
3856 // CHECK9-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
3857 // CHECK9-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
3858 // CHECK9-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
3859 // CHECK9-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
3860 // CHECK9-NEXT:    ]
3861 // CHECK9:       omp_section_loop.inc:
3862 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
3863 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
3864 // CHECK9:       omp_section_loop.exit:
3865 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
3866 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3867 // CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
3868 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
3869 // CHECK9:       omp_section_loop.after:
3870 // CHECK9-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
3871 // CHECK9:       omp_sections.end:
3872 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER12:%.*]]
3873 // CHECK9:       omp_section_loop.preheader12:
3874 // CHECK9-NEXT:    store i32 0, i32* [[P_LOWERBOUND27]], align 4
3875 // CHECK9-NEXT:    store i32 1, i32* [[P_UPPERBOUND28]], align 4
3876 // CHECK9-NEXT:    store i32 1, i32* [[P_STRIDE29]], align 4
3877 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM30:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3878 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]], i32 34, i32* [[P_LASTITER26]], i32* [[P_LOWERBOUND27]], i32* [[P_UPPERBOUND28]], i32* [[P_STRIDE29]], i32 1, i32 1)
3879 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND27]], align 4
3880 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND28]], align 4
3881 // CHECK9-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
3882 // CHECK9-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
3883 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13:%.*]]
3884 // CHECK9:       omp_section_loop.header13:
3885 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_IV19:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER12]] ], [ [[OMP_SECTION_LOOP_NEXT21:%.*]], [[OMP_SECTION_LOOP_INC16:%.*]] ]
3886 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_COND14:%.*]]
3887 // CHECK9:       omp_section_loop.cond14:
3888 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_CMP20:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV19]], [[TMP10]]
3889 // CHECK9-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP20]], label [[OMP_SECTION_LOOP_BODY15:%.*]], label [[OMP_SECTION_LOOP_EXIT17:%.*]]
3890 // CHECK9:       omp_section_loop.body15:
3891 // CHECK9-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV19]], [[TMP7]]
3892 // CHECK9-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
3893 // CHECK9-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
3894 // CHECK9-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC16]] [
3895 // CHECK9-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE22:%.*]]
3896 // CHECK9-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE24:%.*]]
3897 // CHECK9-NEXT:    ]
3898 // CHECK9:       omp_section_loop.inc16:
3899 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_NEXT21]] = add nuw i32 [[OMP_SECTION_LOOP_IV19]], 1
3900 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13]]
3901 // CHECK9:       omp_section_loop.exit17:
3902 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]])
3903 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3904 // CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
3905 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_AFTER18:%.*]]
3906 // CHECK9:       omp_section_loop.after18:
3907 // CHECK9-NEXT:    br label [[OMP_SECTIONS_END32:%.*]]
3908 // CHECK9:       omp_sections.end32:
3909 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3910 // CHECK9-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
3911 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3912 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
3913 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3914 // CHECK9-NEXT:    [[SUB34:%.*]] = sub nsw i32 [[DIV]], 1
3915 // CHECK9-NEXT:    store i32 [[SUB34]], i32* [[DOTCAPTURE_EXPR_33]], align 4
3916 // CHECK9-NEXT:    store i32 0, i32* [[I]], align 4
3917 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3918 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
3919 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3920 // CHECK9:       omp.precond.then:
3921 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3922 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
3923 // CHECK9-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
3924 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3925 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3926 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM36:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
3927 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM36]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3928 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3929 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
3930 // CHECK9-NEXT:    [[CMP37:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
3931 // CHECK9-NEXT:    br i1 [[CMP37]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3932 // CHECK9:       cond.true:
3933 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
3934 // CHECK9-NEXT:    br label [[COND_END:%.*]]
3935 // CHECK9:       cond.false:
3936 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3937 // CHECK9-NEXT:    br label [[COND_END]]
3938 // CHECK9:       cond.end:
3939 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
3940 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3941 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3942 // CHECK9-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
3943 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3944 // CHECK9:       omp.inner.for.cond:
3945 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3946 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3947 // CHECK9-NEXT:    [[CMP38:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
3948 // CHECK9-NEXT:    br i1 [[CMP38]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3949 // CHECK9:       omp.inner.for.body:
3950 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3951 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
3952 // CHECK9-NEXT:    [[ADD39:%.*]] = add nsw i32 0, [[MUL]]
3953 // CHECK9-NEXT:    store i32 [[ADD39]], i32* [[I35]], align 4
3954 // CHECK9-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
3955 // CHECK9-NEXT:    [[TOBOOL40:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
3956 // CHECK9-NEXT:    br i1 [[TOBOOL40]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3957 // CHECK9:       omp_if.then:
3958 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM41:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
3959 // CHECK9-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM41]], i32 2)
3960 // CHECK9-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
3961 // CHECK9-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3962 // CHECK9:       .cancel.exit:
3963 // CHECK9-NEXT:    br label [[CANCEL_EXIT:%.*]]
3964 // CHECK9:       omp_section_loop.body.case:
3965 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM9:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3966 // CHECK9-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM9]], i32 3)
