xref: /llvm-project/clang/test/OpenMP/cancel_codegen.cpp (revision f193bcc701de8311e9f40730f2187d6df2e8667c)
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* @[[GLOB4:[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* @[[GLOB4]], i32 [[TMP0]])
153 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
154 // CHECK1:       cancel.cont:
155 // CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5:[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* @[[GLOB4]], 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* @[[GLOB4]], 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* @[[GLOB4]], i32 [[TMP0]])
207 // CHECK1-NEXT:    br label [[CANCEL_CONT20:%.*]]
208 // CHECK1:       cancel.cont20:
209 // CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5]], 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* @[[GLOB6:[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* @[[GLOB4]], 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* @[[GLOB6]], 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* @[[GLOB6]], 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* @[[GLOB2:[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:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
327 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
328 // CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]])
329 // CHECK1-NEXT:    br label [[RETURN:%.*]]
330 // CHECK1:       .cancel.continue:
331 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
332 // CHECK1:       omp_if.else:
333 // CHECK1-NEXT:    br label [[OMP_IF_END]]
334 // CHECK1:       omp_if.end:
335 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP1]], align 4
336 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP10]] to i8
337 // CHECK1-NEXT:    [[TMP11:%.*]] = load i8**, i8*** [[TMP0]], align 8
338 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
339 // CHECK1-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
340 // CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
341 // CHECK1-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX1]], align 1
342 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
343 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
344 // CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP14]])
345 // CHECK1-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
346 // CHECK1-NEXT:    br i1 [[TMP16]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
347 // CHECK1:       .cancel.exit2:
348 // CHECK1-NEXT:    br label [[RETURN]]
349 // CHECK1:       .cancel.continue3:
350 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP1]], align 4
351 // CHECK1-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP0]], align 8
352 // CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP18]], i64 0
353 // CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
354 // CHECK1-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP19]], i64 0
355 // CHECK1-NEXT:    [[TMP20:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
356 // CHECK1-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP20]] to i32
357 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP17]]
358 // CHECK1-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD]] to i8
359 // CHECK1-NEXT:    store i8 [[CONV7]], i8* [[ARRAYIDX5]], align 1
360 // CHECK1-NEXT:    br label [[RETURN]]
361 // CHECK1:       return:
362 // CHECK1-NEXT:    ret void
363 //
364 //
365 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
366 // CHECK1-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
367 // CHECK1-NEXT:  entry:
368 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
369 // CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
370 // CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
371 // CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
372 // CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
373 // CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
374 // CHECK1-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
375 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
376 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
377 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
378 // CHECK1-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
379 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
380 // CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
381 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
382 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
383 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
384 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
385 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
386 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
387 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META5:![0-9]+]])
388 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
389 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
390 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
391 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
392 // CHECK1-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !14
393 // CHECK1-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !14
394 // CHECK1-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !14
395 // CHECK1-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !14
396 // CHECK1-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
397 // CHECK1-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
398 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
399 // CHECK1-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 4) #[[ATTR2:[0-9]+]]
400 // CHECK1-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
401 // CHECK1-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
402 // CHECK1:       .cancel.exit.i:
403 // CHECK1-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
404 // CHECK1-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
405 // CHECK1:       .cancel.continue.i:
406 // CHECK1-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
407 // CHECK1-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
408 // CHECK1:       .omp_outlined..1.exit:
409 // CHECK1-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
410 // CHECK1-NEXT:    ret i32 0
411 //
412 //
413 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
414 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
415 // CHECK1-NEXT:  entry:
416 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
417 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
418 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
419 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
420 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
421 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
422 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
423 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
424 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
425 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
426 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
427 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
428 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
429 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
430 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
431 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
432 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
433 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
434 // CHECK1-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
435 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
436 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
437 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
438 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
439 // CHECK1:       omp.inner.for.cond:
440 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
441 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
442 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
443 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
444 // CHECK1:       omp.inner.for.body:
445 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
446 // CHECK1-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
447 // CHECK1-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
448 // CHECK1-NEXT:    ]
449 // CHECK1:       .omp.sections.case:
450 // CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
451 // CHECK1-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
452 // CHECK1-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
453 // CHECK1:       .cancel.exit:
454 // CHECK1-NEXT:    br label [[CANCEL_EXIT:%.*]]
455 // CHECK1:       .cancel.continue:
456 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
457 // CHECK1:       .omp.sections.exit:
458 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
459 // CHECK1:       omp.inner.for.inc:
460 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
461 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
462 // CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
463 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
464 // CHECK1:       omp.inner.for.end:
465 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
466 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
467 // CHECK1:       cancel.cont:
468 // CHECK1-NEXT:    ret void
469 // CHECK1:       cancel.exit:
470 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
471 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
472 //
473 //
474 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
475 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
476 // CHECK1-NEXT:  entry:
477 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
478 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
479 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
480 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
481 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
482 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
483 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
484 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
485 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
486 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
487 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
488 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
489 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
490 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
491 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
492 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
493 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
494 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
495 // CHECK1-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
496 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
497 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
498 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
499 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
500 // CHECK1:       omp.inner.for.cond:
501 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
502 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
503 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
504 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
505 // CHECK1:       omp.inner.for.body:
506 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
507 // CHECK1-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
508 // CHECK1-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
509 // CHECK1-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE1:%.*]]
510 // CHECK1-NEXT:    ]
511 // CHECK1:       .omp.sections.case:
512 // CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
513 // CHECK1-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
514 // CHECK1-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
515 // CHECK1:       .cancel.exit:
516 // CHECK1-NEXT:    br label [[CANCEL_EXIT:%.*]]
517 // CHECK1:       .cancel.continue:
518 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
519 // CHECK1:       .omp.sections.case1:
520 // CHECK1-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
521 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
522 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
523 // CHECK1:       .cancel.exit2:
524 // CHECK1-NEXT:    br label [[CANCEL_EXIT]]
525 // CHECK1:       .cancel.continue3:
526 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
527 // CHECK1:       .omp.sections.exit:
528 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
529 // CHECK1:       omp.inner.for.inc:
530 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
531 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
532 // CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
533 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
534 // CHECK1:       omp.inner.for.end:
535 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
536 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
537 // CHECK1:       cancel.cont:
538 // CHECK1-NEXT:    ret void
539 // CHECK1:       cancel.exit:
540 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
541 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
542 //
543 //
544 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
545 // 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]] {
546 // CHECK1-NEXT:  entry:
547 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
548 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
549 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
550 // CHECK1-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
551 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
552 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
553 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
554 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
555 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
556 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
557 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
558 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
559 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
560 // CHECK1-NEXT:    [[R3:%.*]] = alloca i32, align 4
561 // CHECK1-NEXT:    [[I4:%.*]] = alloca i32, align 4
562 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
563 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
564 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
565 // CHECK1-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
566 // CHECK1-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
567 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
568 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
569 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
570 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
571 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
572 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
573 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
574 // CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
575 // CHECK1-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
576 // CHECK1-NEXT:    store i32 0, i32* [[I]], align 4
577 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
578 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
579 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
580 // CHECK1:       omp.precond.then:
581 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
582 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
583 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
584 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
585 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
586 // CHECK1-NEXT:    store i32 0, i32* [[R3]], align 4
587 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
588 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
589 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
590 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
591 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
592 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
593 // CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
594 // CHECK1:       cond.true:
595 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
596 // CHECK1-NEXT:    br label [[COND_END:%.*]]
597 // CHECK1:       cond.false:
598 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
599 // CHECK1-NEXT:    br label [[COND_END]]
600 // CHECK1:       cond.end:
601 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
602 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
603 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
604 // CHECK1-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
605 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
606 // CHECK1:       omp.inner.for.cond:
607 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
608 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
609 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
610 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
611 // CHECK1:       omp.inner.for.body:
612 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
613 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
614 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
615 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
616 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
617 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4
618 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP17]], i32 2)
619 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
620 // CHECK1-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
621 // CHECK1:       .cancel.exit:
622 // CHECK1-NEXT:    br label [[CANCEL_EXIT:%.*]]
623 // CHECK1:       .cancel.continue:
624 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4
625 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[R3]], align 4
626 // CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP20]]
627 // CHECK1-NEXT:    store i32 [[ADD7]], i32* [[R3]], align 4
628 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
629 // CHECK1:       omp.body.continue:
630 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
631 // CHECK1:       omp.inner.for.inc:
632 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
633 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
634 // CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
635 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
636 // CHECK1:       omp.inner.for.end:
637 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
638 // CHECK1:       omp.loop.exit:
639 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
640 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
641 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP24]])
642 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
643 // CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i32* [[R3]] to i8*
644 // CHECK1-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
645 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
646 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
647 // CHECK1-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
648 // 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)
649 // CHECK1-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
650 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
651 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
652 // CHECK1-NEXT:    ]
653 // CHECK1:       .omp.reduction.case1:
654 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP1]], align 4
655 // CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[R3]], align 4
656 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
657 // CHECK1-NEXT:    store i32 [[ADD9]], i32* [[TMP1]], align 4
658 // CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB7]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
659 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
660 // CHECK1:       cancel.exit:
661 // CHECK1-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
662 // CHECK1-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
663 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP34]])
664 // CHECK1-NEXT:    br label [[CANCEL_CONT:%.*]]
665 // CHECK1:       .omp.reduction.case2:
666 // CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[R3]], align 4
667 // CHECK1-NEXT:    [[TMP36:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP35]] monotonic, align 4
668 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
669 // CHECK1:       .omp.reduction.default:
670 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
671 // CHECK1:       omp.precond.end:
672 // CHECK1-NEXT:    br label [[CANCEL_CONT]]
673 // CHECK1:       cancel.cont:
674 // CHECK1-NEXT:    ret void
675 //
676 //
677 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
678 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
679 // CHECK1-NEXT:  entry:
680 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
681 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
682 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
683 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
684 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
685 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
686 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
687 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
688 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
689 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
690 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
691 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
692 // CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
693 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
694 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
695 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
696 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
697 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
698 // CHECK1-NEXT:    ret void
699 //
700 //
701 // CHECK2-LABEL: define {{[^@]+}}@main
702 // CHECK2-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
703 // CHECK2-NEXT:  entry:
704 // CHECK2-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
705 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
706 // CHECK2-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
707 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
708 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
709 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
710 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