3967 // CHECK9-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
3968 // CHECK9-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
3969 // CHECK9:       omp_section_loop.body.case.split:
3970 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
3971 // CHECK9:       omp_section_loop.body.case.cncl:
3972 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
3973 // CHECK9:       omp_section_loop.body.case22:
3974 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM23:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3975 // CHECK9-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]], i32 3)
3976 // CHECK9-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
3977 // CHECK9-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE22_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE22_CNCL:%.*]]
3978 // CHECK9:       omp_section_loop.body.case22.split:
3979 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
3980 // CHECK9:       omp_section_loop.body.case22.cncl:
3981 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
3982 // CHECK9:       omp_section_loop.body.case24:
3983 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM25:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3984 // CHECK9-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM25]], i32 3)
3985 // CHECK9-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
3986 // CHECK9-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE24_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE24_CNCL:%.*]]
3987 // CHECK9:       omp_section_loop.body.case24.split:
3988 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
3989 // CHECK9:       omp_section_loop.body.case24.cncl:
3990 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
3991 // CHECK9:       .cancel.continue:
3992 // CHECK9-NEXT:    br label [[OMP_IF_END:%.*]]
3993 // CHECK9:       omp_if.else:
3994 // CHECK9-NEXT:    br label [[OMP_IF_END]]
3995 // CHECK9:       omp_if.end:
3996 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3997 // CHECK9:       omp.body.continue:
3998 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3999 // CHECK9:       omp.inner.for.inc:
4000 // CHECK9-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4001 // CHECK9-NEXT:    [[ADD42:%.*]] = add nsw i32 [[TMP35]], 1
4002 // CHECK9-NEXT:    store i32 [[ADD42]], i32* [[DOTOMP_IV]], align 4
4003 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
4004 // CHECK9:       omp.inner.for.end:
4005 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4006 // CHECK9:       omp.loop.exit:
4007 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
4008 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
4009 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
4010 // CHECK9:       cancel.exit:
4011 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM43:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
4012 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM43]])
4013 // CHECK9-NEXT:    br label [[CANCEL_CONT:%.*]]
4014 // CHECK9:       omp.precond.end:
4015 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
4016 // CHECK9:       cancel.cont:
4017 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4018 // CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
4019 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
4020 // CHECK9-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM46]], 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*)*))
4021 // CHECK9-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
4022 // CHECK9-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
4023 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
4024 // CHECK9-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], i8* [[TMP36]])
4025 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
4026 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
4027 // CHECK9-NEXT:    store i32 0, i32* [[R]], align 4
4028 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])
4029 // CHECK9-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4030 // CHECK9-NEXT:    ret i32 [[TMP40]]
4031 //
4032 //
4033 // CHECK9-LABEL: define {{[^@]+}}@main..omp_par
4034 // CHECK9-SAME: (i32* noalias [[TID_ADDR:%.*]], i32* noalias [[ZERO_ADDR:%.*]], i32* [[ARGC_ADDR:%.*]], i8*** [[ARGV_ADDR:%.*]]) #[[ATTR1:[0-9]+]] {
4035 // CHECK9-NEXT:  omp.par.entry:
4036 // CHECK9-NEXT:    [[TID_ADDR_LOCAL:%.*]] = alloca i32, align 4
4037 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[TID_ADDR]], align 4
4038 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[TID_ADDR_LOCAL]], align 4
4039 // CHECK9-NEXT:    [[TID:%.*]] = load i32, i32* [[TID_ADDR_LOCAL]], align 4
4040 // CHECK9-NEXT:    br label [[OMP_PAR_REGION:%.*]]
4041 // CHECK9:       omp.par.outlined.exit.exitStub:
4042 // CHECK9-NEXT:    ret void
4043 // CHECK9:       omp.par.region:
4044 // CHECK9-NEXT:    [[TMP1:%.*]] = load float, float* @flag, align 4
4045 // CHECK9-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP1]], 0.000000e+00
4046 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[TMP13:%.*]], label [[TMP2:%.*]]
4047 // CHECK9:       2:
4048 // CHECK9-NEXT:    br label [[TMP3:%.*]]
4049 // CHECK9:       3:
4050 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4051 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP4]] to i8
4052 // CHECK9-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
4053 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 0
4054 // CHECK9-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
4055 // CHECK9-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
4056 // CHECK9-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX2]], align 1
4057 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4058 // CHECK9-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
4059 // CHECK9-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
4060 // CHECK9-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL4:%.*]]
4061 // CHECK9:       .cncl4:
4062 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
4063 // CHECK9:       .cont:
4064 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4065 // CHECK9-NEXT:    [[TMP10:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
4066 // CHECK9-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
4067 // CHECK9-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX5]], align 8
4068 // CHECK9-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
4069 // CHECK9-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX6]], align 1
4070 // CHECK9-NEXT:    [[CONV7:%.*]] = sext i8 [[TMP12]] to i32
4071 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV7]], [[TMP9]]
4072 // CHECK9-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD]] to i8
4073 // CHECK9-NEXT:    store i8 [[CONV8]], i8* [[ARRAYIDX6]], align 1
4074 // CHECK9-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
4075 // CHECK9:       omp.par.pre_finalize:
4076 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
4077 // CHECK9:       13:
4078 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4079 // CHECK9-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
4080 // CHECK9-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 0
4081 // CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
4082 // CHECK9:       .cncl:
4083 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
4084 // CHECK9:       .split:
4085 // CHECK9-NEXT:    br label [[TMP3]]
4086 //
4087 //
4088 // CHECK9-LABEL: define {{[^@]+}}@.omp_task_entry.