711 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
712 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_LB_1:%.*]] = alloca i32, align 4
713 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_UB_2:%.*]] = alloca i32, align 4
714 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_ST_3:%.*]] = alloca i32, align 4
715 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IL_4:%.*]] = alloca i32, align 4
716 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IV_5:%.*]] = alloca i32, align 4
717 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
718 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
719 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
720 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_21:%.*]] = alloca i32, align 4
721 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
722 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
723 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
724 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
725 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
726 // CHECK2-NEXT:    [[I24:%.*]] = alloca i32, align 4
727 // CHECK2-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
728 // CHECK2-NEXT:    [[R:%.*]] = alloca i32, align 4
729 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
730 // CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
731 // CHECK2-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
732 // CHECK2-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
733 // 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]])
734 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
735 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
736 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
737 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
738 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[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)
739 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
740 // CHECK2-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
741 // CHECK2-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
742 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
743 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
744 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
745 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
746 // CHECK2:       omp.inner.for.cond:
747 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
748 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
749 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
750 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
751 // CHECK2:       omp.inner.for.body:
752 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
753 // CHECK2-NEXT:    switch i32 [[TMP7]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
754 // CHECK2-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
755 // CHECK2-NEXT:    ]
756 // CHECK2:       .omp.sections.case:
757 // CHECK2-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
758 // CHECK2-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
759 // CHECK2-NEXT:    br i1 [[TMP9]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
760 // CHECK2:       .cancel.exit:
761 // CHECK2-NEXT:    br label [[CANCEL_EXIT:%.*]]
762 // CHECK2:       .cancel.continue:
763 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
764 // CHECK2:       .omp.sections.exit:
765 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
766 // CHECK2:       omp.inner.for.inc:
767 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
768 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP10]], 1
769 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
770 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
771 // CHECK2:       omp.inner.for.end:
772 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
773 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
774 // CHECK2:       cancel.cont:
775 // CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]])
776 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
777 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
778 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_3]], align 4
779 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_4]], align 4
780 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], 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)
781 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
782 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp slt i32 [[TMP11]], 1
783 // CHECK2-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 1
784 // CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_SECTIONS_UB_2]], align 4
785 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
786 // CHECK2-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
787 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND6:%.*]]
788 // CHECK2:       omp.inner.for.cond6:
789 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
790 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
791 // CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
792 // CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY8:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
793 // CHECK2:       omp.inner.for.body8:
794 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
795 // CHECK2-NEXT:    switch i32 [[TMP17]], label [[DOTOMP_SECTIONS_EXIT15:%.*]] [
796 // CHECK2-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE9:%.*]]
797 // CHECK2-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE12:%.*]]
798 // CHECK2-NEXT:    ]
799 // CHECK2:       .omp.sections.case9:
800 // CHECK2-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
801 // CHECK2-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
802 // CHECK2-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT10:%.*]], label [[DOTCANCEL_CONTINUE11:%.*]]
803 // CHECK2:       .cancel.exit10:
804 // CHECK2-NEXT:    br label [[CANCEL_EXIT19:%.*]]
805 // CHECK2:       cancel.exit:
806 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
807 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
808 // CHECK2:       .cancel.continue11:
809 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
810 // CHECK2:       .omp.sections.case12:
811 // CHECK2-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
812 // CHECK2-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
813 // CHECK2-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT13:%.*]], label [[DOTCANCEL_CONTINUE14:%.*]]
814 // CHECK2:       .cancel.exit13:
815 // CHECK2-NEXT:    br label [[CANCEL_EXIT19]]
816 // CHECK2:       .cancel.continue14:
817 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
818 // CHECK2:       .omp.sections.exit15:
819 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
820 // CHECK2:       omp.inner.for.inc16:
821 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
822 // CHECK2-NEXT:    [[INC17:%.*]] = add nsw i32 [[TMP22]], 1
823 // CHECK2-NEXT:    store i32 [[INC17]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
824 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND6]]
825 // CHECK2:       omp.inner.for.end18:
826 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
827 // CHECK2-NEXT:    br label [[CANCEL_CONT20:%.*]]
828 // CHECK2:       cancel.cont20:
829 // CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
830 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
831 // CHECK2-NEXT:    store i32 [[TMP23]], i32* [[DOTCAPTURE_EXPR_]], align 4
832 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
833 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
834 // CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
835 // CHECK2-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[DIV]], 1
836 // CHECK2-NEXT:    store i32 [[SUB22]], i32* [[DOTCAPTURE_EXPR_21]], align 4
837 // CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
838 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
839 // CHECK2-NEXT:    [[CMP23:%.*]] = icmp slt i32 0, [[TMP25]]
840 // CHECK2-NEXT:    br i1 [[CMP23]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
841 // CHECK2:       omp.precond.then:
842 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
843 // CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
844 // CHECK2-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_UB]], align 4
845 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
846 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
847 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
848 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
849 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
850 // CHECK2-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
851 // CHECK2-NEXT:    br i1 [[CMP25]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
852 // CHECK2:       cond.true:
853 // CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
854 // CHECK2-NEXT:    br label [[COND_END:%.*]]
855 // CHECK2:       cond.false:
856 // CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
857 // CHECK2-NEXT:    br label [[COND_END]]
858 // CHECK2:       cond.end:
859 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE]] ], [ [[TMP30]], [[COND_FALSE]] ]
860 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
861 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
862 // CHECK2-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
863 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND26:%.*]]
864 // CHECK2:       omp.inner.for.cond26:
865 // CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
866 // CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
867 // CHECK2-NEXT:    [[CMP27:%.*]] = icmp sle i32 [[TMP32]], [[TMP33]]
868 // CHECK2-NEXT:    br i1 [[CMP27]], label [[OMP_INNER_FOR_BODY28:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
869 // CHECK2:       omp.inner.for.body28:
870 // CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
871 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP34]], 1
872 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
873 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I24]], align 4
874 // CHECK2-NEXT:    [[TMP35:%.*]] = load float, float* @flag, align 4
875 // CHECK2-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP35]], 0.000000e+00
876 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
877 // CHECK2:       omp_if.then:
878 // CHECK2-NEXT:    [[TMP36:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
879 // CHECK2-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
880 // CHECK2-NEXT:    br i1 [[TMP37]], label [[DOTCANCEL_EXIT29:%.*]], label [[DOTCANCEL_CONTINUE30:%.*]]
881 // CHECK2:       .cancel.exit29:
882 // CHECK2-NEXT:    br label [[CANCEL_EXIT34:%.*]]
883 // CHECK2:       cancel.exit19:
884 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
885 // CHECK2-NEXT:    br label [[CANCEL_CONT20]]
886 // CHECK2:       .cancel.continue30:
887 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
888 // CHECK2:       omp_if.else:
889 // CHECK2-NEXT:    br label [[OMP_IF_END]]
890 // CHECK2:       omp_if.end:
891 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
892 // CHECK2:       omp.body.continue:
893 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
894 // CHECK2:       omp.inner.for.inc31:
895 // CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
896 // CHECK2-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP38]], 1
897 // CHECK2-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV]], align 4
898 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND26]]
899 // CHECK2:       omp.inner.for.end33:
900 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
901 // CHECK2:       omp.loop.exit:
902 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
903 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
904 // CHECK2:       cancel.exit34:
905 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
906 // CHECK2-NEXT:    br label [[CANCEL_CONT35:%.*]]
907 // CHECK2:       omp.precond.end:
908 // CHECK2-NEXT:    br label [[CANCEL_CONT35]]
909 // CHECK2:       cancel.cont35:
910 // CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
911 // 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*)*))
912 // CHECK2-NEXT:    [[TMP40:%.*]] = bitcast i8* [[TMP39]] to %struct.kmp_task_t_with_privates*
913 // CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP40]], i32 0, i32 0
914 // CHECK2-NEXT:    [[TMP42:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP39]])
915 // 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*, ...)*))
916 // 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*, ...)*))
917 // CHECK2-NEXT:    store i32 0, i32* [[R]], align 4
918 // 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]])
919 // CHECK2-NEXT:    [[TMP43:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
920 // CHECK2-NEXT:    ret i32 [[TMP43]]
921 //
922 //
923 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
924 // 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]+]] {
925 // CHECK2-NEXT:  entry:
926 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
927 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
928 // CHECK2-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8***, align 8
929 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
930 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
931 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
932 // CHECK2-NEXT:    store i8*** [[ARGV]], i8**** [[ARGV_ADDR]], align 8
933 // CHECK2-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
934 // CHECK2-NEXT:    [[TMP0:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
935 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
936 // CHECK2-NEXT:    [[TMP2:%.*]] = load float, float* @flag, align 4
937 // CHECK2-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP2]], 0.000000e+00
938 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
939 // CHECK2:       omp_if.then:
940 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
941 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
942 // CHECK2-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 1)
943 // CHECK2-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[TMP5]], 0
944 // CHECK2-NEXT:    br i1 [[TMP6]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
945 // CHECK2:       .cancel.exit:
946 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
947 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
948 // CHECK2-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]])
949 // CHECK2-NEXT:    br label [[RETURN:%.*]]
950 // CHECK2:       .cancel.continue:
951 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
952 // CHECK2:       omp_if.else:
953 // CHECK2-NEXT:    br label [[OMP_IF_END]]
954 // CHECK2:       omp_if.end:
955 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP1]], align 4
956 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP10]] to i8
957 // CHECK2-NEXT:    [[TMP11:%.*]] = load i8**, i8*** [[TMP0]], align 8
958 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
959 // CHECK2-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
960 // CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
961 // CHECK2-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX1]], align 1
962 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
963 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
964 // CHECK2-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP14]])
965 // CHECK2-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
966 // CHECK2-NEXT:    br i1 [[TMP16]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
967 // CHECK2:       .cancel.exit2:
968 // CHECK2-NEXT:    br label [[RETURN]]
969 // CHECK2:       .cancel.continue3:
970 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP1]], align 4
971 // CHECK2-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP0]], align 8
972 // CHECK2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP18]], i64 0
973 // CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
974 // CHECK2-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP19]], i64 0
975 // CHECK2-NEXT:    [[TMP20:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
976 // CHECK2-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP20]] to i32
977 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP17]]
978 // CHECK2-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD]] to i8
979 // CHECK2-NEXT:    store i8 [[CONV7]], i8* [[ARRAYIDX5]], align 1
980 // CHECK2-NEXT:    br label [[RETURN]]
981 // CHECK2:       return:
982 // CHECK2-NEXT:    ret void
983 //
984 //
985 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
986 // CHECK2-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
987 // CHECK2-NEXT:  entry:
988 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
989 // CHECK2-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
990 // CHECK2-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
991 // CHECK2-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
992 // CHECK2-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
993 // CHECK2-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
994 // CHECK2-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
995 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
996 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
997 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
998 // CHECK2-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
999 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
1000 // CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
1001 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
1002 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
1003 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
1004 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1005 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
1006 // CHECK2-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
1007 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META5:![0-9]+]])
1008 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
1009 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
1010 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
1011 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
1012 // CHECK2-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !14
1013 // CHECK2-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !14
1014 // CHECK2-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !14
1015 // CHECK2-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !14
1016 // CHECK2-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
1017 // CHECK2-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
1018 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
1019 // CHECK2-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 4) #[[ATTR2:[0-9]+]]
1020 // CHECK2-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
1021 // CHECK2-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
1022 // CHECK2:       .cancel.exit.i:
1023 // CHECK2-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
1024 // CHECK2-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
1025 // CHECK2:       .cancel.continue.i:
1026 // CHECK2-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
1027 // CHECK2-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
1028 // CHECK2:       .omp_outlined..1.exit:
1029 // CHECK2-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
1030 // CHECK2-NEXT:    ret i32 0
1031 //
1032 //
1033 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
1034 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1035 // CHECK2-NEXT:  entry:
1036 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1037 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1038 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
1039 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
1040 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
1041 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
1042 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
1043 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1044 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1045 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1046 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1047 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
1048 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
1049 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1050 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1051 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
1052 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1053 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
1054 // CHECK2-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
1055 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
1056 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1057 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1058 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1059 // CHECK2:       omp.inner.for.cond:
1060 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1061 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1062 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1063 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1064 // CHECK2:       omp.inner.for.body:
1065 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1066 // CHECK2-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
1067 // CHECK2-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
1068 // CHECK2-NEXT:    ]
1069 // CHECK2:       .omp.sections.case:
1070 // CHECK2-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
1071 // CHECK2-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
1072 // CHECK2-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
1073 // CHECK2:       .cancel.exit:
1074 // CHECK2-NEXT:    br label [[CANCEL_EXIT:%.*]]
1075 // CHECK2:       .cancel.continue:
1076 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1077 // CHECK2:       .omp.sections.exit:
1078 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1079 // CHECK2:       omp.inner.for.inc:
1080 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1081 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
1082 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1083 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
1084 // CHECK2:       omp.inner.for.end:
1085 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
1086 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
1087 // CHECK2:       cancel.cont:
1088 // CHECK2-NEXT:    ret void
1089 // CHECK2:       cancel.exit:
1090 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
1091 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
1092 //
1093 //
1094 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
1095 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1096 // CHECK2-NEXT:  entry:
1097 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1098 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1099 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
1100 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
1101 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
1102 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
1103 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
1104 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1105 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1106 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1107 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1108 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
1109 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
1110 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1111 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1112 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
1113 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1114 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
1115 // CHECK2-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
1116 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
1117 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1118 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1119 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1120 // CHECK2:       omp.inner.for.cond:
1121 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1122 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1123 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1124 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1125 // CHECK2:       omp.inner.for.body:
1126 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1127 // CHECK2-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
1128 // CHECK2-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
1129 // CHECK2-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE1:%.*]]
1130 // CHECK2-NEXT:    ]
1131 // CHECK2:       .omp.sections.case:
1132 // CHECK2-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
1133 // CHECK2-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
1134 // CHECK2-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
1135 // CHECK2:       .cancel.exit:
1136 // CHECK2-NEXT:    br label [[CANCEL_EXIT:%.*]]
1137 // CHECK2:       .cancel.continue:
1138 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1139 // CHECK2:       .omp.sections.case1:
1140 // CHECK2-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
1141 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1142 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
1143 // CHECK2:       .cancel.exit2:
1144 // CHECK2-NEXT:    br label [[CANCEL_EXIT]]
1145 // CHECK2:       .cancel.continue3:
1146 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1147 // CHECK2:       .omp.sections.exit:
1148 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1149 // CHECK2:       omp.inner.for.inc:
1150 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1151 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
1152 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1153 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
1154 // CHECK2:       omp.inner.for.end:
1155 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
1156 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
1157 // CHECK2:       cancel.cont:
1158 // CHECK2-NEXT:    ret void
1159 // CHECK2:       cancel.exit:
1160 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
1161 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
1162 //
1163 //
1164 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
1165 // 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]] {
1166 // CHECK2-NEXT:  entry:
1167 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1168 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1169 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
1170 // CHECK2-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
1171 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1172 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1173 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1174 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1175 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
1176 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1177 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1178 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1179 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1180 // CHECK2-NEXT:    [[R3:%.*]] = alloca i32, align 4
1181 // CHECK2-NEXT:    [[I4:%.*]] = alloca i32, align 4
1182 // CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
1183 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1184 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1185 // CHECK2-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
1186 // CHECK2-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
1187 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
1188 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
1189 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
1190 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
1191 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1192 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1193 // CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1194 // CHECK2-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1195 // CHECK2-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1196 // CHECK2-NEXT:    store i32 0, i32* [[I]], align 4
1197 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1198 // CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1199 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1200 // CHECK2:       omp.precond.then:
1201 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1202 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1203 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
1204 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1205 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1206 // CHECK2-NEXT:    store i32 0, i32* [[R3]], align 4
1207 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1208 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
1209 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1210 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1211 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1212 // CHECK2-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
1213 // CHECK2-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1214 // CHECK2:       cond.true:
1215 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1216 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1217 // CHECK2:       cond.false:
1218 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1219 // CHECK2-NEXT:    br label [[COND_END]]
1220 // CHECK2:       cond.end:
1221 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1222 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1223 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1224 // CHECK2-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
1225 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1226 // CHECK2:       omp.inner.for.cond:
1227 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1228 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1229 // CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1230 // CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1231 // CHECK2:       omp.inner.for.body:
1232 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1233 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
1234 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1235 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
1236 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1237 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4
1238 // CHECK2-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP17]], i32 2)
1239 // CHECK2-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
1240 // CHECK2-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
1241 // CHECK2:       .cancel.exit:
1242 // CHECK2-NEXT:    br label [[CANCEL_EXIT:%.*]]
1243 // CHECK2:       .cancel.continue:
1244 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4
1245 // CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[R3]], align 4
1246 // CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP20]]
1247 // CHECK2-NEXT:    store i32 [[ADD7]], i32* [[R3]], align 4
1248 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1249 // CHECK2:       omp.body.continue:
1250 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1251 // CHECK2:       omp.inner.for.inc:
1252 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1253 // CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
1254 // CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
1255 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
1256 // CHECK2:       omp.inner.for.end:
1257 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1258 // CHECK2:       omp.loop.exit:
1259 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1260 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
1261 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP24]])
1262 // CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1263 // CHECK2-NEXT:    [[TMP26:%.*]] = bitcast i32* [[R3]] to i8*
1264 // CHECK2-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
1265 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1266 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
1267 // CHECK2-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
1268 // 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)
1269 // CHECK2-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1270 // CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1271 // CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1272 // CHECK2-NEXT:    ]
1273 // CHECK2:       .omp.reduction.case1:
1274 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP1]], align 4
1275 // CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[R3]], align 4
1276 // CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
1277 // CHECK2-NEXT:    store i32 [[ADD9]], i32* [[TMP1]], align 4
1278 // CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB7]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
1279 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1280 // CHECK2:       cancel.exit:
1281 // CHECK2-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1282 // CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
1283 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP34]])
1284 // CHECK2-NEXT:    br label [[CANCEL_CONT:%.*]]
1285 // CHECK2:       .omp.reduction.case2:
1286 // CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[R3]], align 4
1287 // CHECK2-NEXT:    [[TMP36:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP35]] monotonic, align 4
1288 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1289 // CHECK2:       .omp.reduction.default:
1290 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
1291 // CHECK2:       omp.precond.end:
1292 // CHECK2-NEXT:    br label [[CANCEL_CONT]]
1293 // CHECK2:       cancel.cont:
1294 // CHECK2-NEXT:    ret void
1295 //
1296 //
1297 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
1298 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
1299 // CHECK2-NEXT:  entry:
1300 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
1301 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
1302 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1303 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1304 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1305 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
1306 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1307 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
1308 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
1309 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1310 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1311 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
1312 // CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1313 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1314 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
1315 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
1316 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1317 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
1318 // CHECK2-NEXT:    ret void
1319 //
1320 //
1321 // CHECK3-LABEL: define {{[^@]+}}@main
1322 // CHECK3-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1323 // CHECK3-NEXT:  entry:
1324 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1325 // CHECK3-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1326 // CHECK3-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
1327 // CHECK3-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
1328 // CHECK3-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
1329 // CHECK3-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
1330 // CHECK3-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
1331 // CHECK3-NEXT:    [[P_LASTITER27:%.*]] = alloca i32, align 4
1332 // CHECK3-NEXT:    [[P_LOWERBOUND28:%.*]] = alloca i32, align 4
1333 // CHECK3-NEXT:    [[P_UPPERBOUND29:%.*]] = alloca i32, align 4
1334 // CHECK3-NEXT:    [[P_STRIDE30:%.*]] = alloca i32, align 4
1335 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1336 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1337 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1338 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
1339 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1340 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1341 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1342 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1343 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1344 // CHECK3-NEXT:    [[I36:%.*]] = alloca i32, align 4
1345 // CHECK3-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
1346 // CHECK3-NEXT:    [[R:%.*]] = alloca i32, align 4
1347 // CHECK3-NEXT:    store i32 0, i32* [[RETVAL]], align 4
1348 // CHECK3-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1349 // CHECK3-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
1350 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
1351 // CHECK3-NEXT:    br label [[OMP_PARALLEL:%.*]]
1352 // CHECK3:       omp_parallel:
1353 // 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]])
1354 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
1355 // CHECK3:       omp.par.outlined.exit:
1356 // CHECK3-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
1357 // CHECK3:       omp.par.exit.split:
1358 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
1359 // CHECK3:       omp_section_loop.preheader:
1360 // CHECK3-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
1361 // CHECK3-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
1362 // CHECK3-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
1363 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1364 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
1365 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
1366 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
1367 // CHECK3-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
1368 // CHECK3-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
1369 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
1370 // CHECK3:       omp_section_loop.header:
1371 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
1372 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
1373 // CHECK3:       omp_section_loop.cond:
1374 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
1375 // CHECK3-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
1376 // CHECK3:       omp_section_loop.body:
1377 // CHECK3-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
1378 // CHECK3-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
1379 // CHECK3-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
1380 // CHECK3-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
1381 // CHECK3-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
1382 // CHECK3-NEXT:    ]
1383 // CHECK3:       omp_section_loop.inc:
1384 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
1385 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
1386 // CHECK3:       omp_section_loop.exit:
1387 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
1388 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1389 // CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
1390 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
1391 // CHECK3:       omp_section_loop.after:
1392 // CHECK3-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
1393 // CHECK3:       omp_sections.end:
1394 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
1395 // CHECK3:       omp_section_loop.preheader13:
1396 // CHECK3-NEXT:    store i32 0, i32* [[P_LOWERBOUND28]], align 4
1397 // CHECK3-NEXT:    store i32 1, i32* [[P_UPPERBOUND29]], align 4
1398 // CHECK3-NEXT:    store i32 1, i32* [[P_STRIDE30]], align 4
1399 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1400 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]], i32 34, i32* [[P_LASTITER27]], i32* [[P_LOWERBOUND28]], i32* [[P_UPPERBOUND29]], i32* [[P_STRIDE30]], i32 1, i32 1)
1401 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND28]], align 4
1402 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND29]], align 4
1403 // CHECK3-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
1404 // CHECK3-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
1405 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
1406 // CHECK3:       omp_section_loop.header14:
1407 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
1408 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
1409 // CHECK3:       omp_section_loop.cond15:
1410 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP10]]
1411 // CHECK3-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
1412 // CHECK3:       omp_section_loop.body16:
1413 // CHECK3-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP7]]
1414 // CHECK3-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
1415 // CHECK3-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
1416 // CHECK3-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC17]] [
1417 // CHECK3-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
1418 // CHECK3-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
1419 // CHECK3-NEXT:    ]
1420 // CHECK3:       omp_section_loop.inc17:
1421 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
1422 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
1423 // CHECK3:       omp_section_loop.exit18:
1424 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
1425 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1426 // CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
1427 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
1428 // CHECK3:       omp_section_loop.after19:
1429 // CHECK3-NEXT:    br label [[OMP_SECTIONS_END33:%.*]]
1430 // CHECK3:       omp_sections.end33:
1431 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1432 // CHECK3-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
1433 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1434 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
1435 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1436 // CHECK3-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
1437 // CHECK3-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
1438 // CHECK3-NEXT:    store i32 0, i32* [[I]], align 4
1439 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1440 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
1441 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1442 // CHECK3:       omp.precond.then:
1443 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1444 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
1445 // CHECK3-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
1446 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1447 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1448 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
1449 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM37]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1450 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1451 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
1452 // CHECK3-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
1453 // CHECK3-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1454 // CHECK3:       cond.true:
1455 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
1456 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1457 // CHECK3:       cond.false:
1458 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1459 // CHECK3-NEXT:    br label [[COND_END]]
1460 // CHECK3:       cond.end:
1461 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
1462 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1463 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1464 // CHECK3-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
1465 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1466 // CHECK3:       omp.inner.for.cond:
1467 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1468 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1469 // CHECK3-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
1470 // CHECK3-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1471 // CHECK3:       omp.inner.for.body:
1472 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1473 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
1474 // CHECK3-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
1475 // CHECK3-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
1476 // CHECK3-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
1477 // CHECK3-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
1478 // CHECK3-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1479 // CHECK3:       omp_if.then:
1480 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
1481 // CHECK3-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
1482 // CHECK3-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
1483 // CHECK3-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
1484 // CHECK3:       .cancel.exit:
1485 // CHECK3-NEXT:    br label [[CANCEL_EXIT:%.*]]
1486 // CHECK3:       omp_section_loop.body.case:
1487 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1488 // CHECK3-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
1489 // CHECK3-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
1490 // CHECK3-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
1491 // CHECK3:       omp_section_loop.body.case.split:
1492 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
1493 // CHECK3:       omp_section_loop.body.case.cncl:
1494 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
1495 // CHECK3:       omp_section_loop.body.case23:
1496 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1497 // CHECK3-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
1498 // CHECK3-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
1499 // CHECK3-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
1500 // CHECK3:       omp_section_loop.body.case23.split:
1501 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
1502 // CHECK3:       omp_section_loop.body.case23.cncl:
1503 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
1504 // CHECK3:       omp_section_loop.body.case25:
1505 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM26:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1506 // CHECK3-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM26]], i32 3)
1507 // CHECK3-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
1508 // CHECK3-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
1509 // CHECK3:       omp_section_loop.body.case25.split:
1510 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
1511 // CHECK3:       omp_section_loop.body.case25.cncl:
1512 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
1513 // CHECK3:       .cancel.continue:
1514 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
1515 // CHECK3:       omp_if.else:
1516 // CHECK3-NEXT:    br label [[OMP_IF_END]]
1517 // CHECK3:       omp_if.end:
1518 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1519 // CHECK3:       omp.body.continue:
1520 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1521 // CHECK3:       omp.inner.for.inc:
1522 // CHECK3-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1523 // CHECK3-NEXT:    [[ADD43:%.*]] = add nsw i32 [[TMP35]], 1
1524 // CHECK3-NEXT:    store i32 [[ADD43]], i32* [[DOTOMP_IV]], align 4
1525 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1526 // CHECK3:       omp.inner.for.end:
1527 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1528 // CHECK3:       omp.loop.exit:
1529 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
1530 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
1531 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
1532 // CHECK3:       cancel.exit:
1533 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
1534 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
1535 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
1536 // CHECK3:       omp.precond.end:
1537 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1538 // CHECK3:       cancel.cont:
1539 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1540 // CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM46]])
1541 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
1542 // CHECK3-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], 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*)*))
1543 // CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
1544 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
1545 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM48:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
1546 // CHECK3-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM48]], i8* [[TMP36]])
1547 // 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*, ...)*))
1548 // 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*, ...)*))
1549 // CHECK3-NEXT:    store i32 0, i32* [[R]], align 4
1550 // 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]])
1551 // CHECK3-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1552 // CHECK3-NEXT:    ret i32 [[TMP40]]
1553 //
1554 //
1555 // CHECK3-LABEL: define {{[^@]+}}@main..omp_par
1556 // CHECK3-SAME: (i32* noalias [[TID_ADDR:%.*]], i32* noalias [[ZERO_ADDR:%.*]], i32* [[ARGC_ADDR:%.*]], i8*** [[ARGV_ADDR:%.*]]) #[[ATTR1:[0-9]+]] {
1557 // CHECK3-NEXT:  omp.par.entry:
1558 // CHECK3-NEXT:    [[TID_ADDR_LOCAL:%.*]] = alloca i32, align 4
1559 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[TID_ADDR]], align 4
1560 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[TID_ADDR_LOCAL]], align 4
1561 // CHECK3-NEXT:    [[TID:%.*]] = load i32, i32* [[TID_ADDR_LOCAL]], align 4
1562 // CHECK3-NEXT:    br label [[OMP_PAR_REGION:%.*]]
1563 // CHECK3:       omp.par.region:
1564 // CHECK3-NEXT:    [[TMP1:%.*]] = load float, float* @flag, align 4
1565 // CHECK3-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP1]], 0.000000e+00
1566 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[TMP13:%.*]], label [[TMP2:%.*]]
1567 // CHECK3:       2:
1568 // CHECK3-NEXT:    br label [[TMP3:%.*]]
1569 // CHECK3:       3:
1570 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1571 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP4]] to i8
1572 // CHECK3-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1573 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 0
1574 // CHECK3-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
1575 // CHECK3-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
1576 // CHECK3-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX3]], align 1
1577 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1578 // CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
1579 // CHECK3-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
1580 // CHECK3-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL5:%.*]]
1581 // CHECK3:       .cncl5:
1582 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
1583 // CHECK3:       .cont:
1584 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1585 // CHECK3-NEXT:    [[TMP10:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1586 // CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
1587 // CHECK3-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
1588 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
1589 // CHECK3-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
1590 // CHECK3-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP12]] to i32
1591 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP9]]
1592 // CHECK3-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
1593 // CHECK3-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
1594 // CHECK3-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
1595 // CHECK3:       omp.par.pre_finalize:
1596 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
1597 // CHECK3:       13:
1598 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1599 // CHECK3-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
1600 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 0
1601 // CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
1602 // CHECK3:       .cncl:
1603 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1604 // CHECK3-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
1605 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
1606 // CHECK3:       .split:
1607 // CHECK3-NEXT:    br label [[TMP3]]
1608 // CHECK3:       omp.par.outlined.exit.exitStub:
1609 // CHECK3-NEXT:    ret void
1610 //
1611 //
1612 // CHECK3-LABEL: define {{[^@]+}}@.omp_task_entry.
1613 // CHECK3-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
1614 // CHECK3-NEXT:  entry:
1615 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
1616 // CHECK3-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
1617 // CHECK3-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
1618 // CHECK3-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
1619 // CHECK3-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
1620 // CHECK3-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
1621 // CHECK3-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
1622 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
1623 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
1624 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
1625 // CHECK3-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
1626 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
1627 // CHECK3-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
1628 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
1629 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
1630 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
1631 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1632 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
1633 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
1634 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META5:![0-9]+]])
1635 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
1636 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
1637 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
1638 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
1639 // CHECK3-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !14
1640 // CHECK3-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !14
1641 // CHECK3-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !14
1642 // CHECK3-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !14
1643 // CHECK3-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
1644 // CHECK3-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
1645 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM_I:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]]) #[[ATTR2:[0-9]+]]
1646 // CHECK3-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM_I]], i32 4) #[[ATTR2]]
1647 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1648 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
1649 // CHECK3:       .cancel.exit.i:
1650 // CHECK3-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
1651 // CHECK3-NEXT:    br label [[DOTOMP_OUTLINED__EXIT:%.*]]
1652 // CHECK3:       .cancel.continue.i:
1653 // CHECK3-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
1654 // CHECK3-NEXT:    br label [[DOTOMP_OUTLINED__EXIT]]
1655 // CHECK3:       .omp_outlined..exit:
1656 // CHECK3-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
1657 // CHECK3-NEXT:    ret i32 0
1658 //
1659 //
1660 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
1661 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5:[0-9]+]] {
1662 // CHECK3-NEXT:  entry:
1663 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1664 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1665 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
1666 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
1667 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
1668 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
1669 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
1670 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1671 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1672 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1673 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1674 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
1675 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
1676 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB17:[0-9]+]])
1677 // 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)
1678 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1679 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 0
1680 // CHECK3-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 0
1681 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
1682 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1683 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1684 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1685 // CHECK3:       omp.inner.for.cond:
1686 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1687 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1688 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
1689 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1690 // CHECK3:       omp.inner.for.body:
1691 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1692 // CHECK3-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
1693 // CHECK3-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
1694 // CHECK3-NEXT:    ]
1695 // CHECK3:       .omp.sections.case:
1696 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1697 // CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
1698 // CHECK3-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
1699 // CHECK3-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
1700 // CHECK3:       .omp.sections.case.split:
1701 // CHECK3-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1702 // CHECK3:       .omp.sections.case.cncl:
1703 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
1704 // CHECK3:       .omp.sections.exit:
1705 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1706 // CHECK3:       omp.inner.for.inc:
1707 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1708 // CHECK3-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
1709 // CHECK3-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1710 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1711 // CHECK3:       omp.inner.for.end:
1712 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19:[0-9]+]])
1713 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
1714 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1715 // CHECK3:       cancel.cont:
1716 // CHECK3-NEXT:    ret void
1717 // CHECK3:       cancel.exit:
1718 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19]])
1719 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
1720 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1721 //
1722 //
1723 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
1724 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5]] {
1725 // CHECK3-NEXT:  entry:
1726 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1727 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1728 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
1729 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
1730 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
1731 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
1732 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
1733 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1734 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1735 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1736 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1737 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
1738 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
1739 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB21:[0-9]+]])
1740 // 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)
1741 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1742 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 1
1743 // CHECK3-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 1
1744 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
1745 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1746 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1747 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1748 // CHECK3:       omp.inner.for.cond:
1749 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1750 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1751 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
1752 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1753 // CHECK3:       omp.inner.for.body:
1754 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1755 // CHECK3-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
1756 // CHECK3-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
1757 // CHECK3-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE2:%.*]]
1758 // CHECK3-NEXT:    ]
1759 // CHECK3:       .omp.sections.case:
1760 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1761 // CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
1762 // CHECK3-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
1763 // CHECK3-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
1764 // CHECK3:       .omp.sections.case.split:
1765 // CHECK3-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1766 // CHECK3:       .omp.sections.case.cncl:
1767 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
1768 // CHECK3:       .omp.sections.case2:
1769 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1770 // CHECK3-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
1771 // CHECK3-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
1772 // CHECK3-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
1773 // CHECK3:       .omp.sections.case2.split:
1774 // CHECK3-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1775 // CHECK3:       .omp.sections.case2.cncl:
1776 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END]]
1777 // CHECK3:       .omp.sections.exit:
1778 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1779 // CHECK3:       omp.inner.for.inc:
1780 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1781 // CHECK3-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
1782 // CHECK3-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1783 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1784 // CHECK3:       omp.inner.for.end:
1785 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM5:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23:[0-9]+]])
1786 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM5]])
1787 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1788 // CHECK3:       cancel.cont:
1789 // CHECK3-NEXT:    ret void
1790 // CHECK3:       cancel.exit:
1791 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23]])
1792 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
1793 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1794 //
1795 //
1796 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
1797 // 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]] {
1798 // CHECK3-NEXT:  entry:
1799 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1800 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1801 // CHECK3-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
1802 // CHECK3-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
1803 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1804 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1805 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1806 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1807 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1808 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1809 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1810 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1811 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1812 // CHECK3-NEXT:    [[R3:%.*]] = alloca i32, align 4
1813 // CHECK3-NEXT:    [[I4:%.*]] = alloca i32, align 4
1814 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
1815 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1816 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1817 // CHECK3-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
1818 // CHECK3-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
1819 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
1820 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
1821 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
1822 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
1823 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1824 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1825 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1826 // CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1827 // CHECK3-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1828 // CHECK3-NEXT:    store i32 0, i32* [[I]], align 4
1829 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1830 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1831 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1832 // CHECK3:       omp.precond.then:
1833 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1834 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1835 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
1836 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1837 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1838 // CHECK3-NEXT:    store i32 0, i32* [[R3]], align 4
1839 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB25:[0-9]+]])
1840 // 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)
1841 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1842 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1843 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
1844 // CHECK3-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1845 // CHECK3:       cond.true:
1846 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1847 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1848 // CHECK3:       cond.false:
1849 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1850 // CHECK3-NEXT:    br label [[COND_END]]
1851 // CHECK3:       cond.end:
1852 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
1853 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1854 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1855 // CHECK3-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
1856 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1857 // CHECK3:       omp.inner.for.cond:
1858 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1859 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1860 // CHECK3-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
1861 // CHECK3-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1862 // CHECK3:       omp.inner.for.body:
1863 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1864 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1
1865 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1866 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
1867 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM7:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB27:[0-9]+]])
1868 // CHECK3-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM7]], i32 2)
1869 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
1870 // CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
1871 // CHECK3:       .cancel.exit:
1872 // CHECK3-NEXT:    br label [[CANCEL_EXIT:%.*]]
1873 // CHECK3:       .cancel.continue:
1874 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I4]], align 4
1875 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[R3]], align 4
1876 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP16]]
1877 // CHECK3-NEXT:    store i32 [[ADD8]], i32* [[R3]], align 4
1878 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1879 // CHECK3:       omp.body.continue:
1880 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1881 // CHECK3:       omp.inner.for.inc:
1882 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1883 // CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], 1
1884 // CHECK3-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
1885 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1886 // CHECK3:       omp.inner.for.end:
1887 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1888 // CHECK3:       omp.loop.exit:
1889 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29:[0-9]+]])
1890 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
1891 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1892 // CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i32* [[R3]] to i8*
1893 // CHECK3-NEXT:    store i8* [[TMP20]], i8** [[TMP19]], align 8
1894 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
1895 // CHECK3-NEXT:    [[TMP21:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
1896 // 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)
1897 // CHECK3-NEXT:    switch i32 [[TMP22]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1898 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1899 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1900 // CHECK3-NEXT:    ]
1901 // CHECK3:       .omp.reduction.case1:
1902 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP1]], align 4
1903 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[R3]], align 4
1904 // CHECK3-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
1905 // CHECK3-NEXT:    store i32 [[ADD13]], i32* [[TMP1]], align 4
1906 // 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)
1907 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1908 // CHECK3:       cancel.exit:
1909 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
1910 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
1911 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
1912 // CHECK3:       .omp.reduction.case2:
1913 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[R3]], align 4
1914 // CHECK3-NEXT:    [[TMP26:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP25]] monotonic, align 4
1915 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1916 // CHECK3:       .omp.reduction.default:
1917 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
1918 // CHECK3:       omp.precond.end:
1919 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1920 // CHECK3:       cancel.cont:
1921 // CHECK3-NEXT:    ret void
1922 //
1923 //
1924 // CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
1925 // CHECK3-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
1926 // CHECK3-NEXT:  entry:
1927 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
1928 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
1929 // CHECK3-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1930 // CHECK3-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1931 // CHECK3-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1932 // CHECK3-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
1933 // CHECK3-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1934 // CHECK3-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
1935 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
1936 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1937 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1938 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
1939 // CHECK3-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1940 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1941 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
1942 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
1943 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1944 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
1945 // CHECK3-NEXT:    ret void
1946 //
1947 //
1948 // CHECK4-LABEL: define {{[^@]+}}@main
1949 // CHECK4-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1950 // CHECK4-NEXT:  entry:
1951 // CHECK4-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1952 // CHECK4-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1953 // CHECK4-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
1954 // CHECK4-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
1955 // CHECK4-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
1956 // CHECK4-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
1957 // CHECK4-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
1958 // CHECK4-NEXT:    [[P_LASTITER27:%.*]] = alloca i32, align 4
1959 // CHECK4-NEXT:    [[P_LOWERBOUND28:%.*]] = alloca i32, align 4
1960 // CHECK4-NEXT:    [[P_UPPERBOUND29:%.*]] = alloca i32, align 4
1961 // CHECK4-NEXT:    [[P_STRIDE30:%.*]] = alloca i32, align 4
1962 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1963 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1964 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1965 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
1966 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
1967 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1968 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1969 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1970 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1971 // CHECK4-NEXT:    [[I36:%.*]] = alloca i32, align 4
1972 // CHECK4-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
1973 // CHECK4-NEXT:    [[R:%.*]] = alloca i32, align 4
1974 // CHECK4-NEXT:    store i32 0, i32* [[RETVAL]], align 4
1975 // CHECK4-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1976 // CHECK4-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
1977 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
1978 // CHECK4-NEXT:    br label [[OMP_PARALLEL:%.*]]
1979 // CHECK4:       omp_parallel:
1980 // 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]])
1981 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
1982 // CHECK4:       omp.par.outlined.exit:
1983 // CHECK4-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
1984 // CHECK4:       omp.par.exit.split:
1985 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
1986 // CHECK4:       omp_section_loop.preheader:
1987 // CHECK4-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
1988 // CHECK4-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
1989 // CHECK4-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
1990 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1991 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
1992 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
1993 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
1994 // CHECK4-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
1995 // CHECK4-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
1996 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
1997 // CHECK4:       omp_section_loop.header:
1998 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
1999 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
2000 // CHECK4:       omp_section_loop.cond:
2001 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
2002 // CHECK4-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
2003 // CHECK4:       omp_section_loop.body:
2004 // CHECK4-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
2005 // CHECK4-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
2006 // CHECK4-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
2007 // CHECK4-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
2008 // CHECK4-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
2009 // CHECK4-NEXT:    ]
2010 // CHECK4:       omp_section_loop.inc:
2011 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
2012 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
2013 // CHECK4:       omp_section_loop.exit:
2014 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
2015 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2016 // CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
2017 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
2018 // CHECK4:       omp_section_loop.after:
2019 // CHECK4-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
2020 // CHECK4:       omp_sections.end:
2021 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
2022 // CHECK4:       omp_section_loop.preheader13:
2023 // CHECK4-NEXT:    store i32 0, i32* [[P_LOWERBOUND28]], align 4
2024 // CHECK4-NEXT:    store i32 1, i32* [[P_UPPERBOUND29]], align 4
2025 // CHECK4-NEXT:    store i32 1, i32* [[P_STRIDE30]], align 4
2026 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2027 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]], i32 34, i32* [[P_LASTITER27]], i32* [[P_LOWERBOUND28]], i32* [[P_UPPERBOUND29]], i32* [[P_STRIDE30]], i32 1, i32 1)
2028 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND28]], align 4
2029 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND29]], align 4
2030 // CHECK4-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
2031 // CHECK4-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
2032 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
2033 // CHECK4:       omp_section_loop.header14:
2034 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
2035 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
2036 // CHECK4:       omp_section_loop.cond15:
2037 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP10]]
2038 // CHECK4-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
2039 // CHECK4:       omp_section_loop.body16:
2040 // CHECK4-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP7]]
2041 // CHECK4-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
2042 // CHECK4-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
2043 // CHECK4-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC17]] [
2044 // CHECK4-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
2045 // CHECK4-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
2046 // CHECK4-NEXT:    ]
2047 // CHECK4:       omp_section_loop.inc17:
2048 // CHECK4-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
2049 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
2050 // CHECK4:       omp_section_loop.exit18:
2051 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
2052 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2053 // CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
2054 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
2055 // CHECK4:       omp_section_loop.after19:
2056 // CHECK4-NEXT:    br label [[OMP_SECTIONS_END33:%.*]]
2057 // CHECK4:       omp_sections.end33:
2058 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2059 // CHECK4-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
2060 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2061 // CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
2062 // CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2063 // CHECK4-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
2064 // CHECK4-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
2065 // CHECK4-NEXT:    store i32 0, i32* [[I]], align 4
2066 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2067 // CHECK4-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
2068 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2069 // CHECK4:       omp.precond.then:
2070 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2071 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
2072 // CHECK4-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
2073 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2074 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2075 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
2076 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM37]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2077 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2078 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
2079 // CHECK4-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
2080 // CHECK4-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2081 // CHECK4:       cond.true:
2082 // CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
2083 // CHECK4-NEXT:    br label [[COND_END:%.*]]
2084 // CHECK4:       cond.false:
2085 // CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2086 // CHECK4-NEXT:    br label [[COND_END]]
2087 // CHECK4:       cond.end:
2088 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
2089 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2090 // CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2091 // CHECK4-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
2092 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2093 // CHECK4:       omp.inner.for.cond:
2094 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2095 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2096 // CHECK4-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
2097 // CHECK4-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2098 // CHECK4:       omp.inner.for.body:
2099 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2100 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
2101 // CHECK4-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
2102 // CHECK4-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
2103 // CHECK4-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
2104 // CHECK4-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
2105 // CHECK4-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2106 // CHECK4:       omp_if.then:
2107 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
2108 // CHECK4-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
2109 // CHECK4-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
2110 // CHECK4-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
2111 // CHECK4:       .cancel.exit:
2112 // CHECK4-NEXT:    br label [[CANCEL_EXIT:%.*]]
2113 // CHECK4:       omp_section_loop.body.case:
2114 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2115 // CHECK4-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
2116 // CHECK4-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
2117 // CHECK4-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
2118 // CHECK4:       omp_section_loop.body.case.split:
2119 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
2120 // CHECK4:       omp_section_loop.body.case.cncl:
2121 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
2122 // CHECK4:       omp_section_loop.body.case23:
2123 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2124 // CHECK4-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
2125 // CHECK4-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
2126 // CHECK4-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
2127 // CHECK4:       omp_section_loop.body.case23.split:
2128 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
2129 // CHECK4:       omp_section_loop.body.case23.cncl:
2130 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
2131 // CHECK4:       omp_section_loop.body.case25:
2132 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM26:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2133 // CHECK4-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM26]], i32 3)
2134 // CHECK4-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
2135 // CHECK4-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
2136 // CHECK4:       omp_section_loop.body.case25.split:
2137 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
2138 // CHECK4:       omp_section_loop.body.case25.cncl:
2139 // CHECK4-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
2140 // CHECK4:       .cancel.continue:
2141 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
2142 // CHECK4:       omp_if.else:
2143 // CHECK4-NEXT:    br label [[OMP_IF_END]]
2144 // CHECK4:       omp_if.end:
2145 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2146 // CHECK4:       omp.body.continue:
2147 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2148 // CHECK4:       omp.inner.for.inc:
2149 // CHECK4-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2150 // CHECK4-NEXT:    [[ADD43:%.*]] = add nsw i32 [[TMP35]], 1
2151 // CHECK4-NEXT:    store i32 [[ADD43]], i32* [[DOTOMP_IV]], align 4
2152 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
2153 // CHECK4:       omp.inner.for.end:
2154 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2155 // CHECK4:       omp.loop.exit:
2156 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
2157 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
2158 // CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
2159 // CHECK4:       cancel.exit:
2160 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
2161 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
2162 // CHECK4-NEXT:    br label [[CANCEL_CONT:%.*]]
2163 // CHECK4:       omp.precond.end:
2164 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
2165 // CHECK4:       cancel.cont:
2166 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2167 // CHECK4-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM46]])
2168 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
2169 // CHECK4-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], 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*)*))
2170 // CHECK4-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
2171 // CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
2172 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM48:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
2173 // CHECK4-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM48]], i8* [[TMP36]])
2174 // 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*, ...)*))
2175 // 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*, ...)*))
2176 // CHECK4-NEXT:    store i32 0, i32* [[R]], align 4
2177 // 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]])
2178 // CHECK4-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2179 // CHECK4-NEXT:    ret i32 [[TMP40]]
2180 //
2181 //
2182 // CHECK4-LABEL: define {{[^@]+}}@main..omp_par
2183 // CHECK4-SAME: (i32* noalias [[TID_ADDR:%.*]], i32* noalias [[ZERO_ADDR:%.*]], i32* [[ARGC_ADDR:%.*]], i8*** [[ARGV_ADDR:%.*]]) #[[ATTR1:[0-9]+]] {
2184 // CHECK4-NEXT:  omp.par.entry:
2185 // CHECK4-NEXT:    [[TID_ADDR_LOCAL:%.*]] = alloca i32, align 4
2186 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[TID_ADDR]], align 4
2187 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[TID_ADDR_LOCAL]], align 4
2188 // CHECK4-NEXT:    [[TID:%.*]] = load i32, i32* [[TID_ADDR_LOCAL]], align 4
2189 // CHECK4-NEXT:    br label [[OMP_PAR_REGION:%.*]]
2190 // CHECK4:       omp.par.region:
2191 // CHECK4-NEXT:    [[TMP1:%.*]] = load float, float* @flag, align 4
2192 // CHECK4-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP1]], 0.000000e+00
2193 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[TMP13:%.*]], label [[TMP2:%.*]]
2194 // CHECK4:       2:
2195 // CHECK4-NEXT:    br label [[TMP3:%.*]]
2196 // CHECK4:       3:
2197 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2198 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP4]] to i8
2199 // CHECK4-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
2200 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 0
2201 // CHECK4-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
2202 // CHECK4-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
2203 // CHECK4-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX3]], align 1
2204 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2205 // CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
2206 // CHECK4-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
2207 // CHECK4-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL5:%.*]]
2208 // CHECK4:       .cncl5:
2209 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
2210 // CHECK4:       .cont:
2211 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2212 // CHECK4-NEXT:    [[TMP10:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
2213 // CHECK4-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
2214 // CHECK4-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
2215 // CHECK4-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
2216 // CHECK4-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
2217 // CHECK4-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP12]] to i32
2218 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP9]]
2219 // CHECK4-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
2220 // CHECK4-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
2221 // CHECK4-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
2222 // CHECK4:       omp.par.pre_finalize:
2223 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
2224 // CHECK4:       13:
2225 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2226 // CHECK4-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
2227 // CHECK4-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 0
2228 // CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
2229 // CHECK4:       .cncl:
2230 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2231 // CHECK4-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
2232 // CHECK4-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
2233 // CHECK4:       .split:
2234 // CHECK4-NEXT:    br label [[TMP3]]
2235 // CHECK4:       omp.par.outlined.exit.exitStub:
2236 // CHECK4-NEXT:    ret void
2237 //
2238 //
2239 // CHECK4-LABEL: define {{[^@]+}}@.omp_task_entry.