4089 // CHECK9-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
4090 // CHECK9-NEXT:  entry:
4091 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
4092 // CHECK9-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
4093 // CHECK9-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
4094 // CHECK9-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
4095 // CHECK9-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
4096 // CHECK9-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
4097 // CHECK9-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
4098 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
4099 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
4100 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
4101 // CHECK9-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
4102 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
4103 // CHECK9-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
4104 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
4105 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
4106 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
4107 // CHECK9-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
4108 // CHECK9-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
4109 // CHECK9-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
4110 // CHECK9-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META4:![0-9]+]])
4111 // CHECK9-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META7:![0-9]+]])
4112 // CHECK9-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META9:![0-9]+]])
4113 // CHECK9-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META11:![0-9]+]])
4114 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !13
4115 // CHECK9-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !13
4116 // CHECK9-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !13
4117 // CHECK9-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !13
4118 // CHECK9-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !13
4119 // CHECK9-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
4120 // CHECK9-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
4121 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM_I:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]]) #[[ATTR2:[0-9]+]]
4122 // CHECK9-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM_I]], i32 4) #[[ATTR2]]
4123 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4124 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
4125 // CHECK9:       .cancel.exit.i:
4126 // CHECK9-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
4127 // CHECK9-NEXT:    br label [[DOTOMP_OUTLINED__EXIT:%.*]]
4128 // CHECK9:       .cancel.continue.i:
4129 // CHECK9-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
4130 // CHECK9-NEXT:    br label [[DOTOMP_OUTLINED__EXIT]]
4131 // CHECK9:       .omp_outlined..exit:
4132 // CHECK9-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
4133 // CHECK9-NEXT:    ret i32 0
4134 //
4135 //
4136 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
4137 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5:[0-9]+]] {
4138 // CHECK9-NEXT:  entry:
4139 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4140 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4141 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
4142 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
4143 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
4144 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
4145 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
4146 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4147 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4148 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4149 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4150 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
4151 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
4152 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB17:[0-9]+]])
4153 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB15:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
4154 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4155 // CHECK9-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 0
4156 // CHECK9-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 0
4157 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
4158 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4159 // CHECK9-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4160 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4161 // CHECK9:       omp.inner.for.cond:
4162 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4163 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4164 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
4165 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4166 // CHECK9:       omp.inner.for.body:
4167 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4168 // CHECK9-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
4169 // CHECK9-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
4170 // CHECK9-NEXT:    ]
4171 // CHECK9:       .omp.sections.case:
4172 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4173 // CHECK9-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
4174 // CHECK9-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
4175 // CHECK9-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
4176 // CHECK9:       .omp.sections.case.split:
4177 // CHECK9-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
4178 // CHECK9:       .omp.sections.case.cncl:
4179 // CHECK9-NEXT:    br label [[CANCEL_CONT:%.*]]
4180 // CHECK9:       .omp.sections.exit:
4181 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4182 // CHECK9:       omp.inner.for.inc:
4183 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4184 // CHECK9-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
4185 // CHECK9-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4186 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
4187 // CHECK9:       omp.inner.for.end:
4188 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19:[0-9]+]])
4189 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
4190 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
4191 // CHECK9:       cancel.cont:
4192 // CHECK9-NEXT:    ret void
4193 // CHECK9:       cancel.exit:
4194 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19]])
4195 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
4196 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
4197 //
4198 //
4199 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
4200 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5]] {
4201 // CHECK9-NEXT:  entry:
4202 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4203 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4204 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
4205 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
4206 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
4207 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
4208 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
4209 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4210 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4211 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4212 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4213 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
4214 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
4215 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB21:[0-9]+]])
4216 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
4217 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4218 // CHECK9-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 1
4219 // CHECK9-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 1
4220 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
4221 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4222 // CHECK9-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4223 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4224 // CHECK9:       omp.inner.for.cond:
4225 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4226 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4227 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
4228 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4229 // CHECK9:       omp.inner.for.body:
4230 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4231 // CHECK9-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
4232 // CHECK9-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
4233 // CHECK9-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE2:%.*]]
4234 // CHECK9-NEXT:    ]
4235 // CHECK9:       .omp.sections.case:
4236 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4237 // CHECK9-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
4238 // CHECK9-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
4239 // CHECK9-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
4240 // CHECK9:       .omp.sections.case.split:
4241 // CHECK9-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
4242 // CHECK9:       .omp.sections.case.cncl:
4243 // CHECK9-NEXT:    br label [[CANCEL_CONT:%.*]]
4244 // CHECK9:       .omp.sections.case2:
4245 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4246 // CHECK9-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
4247 // CHECK9-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
4248 // CHECK9-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
4249 // CHECK9:       .omp.sections.case2.split:
4250 // CHECK9-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
4251 // CHECK9:       .omp.sections.case2.cncl:
4252 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_END]]
4253 // CHECK9:       .omp.sections.exit:
4254 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4255 // CHECK9:       omp.inner.for.inc:
4256 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4257 // CHECK9-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
4258 // CHECK9-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4259 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
4260 // CHECK9:       omp.inner.for.end:
4261 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM5:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23:[0-9]+]])
4262 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM5]])
4263 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
4264 // CHECK9:       cancel.cont:
4265 // CHECK9-NEXT:    ret void
4266 // CHECK9:       cancel.exit:
4267 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23]])
4268 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
4269 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
4270 //
4271 //
4272 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
4273 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR5]] {
4274 // CHECK9-NEXT:  entry:
4275 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4276 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4277 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
4278 // CHECK9-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
4279 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4280 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4281 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4282 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4283 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4284 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4285 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4286 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4287 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4288 // CHECK9-NEXT:    [[R3:%.*]] = alloca i32, align 4
4289 // CHECK9-NEXT:    [[I4:%.*]] = alloca i32, align 4
4290 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
4291 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4292 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4293 // CHECK9-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
4294 // CHECK9-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
4295 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
4296 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
4297 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
4298 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
4299 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4300 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
4301 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4302 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
4303 // CHECK9-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4304 // CHECK9-NEXT:    store i32 0, i32* [[I]], align 4
4305 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4306 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
4307 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4308 // CHECK9:       omp.precond.then:
4309 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4310 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4311 // CHECK9-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
4312 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4313 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4314 // CHECK9-NEXT:    store i32 0, i32* [[R3]], align 4
4315 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB25:[0-9]+]])
4316 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4317 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4318 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4319 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
4320 // CHECK9-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4321 // CHECK9:       cond.true:
4322 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4323 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4324 // CHECK9:       cond.false:
4325 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4326 // CHECK9-NEXT:    br label [[COND_END]]
4327 // CHECK9:       cond.end:
4328 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
4329 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4330 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4331 // CHECK9-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
4332 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4333 // CHECK9:       omp.inner.for.cond:
4334 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4335 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4336 // CHECK9-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
4337 // CHECK9-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4338 // CHECK9:       omp.inner.for.body:
4339 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4340 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1
4341 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4342 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
4343 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM7:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB27:[0-9]+]])
4344 // CHECK9-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM7]], i32 2)
4345 // CHECK9-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
4346 // CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
4347 // CHECK9:       .cancel.exit:
4348 // CHECK9-NEXT:    br label [[CANCEL_EXIT:%.*]]
4349 // CHECK9:       .cancel.continue:
4350 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I4]], align 4
4351 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[R3]], align 4
4352 // CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP16]]
4353 // CHECK9-NEXT:    store i32 [[ADD8]], i32* [[R3]], align 4
4354 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4355 // CHECK9:       omp.body.continue:
4356 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4357 // CHECK9:       omp.inner.for.inc:
4358 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4359 // CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], 1
4360 // CHECK9-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
4361 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
4362 // CHECK9:       omp.inner.for.end:
4363 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4364 // CHECK9:       omp.loop.exit:
4365 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29:[0-9]+]])
4366 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
4367 // CHECK9-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
4368 // CHECK9-NEXT:    [[TMP20:%.*]] = bitcast i32* [[R3]] to i8*
4369 // CHECK9-NEXT:    store i8* [[TMP20]], i8** [[TMP19]], align 8
4370 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
4371 // CHECK9-NEXT:    [[TMP21:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
4372 // CHECK9-NEXT:    [[TMP22:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB30:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]], i32 1, i64 8, i8* [[TMP21]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
4373 // CHECK9-NEXT:    switch i32 [[TMP22]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
4374 // CHECK9-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
4375 // CHECK9-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
4376 // CHECK9-NEXT:    ]
4377 // CHECK9:       .omp.reduction.case1:
4378 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP1]], align 4
4379 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[R3]], align 4
4380 // CHECK9-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
4381 // CHECK9-NEXT:    store i32 [[ADD13]], i32* [[TMP1]], align 4
4382 // CHECK9-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB30]], i32 [[OMP_GLOBAL_THREAD_NUM12]], [8 x i32]* @.gomp_critical_user_.reduction.var)
4383 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
4384 // CHECK9:       cancel.exit:
4385 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
4386 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
4387 // CHECK9-NEXT:    br label [[CANCEL_CONT:%.*]]
4388 // CHECK9:       .omp.reduction.case2:
4389 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[R3]], align 4
4390 // CHECK9-NEXT:    [[TMP26:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP25]] monotonic, align 4
4391 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
4392 // CHECK9:       .omp.reduction.default:
4393 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
4394 // CHECK9:       omp.precond.end:
4395 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
4396 // CHECK9:       cancel.cont:
4397 // CHECK9-NEXT:    ret void
4398 //
4399 //
4400 // CHECK9-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
4401 // CHECK9-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
4402 // CHECK9-NEXT:  entry:
4403 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
4404 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
4405 // CHECK9-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
4406 // CHECK9-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
4407 // CHECK9-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
4408 // CHECK9-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
4409 // CHECK9-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
4410 // CHECK9-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
4411 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
4412 // CHECK9-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
4413 // CHECK9-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