2240 // CHECK4-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
2241 // CHECK4-NEXT:  entry:
2242 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
2243 // CHECK4-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
2244 // CHECK4-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
2245 // CHECK4-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
2246 // CHECK4-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
2247 // CHECK4-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
2248 // CHECK4-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
2249 // CHECK4-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
2250 // CHECK4-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
2251 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
2252 // CHECK4-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
2253 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
2254 // CHECK4-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
2255 // CHECK4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
2256 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
2257 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
2258 // CHECK4-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2259 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
2260 // CHECK4-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
2261 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META5:![0-9]+]])
2262 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
2263 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
2264 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
2265 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
2266 // CHECK4-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !14
2267 // CHECK4-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !14
2268 // CHECK4-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !14
2269 // CHECK4-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !14
2270 // CHECK4-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
2271 // CHECK4-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
2272 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM_I:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]]) #[[ATTR2:[0-9]+]]
2273 // CHECK4-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM_I]], i32 4) #[[ATTR2]]
2274 // CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2275 // CHECK4-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
2276 // CHECK4:       .cancel.exit.i:
2277 // CHECK4-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
2278 // CHECK4-NEXT:    br label [[DOTOMP_OUTLINED__EXIT:%.*]]
2279 // CHECK4:       .cancel.continue.i:
2280 // CHECK4-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
2281 // CHECK4-NEXT:    br label [[DOTOMP_OUTLINED__EXIT]]
2282 // CHECK4:       .omp_outlined..exit:
2283 // CHECK4-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
2284 // CHECK4-NEXT:    ret i32 0
2285 //
2286 //
2287 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
2288 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5:[0-9]+]] {
2289 // CHECK4-NEXT:  entry:
2290 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2291 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2292 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
2293 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
2294 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
2295 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
2296 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
2297 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2298 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2299 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2300 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2301 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
2302 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
2303 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB17:[0-9]+]])
2304 // 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)
2305 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2306 // CHECK4-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 0
2307 // CHECK4-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 0
2308 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
2309 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2310 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2311 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2312 // CHECK4:       omp.inner.for.cond:
2313 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2314 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2315 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
2316 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2317 // CHECK4:       omp.inner.for.body:
2318 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2319 // CHECK4-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
2320 // CHECK4-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
2321 // CHECK4-NEXT:    ]
2322 // CHECK4:       .omp.sections.case:
2323 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2324 // CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
2325 // CHECK4-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
2326 // CHECK4-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
2327 // CHECK4:       .omp.sections.case.split:
2328 // CHECK4-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
2329 // CHECK4:       .omp.sections.case.cncl:
2330 // CHECK4-NEXT:    br label [[CANCEL_CONT:%.*]]
2331 // CHECK4:       .omp.sections.exit:
2332 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2333 // CHECK4:       omp.inner.for.inc:
2334 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2335 // CHECK4-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
2336 // CHECK4-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2337 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
2338 // CHECK4:       omp.inner.for.end:
2339 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19:[0-9]+]])
2340 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
2341 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
2342 // CHECK4:       cancel.cont:
2343 // CHECK4-NEXT:    ret void
2344 // CHECK4:       cancel.exit:
2345 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19]])
2346 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
2347 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
2348 //
2349 //
2350 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
2351 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5]] {
2352 // CHECK4-NEXT:  entry:
2353 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2354 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2355 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
2356 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
2357 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
2358 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
2359 // CHECK4-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
2360 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2361 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2362 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2363 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2364 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
2365 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
2366 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB21:[0-9]+]])
2367 // 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)
2368 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2369 // CHECK4-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 1
2370 // CHECK4-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 1
2371 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
2372 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2373 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2374 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2375 // CHECK4:       omp.inner.for.cond:
2376 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2377 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2378 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
2379 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2380 // CHECK4:       omp.inner.for.body:
2381 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2382 // CHECK4-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
2383 // CHECK4-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
2384 // CHECK4-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE2:%.*]]
2385 // CHECK4-NEXT:    ]
2386 // CHECK4:       .omp.sections.case:
2387 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2388 // CHECK4-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
2389 // CHECK4-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
2390 // CHECK4-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
2391 // CHECK4:       .omp.sections.case.split:
2392 // CHECK4-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
2393 // CHECK4:       .omp.sections.case.cncl:
2394 // CHECK4-NEXT:    br label [[CANCEL_CONT:%.*]]
2395 // CHECK4:       .omp.sections.case2:
2396 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2397 // CHECK4-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
2398 // CHECK4-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
2399 // CHECK4-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
2400 // CHECK4:       .omp.sections.case2.split:
2401 // CHECK4-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
2402 // CHECK4:       .omp.sections.case2.cncl:
2403 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_END]]
2404 // CHECK4:       .omp.sections.exit:
2405 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2406 // CHECK4:       omp.inner.for.inc:
2407 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2408 // CHECK4-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
2409 // CHECK4-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2410 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
2411 // CHECK4:       omp.inner.for.end:
2412 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM5:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23:[0-9]+]])
2413 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM5]])
2414 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
2415 // CHECK4:       cancel.cont:
2416 // CHECK4-NEXT:    ret void
2417 // CHECK4:       cancel.exit:
2418 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23]])
2419 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
2420 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
2421 //
2422 //
2423 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3
2424 // 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]] {
2425 // CHECK4-NEXT:  entry:
2426 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2427 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2428 // CHECK4-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
2429 // CHECK4-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
2430 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2431 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2432 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2433 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2434 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
2435 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2436 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2437 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2438 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2439 // CHECK4-NEXT:    [[R3:%.*]] = alloca i32, align 4
2440 // CHECK4-NEXT:    [[I4:%.*]] = alloca i32, align 4
2441 // CHECK4-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
2442 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2443 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2444 // CHECK4-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
2445 // CHECK4-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
2446 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
2447 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
2448 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
2449 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
2450 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2451 // CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
2452 // CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2453 // CHECK4-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2454 // CHECK4-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2455 // CHECK4-NEXT:    store i32 0, i32* [[I]], align 4
2456 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2457 // CHECK4-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
2458 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2459 // CHECK4:       omp.precond.then:
2460 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2461 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2462 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
2463 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2464 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2465 // CHECK4-NEXT:    store i32 0, i32* [[R3]], align 4
2466 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB25:[0-9]+]])
2467 // 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)
2468 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2469 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2470 // CHECK4-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
2471 // CHECK4-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2472 // CHECK4:       cond.true:
2473 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2474 // CHECK4-NEXT:    br label [[COND_END:%.*]]
2475 // CHECK4:       cond.false:
2476 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2477 // CHECK4-NEXT:    br label [[COND_END]]
2478 // CHECK4:       cond.end:
2479 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
2480 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2481 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2482 // CHECK4-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
2483 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2484 // CHECK4:       omp.inner.for.cond:
2485 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2486 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2487 // CHECK4-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
2488 // CHECK4-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2489 // CHECK4:       omp.inner.for.body:
2490 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2491 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1
2492 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2493 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
2494 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM7:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB27:[0-9]+]])
2495 // CHECK4-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM7]], i32 2)
2496 // CHECK4-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
2497 // CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
2498 // CHECK4:       .cancel.exit:
2499 // CHECK4-NEXT:    br label [[CANCEL_EXIT:%.*]]
2500 // CHECK4:       .cancel.continue:
2501 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I4]], align 4
2502 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[R3]], align 4
2503 // CHECK4-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP16]]
2504 // CHECK4-NEXT:    store i32 [[ADD8]], i32* [[R3]], align 4
2505 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2506 // CHECK4:       omp.body.continue:
2507 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2508 // CHECK4:       omp.inner.for.inc:
2509 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2510 // CHECK4-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], 1
2511 // CHECK4-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
2512 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
2513 // CHECK4:       omp.inner.for.end:
2514 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2515 // CHECK4:       omp.loop.exit:
2516 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29:[0-9]+]])
2517 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
2518 // CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
2519 // CHECK4-NEXT:    [[TMP20:%.*]] = bitcast i32* [[R3]] to i8*
2520 // CHECK4-NEXT:    store i8* [[TMP20]], i8** [[TMP19]], align 8
2521 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
2522 // CHECK4-NEXT:    [[TMP21:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
2523 // 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)
2524 // CHECK4-NEXT:    switch i32 [[TMP22]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2525 // CHECK4-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2526 // CHECK4-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2527 // CHECK4-NEXT:    ]
2528 // CHECK4:       .omp.reduction.case1:
2529 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP1]], align 4
2530 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[R3]], align 4
2531 // CHECK4-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
2532 // CHECK4-NEXT:    store i32 [[ADD13]], i32* [[TMP1]], align 4
2533 // 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)
2534 // CHECK4-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
2535 // CHECK4:       cancel.exit:
2536 // CHECK4-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
2537 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
2538 // CHECK4-NEXT:    br label [[CANCEL_CONT:%.*]]
2539 // CHECK4:       .omp.reduction.case2:
2540 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[R3]], align 4
2541 // CHECK4-NEXT:    [[TMP26:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP25]] monotonic, align 4
2542 // CHECK4-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
2543 // CHECK4:       .omp.reduction.default:
2544 // CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
2545 // CHECK4:       omp.precond.end:
2546 // CHECK4-NEXT:    br label [[CANCEL_CONT]]
2547 // CHECK4:       cancel.cont:
2548 // CHECK4-NEXT:    ret void
2549 //
2550 //
2551 // CHECK4-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
2552 // CHECK4-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
2553 // CHECK4-NEXT:  entry:
2554 // CHECK4-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
2555 // CHECK4-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
2556 // CHECK4-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
2557 // CHECK4-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
2558 // CHECK4-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
2559 // CHECK4-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
2560 // CHECK4-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
2561 // CHECK4-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
2562 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
2563 // CHECK4-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2564 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
2565 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
2566 // CHECK4-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
2567 // CHECK4-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
2568 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
2569 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
2570 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2571 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
2572 // CHECK4-NEXT:    ret void
2573 //
2574 //
2575 // CHECK7-LABEL: define {{[^@]+}}@main
2576 // CHECK7-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
2577 // CHECK7-NEXT:  entry:
2578 // CHECK7-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
2579 // CHECK7-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
2580 // CHECK7-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
2581 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
2582 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
2583 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
2584 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
2585 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
2586 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_LB_1:%.*]] = alloca i32, align 4
2587 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_UB_2:%.*]] = alloca i32, align 4
2588 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_ST_3:%.*]] = alloca i32, align 4
2589 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IL_4:%.*]] = alloca i32, align 4
2590 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IV_5:%.*]] = alloca i32, align 4
2591 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2592 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2593 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2594 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_21:%.*]] = alloca i32, align 4
2595 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
2596 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2597 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2598 // CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2599 // CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2600 // CHECK7-NEXT:    [[I24:%.*]] = alloca i32, align 4
2601 // CHECK7-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
2602 // CHECK7-NEXT:    [[R:%.*]] = alloca i32, align 4
2603 // CHECK7-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
2604 // CHECK7-NEXT:    store i32 0, i32* [[RETVAL]], align 4
2605 // CHECK7-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
2606 // CHECK7-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
2607 // 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]])
2608 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2609 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2610 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
2611 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
2612 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[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)
2613 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2614 // CHECK7-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
2615 // CHECK7-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
2616 // CHECK7-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
2617 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2618 // CHECK7-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2619 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2620 // CHECK7:       omp.inner.for.cond:
2621 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2622 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2623 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2624 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2625 // CHECK7:       omp.inner.for.body:
2626 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2627 // CHECK7-NEXT:    switch i32 [[TMP7]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
2628 // CHECK7-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
2629 // CHECK7-NEXT:    ]
2630 // CHECK7:       .omp.sections.case:
2631 // CHECK7-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
2632 // CHECK7-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
2633 // CHECK7-NEXT:    br i1 [[TMP9]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
2634 // CHECK7:       .cancel.exit:
2635 // CHECK7-NEXT:    br label [[CANCEL_EXIT:%.*]]
2636 // CHECK7:       .cancel.continue:
2637 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
2638 // CHECK7:       .omp.sections.exit:
2639 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2640 // CHECK7:       omp.inner.for.inc:
2641 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2642 // CHECK7-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP10]], 1
2643 // CHECK7-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2644 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
2645 // CHECK7:       omp.inner.for.end:
2646 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
2647 // CHECK7-NEXT:    br label [[CANCEL_CONT:%.*]]
2648 // CHECK7:       cancel.cont:
2649 // CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]])
2650 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
2651 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
2652 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_3]], align 4
2653 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_4]], align 4
2654 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], 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)
2655 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
2656 // CHECK7-NEXT:    [[TMP12:%.*]] = icmp slt i32 [[TMP11]], 1
2657 // CHECK7-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 1
2658 // CHECK7-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_SECTIONS_UB_2]], align 4
2659 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
2660 // CHECK7-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
2661 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND6:%.*]]
2662 // CHECK7:       omp.inner.for.cond6:
2663 // CHECK7-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
2664 // CHECK7-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
2665 // CHECK7-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
2666 // CHECK7-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY8:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
2667 // CHECK7:       omp.inner.for.body8:
2668 // CHECK7-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
2669 // CHECK7-NEXT:    switch i32 [[TMP17]], label [[DOTOMP_SECTIONS_EXIT15:%.*]] [
2670 // CHECK7-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE9:%.*]]
2671 // CHECK7-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE12:%.*]]
2672 // CHECK7-NEXT:    ]
2673 // CHECK7:       .omp.sections.case9:
2674 // CHECK7-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
2675 // CHECK7-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
2676 // CHECK7-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT10:%.*]], label [[DOTCANCEL_CONTINUE11:%.*]]
2677 // CHECK7:       .cancel.exit10:
2678 // CHECK7-NEXT:    br label [[CANCEL_EXIT19:%.*]]
2679 // CHECK7:       cancel.exit:
2680 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
2681 // CHECK7-NEXT:    br label [[CANCEL_CONT]]
2682 // CHECK7:       .cancel.continue11:
2683 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
2684 // CHECK7:       .omp.sections.case12:
2685 // CHECK7-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
2686 // CHECK7-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
2687 // CHECK7-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT13:%.*]], label [[DOTCANCEL_CONTINUE14:%.*]]
2688 // CHECK7:       .cancel.exit13:
2689 // CHECK7-NEXT:    br label [[CANCEL_EXIT19]]
2690 // CHECK7:       .cancel.continue14:
2691 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
2692 // CHECK7:       .omp.sections.exit15:
2693 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
2694 // CHECK7:       omp.inner.for.inc16:
2695 // CHECK7-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
2696 // CHECK7-NEXT:    [[INC17:%.*]] = add nsw i32 [[TMP22]], 1
2697 // CHECK7-NEXT:    store i32 [[INC17]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
2698 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND6]]
2699 // CHECK7:       omp.inner.for.end18:
2700 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
2701 // CHECK7-NEXT:    br label [[CANCEL_CONT20:%.*]]
2702 // CHECK7:       cancel.cont20:
2703 // CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
2704 // CHECK7-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2705 // CHECK7-NEXT:    store i32 [[TMP23]], i32* [[DOTCAPTURE_EXPR_]], align 4
2706 // CHECK7-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2707 // CHECK7-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
2708 // CHECK7-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2709 // CHECK7-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[DIV]], 1
2710 // CHECK7-NEXT:    store i32 [[SUB22]], i32* [[DOTCAPTURE_EXPR_21]], align 4
2711 // CHECK7-NEXT:    store i32 0, i32* [[I]], align 4
2712 // CHECK7-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2713 // CHECK7-NEXT:    [[CMP23:%.*]] = icmp slt i32 0, [[TMP25]]
2714 // CHECK7-NEXT:    br i1 [[CMP23]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2715 // CHECK7:       omp.precond.then:
2716 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2717 // CHECK7-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
2718 // CHECK7-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_UB]], align 4
2719 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2720 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2721 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2722 // CHECK7-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2723 // CHECK7-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
2724 // CHECK7-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
2725 // CHECK7-NEXT:    br i1 [[CMP25]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2726 // CHECK7:       cond.true:
2727 // CHECK7-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
2728 // CHECK7-NEXT:    br label [[COND_END:%.*]]
2729 // CHECK7:       cond.false:
2730 // CHECK7-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2731 // CHECK7-NEXT:    br label [[COND_END]]
2732 // CHECK7:       cond.end:
2733 // CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE]] ], [ [[TMP30]], [[COND_FALSE]] ]
2734 // CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2735 // CHECK7-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2736 // CHECK7-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
2737 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND26:%.*]]
2738 // CHECK7:       omp.inner.for.cond26:
2739 // CHECK7-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2740 // CHECK7-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2741 // CHECK7-NEXT:    [[CMP27:%.*]] = icmp sle i32 [[TMP32]], [[TMP33]]
2742 // CHECK7-NEXT:    br i1 [[CMP27]], label [[OMP_INNER_FOR_BODY28:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
2743 // CHECK7:       omp.inner.for.body28:
2744 // CHECK7-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2745 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP34]], 1
2746 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2747 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[I24]], align 4
2748 // CHECK7-NEXT:    [[TMP35:%.*]] = load float, float* @flag, align 4
2749 // CHECK7-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP35]], 0.000000e+00
2750 // CHECK7-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2751 // CHECK7:       omp_if.then:
2752 // CHECK7-NEXT:    [[TMP36:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
2753 // CHECK7-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
2754 // CHECK7-NEXT:    br i1 [[TMP37]], label [[DOTCANCEL_EXIT29:%.*]], label [[DOTCANCEL_CONTINUE30:%.*]]
2755 // CHECK7:       .cancel.exit29:
2756 // CHECK7-NEXT:    br label [[CANCEL_EXIT34:%.*]]
2757 // CHECK7:       cancel.exit19:
2758 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
2759 // CHECK7-NEXT:    br label [[CANCEL_CONT20]]
2760 // CHECK7:       .cancel.continue30:
2761 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
2762 // CHECK7:       omp_if.else:
2763 // CHECK7-NEXT:    br label [[OMP_IF_END]]
2764 // CHECK7:       omp_if.end:
2765 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2766 // CHECK7:       omp.body.continue:
2767 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
2768 // CHECK7:       omp.inner.for.inc31:
2769 // CHECK7-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2770 // CHECK7-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP38]], 1
2771 // CHECK7-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV]], align 4
2772 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND26]]
2773 // CHECK7:       omp.inner.for.end33:
2774 // CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2775 // CHECK7:       omp.loop.exit:
2776 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
2777 // CHECK7-NEXT:    br label [[OMP_PRECOND_END]]
2778 // CHECK7:       cancel.exit34:
2779 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
2780 // CHECK7-NEXT:    br label [[CANCEL_CONT35:%.*]]
2781 // CHECK7:       omp.precond.end:
2782 // CHECK7-NEXT:    br label [[CANCEL_CONT35]]
2783 // CHECK7:       cancel.cont35:
2784 // CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
2785 // 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*)*))
2786 // CHECK7-NEXT:    [[TMP40:%.*]] = bitcast i8* [[TMP39]] to %struct.kmp_task_t_with_privates*
2787 // CHECK7-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP40]], i32 0, i32 0
2788 // CHECK7-NEXT:    [[TMP42:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP39]])
2789 // 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*, ...)*))
2790 // 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*, ...)*))
2791 // CHECK7-NEXT:    store i32 0, i32* [[R]], align 4
2792 // 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]])
2793 // CHECK7-NEXT:    [[TMP43:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2794 // CHECK7-NEXT:    ret i32 [[TMP43]]
2795 //
2796 //
2797 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined.
2798 // 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]+]] {
2799 // CHECK7-NEXT:  entry:
2800 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2801 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2802 // CHECK7-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8***, align 8
2803 // CHECK7-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
2804 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2805 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2806 // CHECK7-NEXT:    store i8*** [[ARGV]], i8**** [[ARGV_ADDR]], align 8
2807 // CHECK7-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
2808 // CHECK7-NEXT:    [[TMP0:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
2809 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
2810 // CHECK7-NEXT:    [[TMP2:%.*]] = load float, float* @flag, align 4
2811 // CHECK7-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP2]], 0.000000e+00
2812 // CHECK7-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2813 // CHECK7:       omp_if.then:
2814 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2815 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2816 // CHECK7-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 1)
2817 // CHECK7-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[TMP5]], 0
2818 // CHECK7-NEXT:    br i1 [[TMP6]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
2819 // CHECK7:       .cancel.exit:
2820 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2821 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
2822 // CHECK7-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]])
2823 // CHECK7-NEXT:    br label [[RETURN:%.*]]
2824 // CHECK7:       .cancel.continue:
2825 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
2826 // CHECK7:       omp_if.else:
2827 // CHECK7-NEXT:    br label [[OMP_IF_END]]
2828 // CHECK7:       omp_if.end:
2829 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP1]], align 4
2830 // CHECK7-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP10]] to i8
2831 // CHECK7-NEXT:    [[TMP11:%.*]] = load i8**, i8*** [[TMP0]], align 8
2832 // CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
2833 // CHECK7-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
2834 // CHECK7-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
2835 // CHECK7-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX1]], align 1
2836 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2837 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
2838 // CHECK7-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP14]])
2839 // CHECK7-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
2840 // CHECK7-NEXT:    br i1 [[TMP16]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
2841 // CHECK7:       .cancel.exit2:
2842 // CHECK7-NEXT:    br label [[RETURN]]
2843 // CHECK7:       .cancel.continue3:
2844 // CHECK7-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP1]], align 4
2845 // CHECK7-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP0]], align 8
2846 // CHECK7-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP18]], i64 0
2847 // CHECK7-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
2848 // CHECK7-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP19]], i64 0
2849 // CHECK7-NEXT:    [[TMP20:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
2850 // CHECK7-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP20]] to i32
2851 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP17]]
2852 // CHECK7-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD]] to i8
2853 // CHECK7-NEXT:    store i8 [[CONV7]], i8* [[ARRAYIDX5]], align 1
2854 // CHECK7-NEXT:    br label [[RETURN]]
2855 // CHECK7:       return:
2856 // CHECK7-NEXT:    ret void
2857 //
2858 //
2859 // CHECK7-LABEL: define {{[^@]+}}@.omp_task_entry.