4414 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
4415 // CHECK9-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
4416 // CHECK9-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
4417 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
4418 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
4419 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4420 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
4421 // CHECK9-NEXT:    ret void
4422 //
4423 //
4424 // CHECK10-LABEL: define {{[^@]+}}@main
4425 // CHECK10-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
4426 // CHECK10-NEXT:  entry:
4427 // CHECK10-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
4428 // CHECK10-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
4429 // CHECK10-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
4430 // CHECK10-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
4431 // CHECK10-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
4432 // CHECK10-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
4433 // CHECK10-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
4434 // CHECK10-NEXT:    [[P_LASTITER26:%.*]] = alloca i32, align 4
4435 // CHECK10-NEXT:    [[P_LOWERBOUND27:%.*]] = alloca i32, align 4
4436 // CHECK10-NEXT:    [[P_UPPERBOUND28:%.*]] = alloca i32, align 4
4437 // CHECK10-NEXT:    [[P_STRIDE29:%.*]] = alloca i32, align 4
4438 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4439 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4440 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4441 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_33:%.*]] = alloca i32, align 4
4442 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
4443 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4444 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4445 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4446 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4447 // CHECK10-NEXT:    [[I35:%.*]] = alloca i32, align 4
4448 // CHECK10-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
4449 // CHECK10-NEXT:    [[R:%.*]] = alloca i32, align 4
4450 // CHECK10-NEXT:    store i32 0, i32* [[RETVAL]], align 4
4451 // CHECK10-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4452 // CHECK10-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
4453 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
4454 // CHECK10-NEXT:    br label [[OMP_PARALLEL:%.*]]
4455 // CHECK10:       omp_parallel:
4456 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8***)* @main..omp_par to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i8*** [[ARGV_ADDR]])
4457 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
4458 // CHECK10:       omp.par.outlined.exit:
4459 // CHECK10-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
4460 // CHECK10:       omp.par.exit.split:
4461 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
4462 // CHECK10:       omp_section_loop.preheader:
4463 // CHECK10-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
4464 // CHECK10-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
4465 // CHECK10-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
4466 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4467 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
4468 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
4469 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
4470 // CHECK10-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
4471 // CHECK10-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
4472 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
4473 // CHECK10:       omp_section_loop.header:
4474 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
4475 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
4476 // CHECK10:       omp_section_loop.cond:
4477 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
4478 // CHECK10-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
4479 // CHECK10:       omp_section_loop.body:
4480 // CHECK10-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
4481 // CHECK10-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
4482 // CHECK10-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
4483 // CHECK10-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
4484 // CHECK10-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
4485 // CHECK10-NEXT:    ]
4486 // CHECK10:       omp_section_loop.inc:
4487 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
4488 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
4489 // CHECK10:       omp_section_loop.exit:
4490 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
4491 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4492 // CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
4493 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
4494 // CHECK10:       omp_section_loop.after:
4495 // CHECK10-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
4496 // CHECK10:       omp_sections.end:
4497 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER12:%.*]]
4498 // CHECK10:       omp_section_loop.preheader12:
4499 // CHECK10-NEXT:    store i32 0, i32* [[P_LOWERBOUND27]], align 4
4500 // CHECK10-NEXT:    store i32 1, i32* [[P_UPPERBOUND28]], align 4
4501 // CHECK10-NEXT:    store i32 1, i32* [[P_STRIDE29]], align 4
4502 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM30:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4503 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]], i32 34, i32* [[P_LASTITER26]], i32* [[P_LOWERBOUND27]], i32* [[P_UPPERBOUND28]], i32* [[P_STRIDE29]], i32 1, i32 1)
4504 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND27]], align 4
4505 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND28]], align 4
4506 // CHECK10-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
4507 // CHECK10-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
4508 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13:%.*]]
4509 // CHECK10:       omp_section_loop.header13:
4510 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_IV19:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER12]] ], [ [[OMP_SECTION_LOOP_NEXT21:%.*]], [[OMP_SECTION_LOOP_INC16:%.*]] ]
4511 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_COND14:%.*]]
4512 // CHECK10:       omp_section_loop.cond14:
4513 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_CMP20:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV19]], [[TMP10]]
4514 // CHECK10-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP20]], label [[OMP_SECTION_LOOP_BODY15:%.*]], label [[OMP_SECTION_LOOP_EXIT17:%.*]]
4515 // CHECK10:       omp_section_loop.body15:
4516 // CHECK10-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV19]], [[TMP7]]
4517 // CHECK10-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
4518 // CHECK10-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
4519 // CHECK10-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC16]] [
4520 // CHECK10-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE22:%.*]]
4521 // CHECK10-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE24:%.*]]
4522 // CHECK10-NEXT:    ]
4523 // CHECK10:       omp_section_loop.inc16:
4524 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_NEXT21]] = add nuw i32 [[OMP_SECTION_LOOP_IV19]], 1
4525 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER13]]
4526 // CHECK10:       omp_section_loop.exit17:
4527 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM30]])
4528 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4529 // CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
4530 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_AFTER18:%.*]]
4531 // CHECK10:       omp_section_loop.after18:
4532 // CHECK10-NEXT:    br label [[OMP_SECTIONS_END32:%.*]]
4533 // CHECK10:       omp_sections.end32:
4534 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4535 // CHECK10-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
4536 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4537 // CHECK10-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
4538 // CHECK10-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4539 // CHECK10-NEXT:    [[SUB34:%.*]] = sub nsw i32 [[DIV]], 1
4540 // CHECK10-NEXT:    store i32 [[SUB34]], i32* [[DOTCAPTURE_EXPR_33]], align 4
4541 // CHECK10-NEXT:    store i32 0, i32* [[I]], align 4
4542 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4543 // CHECK10-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
4544 // CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4545 // CHECK10:       omp.precond.then:
4546 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4547 // CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
4548 // CHECK10-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
4549 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4550 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4551 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM36:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
4552 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM36]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4553 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4554 // CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
4555 // CHECK10-NEXT:    [[CMP37:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
4556 // CHECK10-NEXT:    br i1 [[CMP37]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4557 // CHECK10:       cond.true:
4558 // CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_33]], align 4
4559 // CHECK10-NEXT:    br label [[COND_END:%.*]]
4560 // CHECK10:       cond.false:
4561 // CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4562 // CHECK10-NEXT:    br label [[COND_END]]
4563 // CHECK10:       cond.end:
4564 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
4565 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4566 // CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4567 // CHECK10-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
4568 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4569 // CHECK10:       omp.inner.for.cond:
4570 // CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4571 // CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4572 // CHECK10-NEXT:    [[CMP38:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
4573 // CHECK10-NEXT:    br i1 [[CMP38]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4574 // CHECK10:       omp.inner.for.body:
4575 // CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4576 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
4577 // CHECK10-NEXT:    [[ADD39:%.*]] = add nsw i32 0, [[MUL]]
4578 // CHECK10-NEXT:    store i32 [[ADD39]], i32* [[I35]], align 4
4579 // CHECK10-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
4580 // CHECK10-NEXT:    [[TOBOOL40:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
4581 // CHECK10-NEXT:    br i1 [[TOBOOL40]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4582 // CHECK10:       omp_if.then:
4583 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM41:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
4584 // CHECK10-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM41]], i32 2)
4585 // CHECK10-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
4586 // CHECK10-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
4587 // CHECK10:       .cancel.exit:
4588 // CHECK10-NEXT:    br label [[CANCEL_EXIT:%.*]]
4589 // CHECK10:       omp_section_loop.body.case:
4590 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM9:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4591 // CHECK10-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM9]], i32 3)
4592 // CHECK10-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
4593 // CHECK10-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
4594 // CHECK10:       omp_section_loop.body.case.split:
4595 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
4596 // CHECK10:       omp_section_loop.body.case.cncl:
4597 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
4598 // CHECK10:       omp_section_loop.body.case22:
4599 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM23:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4600 // CHECK10-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM23]], i32 3)
4601 // CHECK10-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
4602 // CHECK10-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE22_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE22_CNCL:%.*]]
4603 // CHECK10:       omp_section_loop.body.case22.split:
4604 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
4605 // CHECK10:       omp_section_loop.body.case22.cncl:
4606 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
4607 // CHECK10:       omp_section_loop.body.case24:
4608 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM25:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4609 // CHECK10-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM25]], i32 3)
4610 // CHECK10-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
4611 // CHECK10-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE24_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE24_CNCL:%.*]]
4612 // CHECK10:       omp_section_loop.body.case24.split:
4613 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
4614 // CHECK10:       omp_section_loop.body.case24.cncl:
4615 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT17]]
4616 // CHECK10:       .cancel.continue:
4617 // CHECK10-NEXT:    br label [[OMP_IF_END:%.*]]
4618 // CHECK10:       omp_if.else:
4619 // CHECK10-NEXT:    br label [[OMP_IF_END]]
4620 // CHECK10:       omp_if.end:
4621 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4622 // CHECK10:       omp.body.continue:
4623 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4624 // CHECK10:       omp.inner.for.inc:
4625 // CHECK10-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4626 // CHECK10-NEXT:    [[ADD42:%.*]] = add nsw i32 [[TMP35]], 1
4627 // CHECK10-NEXT:    store i32 [[ADD42]], i32* [[DOTOMP_IV]], align 4
4628 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
4629 // CHECK10:       omp.inner.for.end:
4630 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4631 // CHECK10:       omp.loop.exit:
4632 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
4633 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
4634 // CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
4635 // CHECK10:       cancel.exit:
4636 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM43:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
4637 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM43]])
4638 // CHECK10-NEXT:    br label [[CANCEL_CONT:%.*]]
4639 // CHECK10:       omp.precond.end:
4640 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
4641 // CHECK10:       cancel.cont:
4642 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4643 // CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
4644 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
4645 // CHECK10-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM46]], 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*)*))
4646 // CHECK10-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
4647 // CHECK10-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
4648 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
4649 // CHECK10-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], i8* [[TMP36]])
4650 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
4651 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
4652 // CHECK10-NEXT:    store i32 0, i32* [[R]], align 4
4653 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i32* [[R]])
4654 // CHECK10-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4655 // CHECK10-NEXT:    ret i32 [[TMP40]]
4656 //
4657 //
4658 // CHECK10-LABEL: define {{[^@]+}}@main..omp_par
4659 // CHECK10-SAME: (i32* noalias [[TID_ADDR:%.*]], i32* noalias [[ZERO_ADDR:%.*]], i32* [[ARGC_ADDR:%.*]], i8*** [[ARGV_ADDR:%.*]]) #[[ATTR1:[0-9]+]] {
4660 // CHECK10-NEXT:  omp.par.entry:
4661 // CHECK10-NEXT:    [[TID_ADDR_LOCAL:%.*]] = alloca i32, align 4
4662 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[TID_ADDR]], align 4
4663 // CHECK10-NEXT:    store i32 [[TMP0]], i32* [[TID_ADDR_LOCAL]], align 4
4664 // CHECK10-NEXT:    [[TID:%.*]] = load i32, i32* [[TID_ADDR_LOCAL]], align 4
4665 // CHECK10-NEXT:    br label [[OMP_PAR_REGION:%.*]]
4666 // CHECK10:       omp.par.outlined.exit.exitStub:
4667 // CHECK10-NEXT:    ret void
4668 // CHECK10:       omp.par.region:
4669 // CHECK10-NEXT:    [[TMP1:%.*]] = load float, float* @flag, align 4
4670 // CHECK10-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP1]], 0.000000e+00
4671 // CHECK10-NEXT:    br i1 [[TOBOOL]], label [[TMP13:%.*]], label [[TMP2:%.*]]
4672 // CHECK10:       2:
4673 // CHECK10-NEXT:    br label [[TMP3:%.*]]
4674 // CHECK10:       3:
4675 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4676 // CHECK10-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP4]] to i8
4677 // CHECK10-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
4678 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 0
4679 // CHECK10-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
4680 // CHECK10-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
4681 // CHECK10-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX2]], align 1
4682 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4683 // CHECK10-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
4684 // CHECK10-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
4685 // CHECK10-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL4:%.*]]
4686 // CHECK10:       .cncl4:
4687 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
4688 // CHECK10:       .cont:
4689 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4690 // CHECK10-NEXT:    [[TMP10:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
4691 // CHECK10-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
4692 // CHECK10-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX5]], align 8
4693 // CHECK10-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
4694 // CHECK10-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX6]], align 1
4695 // CHECK10-NEXT:    [[CONV7:%.*]] = sext i8 [[TMP12]] to i32
4696 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV7]], [[TMP9]]
4697 // CHECK10-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD]] to i8
4698 // CHECK10-NEXT:    store i8 [[CONV8]], i8* [[ARRAYIDX6]], align 1
4699 // CHECK10-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
4700 // CHECK10:       omp.par.pre_finalize:
4701 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
4702 // CHECK10:       13:
4703 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4704 // CHECK10-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
4705 // CHECK10-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 0
4706 // CHECK10-NEXT:    br i1 [[TMP15]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
4707 // CHECK10:       .cncl:
4708 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
4709 // CHECK10:       .split:
4710 // CHECK10-NEXT:    br label [[TMP3]]
4711 //
4712 //
4713 // CHECK10-LABEL: define {{[^@]+}}@.omp_task_entry.
4714 // CHECK10-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
4715 // CHECK10-NEXT:  entry:
4716 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
4717 // CHECK10-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
4718 // CHECK10-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
4719 // CHECK10-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
4720 // CHECK10-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
4721 // CHECK10-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
4722 // CHECK10-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
4723 // CHECK10-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
4724 // CHECK10-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
4725 // CHECK10-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
4726 // CHECK10-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
4727 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
4728 // CHECK10-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
4729 // CHECK10-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
4730 // CHECK10-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
4731 // CHECK10-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
4732 // CHECK10-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
4733 // CHECK10-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
4734 // CHECK10-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
4735 // CHECK10-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META4:![0-9]+]])
4736 // CHECK10-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META7:![0-9]+]])
4737 // CHECK10-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META9:![0-9]+]])
4738 // CHECK10-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META11:![0-9]+]])
4739 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !13
4740 // CHECK10-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !13
4741 // CHECK10-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !13
4742 // CHECK10-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !13
4743 // CHECK10-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !13
4744 // CHECK10-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
4745 // CHECK10-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !13
4746 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM_I:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]]) #[[ATTR2:[0-9]+]]
4747 // CHECK10-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM_I]], i32 4) #[[ATTR2]]
4748 // CHECK10-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4749 // CHECK10-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
4750 // CHECK10:       .cancel.exit.i:
4751 // CHECK10-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
4752 // CHECK10-NEXT:    br label [[DOTOMP_OUTLINED__EXIT:%.*]]
4753 // CHECK10:       .cancel.continue.i:
4754 // CHECK10-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
4755 // CHECK10-NEXT:    br label [[DOTOMP_OUTLINED__EXIT]]
4756 // CHECK10:       .omp_outlined..exit:
4757 // CHECK10-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !13
4758 // CHECK10-NEXT:    ret i32 0
4759 //
4760 //
4761 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
4762 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5:[0-9]+]] {
4763 // CHECK10-NEXT:  entry:
4764 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4765 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4766 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
4767 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
4768 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
4769 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
4770 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
4771 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4772 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4773 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4774 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4775 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
4776 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
4777 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB17:[0-9]+]])
4778 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB15:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
4779 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4780 // CHECK10-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 0
4781 // CHECK10-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 0
4782 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
4783 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4784 // CHECK10-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4785 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4786 // CHECK10:       omp.inner.for.cond:
4787 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4788 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4789 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
4790 // CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4791 // CHECK10:       omp.inner.for.body:
4792 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4793 // CHECK10-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
4794 // CHECK10-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
4795 // CHECK10-NEXT:    ]
4796 // CHECK10:       .omp.sections.case:
4797 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4798 // CHECK10-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
4799 // CHECK10-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
4800 // CHECK10-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
4801 // CHECK10:       .omp.sections.case.split:
4802 // CHECK10-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
4803 // CHECK10:       .omp.sections.case.cncl:
4804 // CHECK10-NEXT:    br label [[CANCEL_CONT:%.*]]
4805 // CHECK10:       .omp.sections.exit:
4806 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4807 // CHECK10:       omp.inner.for.inc:
4808 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4809 // CHECK10-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
4810 // CHECK10-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4811 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
4812 // CHECK10:       omp.inner.for.end:
4813 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19:[0-9]+]])
4814 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
4815 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
4816 // CHECK10:       cancel.cont:
4817 // CHECK10-NEXT:    ret void
4818 // CHECK10:       cancel.exit:
4819 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19]])
4820 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
4821 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
4822 //
4823 //
4824 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
4825 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5]] {
4826 // CHECK10-NEXT:  entry:
4827 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4828 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4829 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
4830 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
4831 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
4832 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
4833 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
4834 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4835 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4836 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4837 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4838 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
4839 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
4840 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB21:[0-9]+]])
4841 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
4842 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4843 // CHECK10-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 1
4844 // CHECK10-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 1
4845 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
4846 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4847 // CHECK10-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4848 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4849 // CHECK10:       omp.inner.for.cond:
4850 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4851 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4852 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
4853 // CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4854 // CHECK10:       omp.inner.for.body:
4855 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4856 // CHECK10-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
4857 // CHECK10-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
4858 // CHECK10-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE2:%.*]]
4859 // CHECK10-NEXT:    ]
4860 // CHECK10:       .omp.sections.case:
4861 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4862 // CHECK10-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
4863 // CHECK10-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
4864 // CHECK10-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
4865 // CHECK10:       .omp.sections.case.split:
4866 // CHECK10-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
4867 // CHECK10:       .omp.sections.case.cncl:
4868 // CHECK10-NEXT:    br label [[CANCEL_CONT:%.*]]
4869 // CHECK10:       .omp.sections.case2:
4870 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4871 // CHECK10-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
4872 // CHECK10-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
4873 // CHECK10-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
4874 // CHECK10:       .omp.sections.case2.split:
4875 // CHECK10-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
4876 // CHECK10:       .omp.sections.case2.cncl:
4877 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_END]]
4878 // CHECK10:       .omp.sections.exit:
4879 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4880 // CHECK10:       omp.inner.for.inc:
4881 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4882 // CHECK10-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
4883 // CHECK10-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4884 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
4885 // CHECK10:       omp.inner.for.end:
4886 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM5:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23:[0-9]+]])
4887 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM5]])
4888 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
4889 // CHECK10:       cancel.cont:
4890 // CHECK10-NEXT:    ret void
4891 // CHECK10:       cancel.exit:
4892 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23]])
4893 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
4894 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
4895 //
4896 //
4897 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
4898 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR5]] {
4899 // CHECK10-NEXT:  entry:
4900 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4901 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4902 // CHECK10-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
4903 // CHECK10-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
4904 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4905 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4906 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4907 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4908 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
4909 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4910 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4911 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4912 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4913 // CHECK10-NEXT:    [[R3:%.*]] = alloca i32, align 4
4914 // CHECK10-NEXT:    [[I4:%.*]] = alloca i32, align 4
4915 // CHECK10-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
4916 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4917 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4918 // CHECK10-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
4919 // CHECK10-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
4920 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
4921 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
4922 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
4923 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
4924 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4925 // CHECK10-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
4926 // CHECK10-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4927 // CHECK10-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
4928 // CHECK10-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4929 // CHECK10-NEXT:    store i32 0, i32* [[I]], align 4
4930 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4931 // CHECK10-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
4932 // CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4933 // CHECK10:       omp.precond.then:
4934 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4935 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4936 // CHECK10-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
4937 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4938 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4939 // CHECK10-NEXT:    store i32 0, i32* [[R3]], align 4
4940 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB25:[0-9]+]])
4941 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4942 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4943 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4944 // CHECK10-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
4945 // CHECK10-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4946 // CHECK10:       cond.true:
4947 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4948 // CHECK10-NEXT:    br label [[COND_END:%.*]]
4949 // CHECK10:       cond.false:
4950 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4951 // CHECK10-NEXT:    br label [[COND_END]]
4952 // CHECK10:       cond.end:
4953 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
4954 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4955 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4956 // CHECK10-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
4957 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4958 // CHECK10:       omp.inner.for.cond:
4959 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4960 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4961 // CHECK10-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
4962 // CHECK10-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4963 // CHECK10:       omp.inner.for.body:
4964 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4965 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1
4966 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4967 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
4968 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM7:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB27:[0-9]+]])
4969 // CHECK10-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM7]], i32 2)
4970 // CHECK10-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
4971 // CHECK10-NEXT:    br i1 [[TMP15]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
4972 // CHECK10:       .cancel.exit:
4973 // CHECK10-NEXT:    br label [[CANCEL_EXIT:%.*]]
4974 // CHECK10:       .cancel.continue:
4975 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I4]], align 4
4976 // CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[R3]], align 4
4977 // CHECK10-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP16]]
4978 // CHECK10-NEXT:    store i32 [[ADD8]], i32* [[R3]], align 4
4979 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4980 // CHECK10:       omp.body.continue:
4981 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4982 // CHECK10:       omp.inner.for.inc:
4983 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4984 // CHECK10-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], 1
4985 // CHECK10-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
4986 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
4987 // CHECK10:       omp.inner.for.end:
4988 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4989 // CHECK10:       omp.loop.exit:
4990 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29:[0-9]+]])
4991 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
4992 // CHECK10-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
4993 // CHECK10-NEXT:    [[TMP20:%.*]] = bitcast i32* [[R3]] to i8*
4994 // CHECK10-NEXT:    store i8* [[TMP20]], i8** [[TMP19]], align 8
4995 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
4996 // CHECK10-NEXT:    [[TMP21:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
4997 // CHECK10-NEXT:    [[TMP22:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB30:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]], i32 1, i64 8, i8* [[TMP21]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
4998 // CHECK10-NEXT:    switch i32 [[TMP22]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
4999 // CHECK10-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
5000 // CHECK10-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
5001 // CHECK10-NEXT:    ]
5002 // CHECK10:       .omp.reduction.case1:
5003 // CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP1]], align 4
5004 // CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[R3]], align 4
5005 // CHECK10-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
5006 // CHECK10-NEXT:    store i32 [[ADD13]], i32* [[TMP1]], align 4
5007 // CHECK10-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB30]], i32 [[OMP_GLOBAL_THREAD_NUM12]], [8 x i32]* @.gomp_critical_user_.reduction.var)
5008 // CHECK10-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5009 // CHECK10:       cancel.exit:
5010 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
5011 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
5012 // CHECK10-NEXT:    br label [[CANCEL_CONT:%.*]]
5013 // CHECK10:       .omp.reduction.case2:
5014 // CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[R3]], align 4
5015 // CHECK10-NEXT:    [[TMP26:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP25]] monotonic, align 4
5016 // CHECK10-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5017 // CHECK10:       .omp.reduction.default:
5018 // CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
5019 // CHECK10:       omp.precond.end:
5020 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
5021 // CHECK10:       cancel.cont:
5022 // CHECK10-NEXT:    ret void
5023 //
5024 //
5025 // CHECK10-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
5026 // CHECK10-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
5027 // CHECK10-NEXT:  entry:
5028 // CHECK10-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
5029 // CHECK10-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
5030 // CHECK10-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
5031 // CHECK10-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
5032 // CHECK10-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
5033 // CHECK10-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
5034 // CHECK10-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
5035 // CHECK10-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
5036 // CHECK10-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
5037 // CHECK10-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
5038 // CHECK10-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
5039 // CHECK10-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
5040 // CHECK10-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
5041 // CHECK10-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
5042 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
5043 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
5044 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5045 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
5046 // CHECK10-NEXT:    ret void
5047 //
5048 //