2860 // CHECK7-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
2861 // CHECK7-NEXT:  entry:
2862 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
2863 // CHECK7-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
2864 // CHECK7-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
2865 // CHECK7-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
2866 // CHECK7-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
2867 // CHECK7-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
2868 // CHECK7-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
2869 // CHECK7-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
2870 // CHECK7-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
2871 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
2872 // CHECK7-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
2873 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
2874 // CHECK7-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
2875 // CHECK7-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
2876 // CHECK7-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
2877 // CHECK7-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
2878 // CHECK7-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2879 // CHECK7-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
2880 // CHECK7-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
2881 // CHECK7-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META5:![0-9]+]])
2882 // CHECK7-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
2883 // CHECK7-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
2884 // CHECK7-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
2885 // CHECK7-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
2886 // CHECK7-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !14
2887 // CHECK7-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !14
2888 // CHECK7-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !14
2889 // CHECK7-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !14
2890 // CHECK7-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
2891 // CHECK7-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
2892 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
2893 // CHECK7-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 4) #[[ATTR2:[0-9]+]]
2894 // CHECK7-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
2895 // CHECK7-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
2896 // CHECK7:       .cancel.exit.i:
2897 // CHECK7-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
2898 // CHECK7-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
2899 // CHECK7:       .cancel.continue.i:
2900 // CHECK7-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
2901 // CHECK7-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
2902 // CHECK7:       .omp_outlined..1.exit:
2903 // CHECK7-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
2904 // CHECK7-NEXT:    ret i32 0
2905 //
2906 //
2907 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..2
2908 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2909 // CHECK7-NEXT:  entry:
2910 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2911 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2912 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
2913 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
2914 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
2915 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
2916 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
2917 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2918 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2919 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2920 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2921 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
2922 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
2923 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2924 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2925 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
2926 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2927 // CHECK7-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
2928 // CHECK7-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
2929 // CHECK7-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
2930 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2931 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2932 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2933 // CHECK7:       omp.inner.for.cond:
2934 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2935 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2936 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2937 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2938 // CHECK7:       omp.inner.for.body:
2939 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2940 // CHECK7-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
2941 // CHECK7-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
2942 // CHECK7-NEXT:    ]
2943 // CHECK7:       .omp.sections.case:
2944 // CHECK7-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
2945 // CHECK7-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
2946 // CHECK7-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
2947 // CHECK7:       .cancel.exit:
2948 // CHECK7-NEXT:    br label [[CANCEL_EXIT:%.*]]
2949 // CHECK7:       .cancel.continue:
2950 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
2951 // CHECK7:       .omp.sections.exit:
2952 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2953 // CHECK7:       omp.inner.for.inc:
2954 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2955 // CHECK7-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
2956 // CHECK7-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2957 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
2958 // CHECK7:       omp.inner.for.end:
2959 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
2960 // CHECK7-NEXT:    br label [[CANCEL_CONT:%.*]]
2961 // CHECK7:       cancel.cont:
2962 // CHECK7-NEXT:    ret void
2963 // CHECK7:       cancel.exit:
2964 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
2965 // CHECK7-NEXT:    br label [[CANCEL_CONT]]
2966 //
2967 //
2968 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..3
2969 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2970 // CHECK7-NEXT:  entry:
2971 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2972 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2973 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
2974 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
2975 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
2976 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
2977 // CHECK7-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
2978 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2979 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2980 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2981 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2982 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
2983 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
2984 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2985 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2986 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
2987 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2988 // CHECK7-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
2989 // CHECK7-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
2990 // CHECK7-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
2991 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
2992 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
2993 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2994 // CHECK7:       omp.inner.for.cond:
2995 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
2996 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
2997 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2998 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2999 // CHECK7:       omp.inner.for.body:
3000 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3001 // CHECK7-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
3002 // CHECK7-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
3003 // CHECK7-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE1:%.*]]
3004 // CHECK7-NEXT:    ]
3005 // CHECK7:       .omp.sections.case:
3006 // CHECK7-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
3007 // CHECK7-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
3008 // CHECK7-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3009 // CHECK7:       .cancel.exit:
3010 // CHECK7-NEXT:    br label [[CANCEL_EXIT:%.*]]
3011 // CHECK7:       .cancel.continue:
3012 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
3013 // CHECK7:       .omp.sections.case1:
3014 // CHECK7-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
3015 // CHECK7-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3016 // CHECK7-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
3017 // CHECK7:       .cancel.exit2:
3018 // CHECK7-NEXT:    br label [[CANCEL_EXIT]]
3019 // CHECK7:       .cancel.continue3:
3020 // CHECK7-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
3021 // CHECK7:       .omp.sections.exit:
3022 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3023 // CHECK7:       omp.inner.for.inc:
3024 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3025 // CHECK7-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
3026 // CHECK7-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3027 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
3028 // CHECK7:       omp.inner.for.end:
3029 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
3030 // CHECK7-NEXT:    br label [[CANCEL_CONT:%.*]]
3031 // CHECK7:       cancel.cont:
3032 // CHECK7-NEXT:    ret void
3033 // CHECK7:       cancel.exit:
3034 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
3035 // CHECK7-NEXT:    br label [[CANCEL_CONT]]
3036 //
3037 //
3038 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..4
3039 // 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]] {
3040 // CHECK7-NEXT:  entry:
3041 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3042 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3043 // CHECK7-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
3044 // CHECK7-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
3045 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3046 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3047 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3048 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3049 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
3050 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3051 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3052 // CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3053 // CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3054 // CHECK7-NEXT:    [[R3:%.*]] = alloca i32, align 4
3055 // CHECK7-NEXT:    [[I4:%.*]] = alloca i32, align 4
3056 // CHECK7-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
3057 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3058 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3059 // CHECK7-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
3060 // CHECK7-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
3061 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
3062 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
3063 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
3064 // CHECK7-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
3065 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3066 // CHECK7-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
3067 // CHECK7-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3068 // CHECK7-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3069 // CHECK7-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3070 // CHECK7-NEXT:    store i32 0, i32* [[I]], align 4
3071 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3072 // CHECK7-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
3073 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3074 // CHECK7:       omp.precond.then:
3075 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3076 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3077 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
3078 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3079 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3080 // CHECK7-NEXT:    store i32 0, i32* [[R3]], align 4
3081 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3082 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
3083 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3084 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3085 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3086 // CHECK7-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
3087 // CHECK7-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3088 // CHECK7:       cond.true:
3089 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3090 // CHECK7-NEXT:    br label [[COND_END:%.*]]
3091 // CHECK7:       cond.false:
3092 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3093 // CHECK7-NEXT:    br label [[COND_END]]
3094 // CHECK7:       cond.end:
3095 // CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
3096 // CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3097 // CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3098 // CHECK7-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
3099 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3100 // CHECK7:       omp.inner.for.cond:
3101 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3102 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3103 // CHECK7-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
3104 // CHECK7-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3105 // CHECK7:       omp.inner.for.body:
3106 // CHECK7-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3107 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
3108 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3109 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
3110 // CHECK7-NEXT:    [[TMP16:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3111 // CHECK7-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4
3112 // CHECK7-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP17]], i32 2)
3113 // CHECK7-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
3114 // CHECK7-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3115 // CHECK7:       .cancel.exit:
3116 // CHECK7-NEXT:    br label [[CANCEL_EXIT:%.*]]
3117 // CHECK7:       .cancel.continue:
3118 // CHECK7-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4
3119 // CHECK7-NEXT:    [[TMP21:%.*]] = load i32, i32* [[R3]], align 4
3120 // CHECK7-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP20]]
3121 // CHECK7-NEXT:    store i32 [[ADD7]], i32* [[R3]], align 4
3122 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3123 // CHECK7:       omp.body.continue:
3124 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3125 // CHECK7:       omp.inner.for.inc:
3126 // CHECK7-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3127 // CHECK7-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
3128 // CHECK7-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
3129 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]]
3130 // CHECK7:       omp.inner.for.end:
3131 // CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3132 // CHECK7:       omp.loop.exit:
3133 // CHECK7-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3134 // CHECK7-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
3135 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP24]])
3136 // CHECK7-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
3137 // CHECK7-NEXT:    [[TMP26:%.*]] = bitcast i32* [[R3]] to i8*
3138 // CHECK7-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
3139 // CHECK7-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3140 // CHECK7-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
3141 // CHECK7-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
3142 // 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)
3143 // CHECK7-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3144 // CHECK7-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3145 // CHECK7-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3146 // CHECK7-NEXT:    ]
3147 // CHECK7:       .omp.reduction.case1:
3148 // CHECK7-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP1]], align 4
3149 // CHECK7-NEXT:    [[TMP32:%.*]] = load i32, i32* [[R3]], align 4
3150 // CHECK7-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
3151 // CHECK7-NEXT:    store i32 [[ADD9]], i32* [[TMP1]], align 4
3152 // CHECK7-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB7]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
3153 // CHECK7-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3154 // CHECK7:       cancel.exit:
3155 // CHECK7-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3156 // CHECK7-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
3157 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP34]])
3158 // CHECK7-NEXT:    br label [[CANCEL_CONT:%.*]]
3159 // CHECK7:       .omp.reduction.case2:
3160 // CHECK7-NEXT:    [[TMP35:%.*]] = load i32, i32* [[R3]], align 4
3161 // CHECK7-NEXT:    [[TMP36:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP35]] monotonic, align 4
3162 // CHECK7-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3163 // CHECK7:       .omp.reduction.default:
3164 // CHECK7-NEXT:    br label [[OMP_PRECOND_END]]
3165 // CHECK7:       omp.precond.end:
3166 // CHECK7-NEXT:    br label [[CANCEL_CONT]]
3167 // CHECK7:       cancel.cont:
3168 // CHECK7-NEXT:    ret void
3169 //
3170 //
3171 // CHECK7-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
3172 // CHECK7-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
3173 // CHECK7-NEXT:  entry:
3174 // CHECK7-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
3175 // CHECK7-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
3176 // CHECK7-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
3177 // CHECK7-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
3178 // CHECK7-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
3179 // CHECK7-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
3180 // CHECK7-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
3181 // CHECK7-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
3182 // CHECK7-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
3183 // CHECK7-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
3184 // CHECK7-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
3185 // CHECK7-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
3186 // CHECK7-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
3187 // CHECK7-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
3188 // CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
3189 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
3190 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3191 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
3192 // CHECK7-NEXT:    ret void
3193 //
3194 //
3195 // CHECK8-LABEL: define {{[^@]+}}@main
3196 // CHECK8-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
3197 // CHECK8-NEXT:  entry:
3198 // CHECK8-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
3199 // CHECK8-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
3200 // CHECK8-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
3201 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
3202 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
3203 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
3204 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
3205 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
3206 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_LB_1:%.*]] = alloca i32, align 4
3207 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_UB_2:%.*]] = alloca i32, align 4
3208 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_ST_3:%.*]] = alloca i32, align 4
3209 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IL_4:%.*]] = alloca i32, align 4
3210 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IV_5:%.*]] = alloca i32, align 4
3211 // CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3212 // CHECK8-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3213 // CHECK8-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3214 // CHECK8-NEXT:    [[DOTCAPTURE_EXPR_21:%.*]] = alloca i32, align 4
3215 // CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
3216 // CHECK8-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3217 // CHECK8-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3218 // CHECK8-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3219 // CHECK8-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3220 // CHECK8-NEXT:    [[I24:%.*]] = alloca i32, align 4
3221 // CHECK8-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
3222 // CHECK8-NEXT:    [[R:%.*]] = alloca i32, align 4
3223 // CHECK8-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
3224 // CHECK8-NEXT:    store i32 0, i32* [[RETVAL]], align 4
3225 // CHECK8-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3226 // CHECK8-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
3227 // 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]])
3228 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
3229 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3230 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
3231 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
3232 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[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)
3233 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3234 // CHECK8-NEXT:    [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
3235 // CHECK8-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i32 [[TMP1]], i32 0
3236 // CHECK8-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
3237 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
3238 // CHECK8-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3239 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3240 // CHECK8:       omp.inner.for.cond:
3241 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3242 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3243 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3244 // CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3245 // CHECK8:       omp.inner.for.body:
3246 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3247 // CHECK8-NEXT:    switch i32 [[TMP7]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
3248 // CHECK8-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
3249 // CHECK8-NEXT:    ]
3250 // CHECK8:       .omp.sections.case:
3251 // CHECK8-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
3252 // CHECK8-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
3253 // CHECK8-NEXT:    br i1 [[TMP9]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3254 // CHECK8:       .cancel.exit:
3255 // CHECK8-NEXT:    br label [[CANCEL_EXIT:%.*]]
3256 // CHECK8:       .cancel.continue:
3257 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
3258 // CHECK8:       .omp.sections.exit:
3259 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3260 // CHECK8:       omp.inner.for.inc:
3261 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3262 // CHECK8-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP10]], 1
3263 // CHECK8-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3264 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
3265 // CHECK8:       omp.inner.for.end:
3266 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
3267 // CHECK8-NEXT:    br label [[CANCEL_CONT:%.*]]
3268 // CHECK8:       cancel.cont:
3269 // CHECK8-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5:[0-9]+]], i32 [[TMP0]])
3270 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
3271 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
3272 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_3]], align 4
3273 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_4]], align 4
3274 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], 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)
3275 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
3276 // CHECK8-NEXT:    [[TMP12:%.*]] = icmp slt i32 [[TMP11]], 1
3277 // CHECK8-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 1
3278 // CHECK8-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_SECTIONS_UB_2]], align 4
3279 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_1]], align 4
3280 // CHECK8-NEXT:    store i32 [[TMP14]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
3281 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND6:%.*]]
3282 // CHECK8:       omp.inner.for.cond6:
3283 // CHECK8-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
3284 // CHECK8-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_2]], align 4
3285 // CHECK8-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
3286 // CHECK8-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY8:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
3287 // CHECK8:       omp.inner.for.body8:
3288 // CHECK8-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
3289 // CHECK8-NEXT:    switch i32 [[TMP17]], label [[DOTOMP_SECTIONS_EXIT15:%.*]] [
3290 // CHECK8-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE9:%.*]]
3291 // CHECK8-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE12:%.*]]
3292 // CHECK8-NEXT:    ]
3293 // CHECK8:       .omp.sections.case9:
3294 // CHECK8-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
3295 // CHECK8-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
3296 // CHECK8-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT10:%.*]], label [[DOTCANCEL_CONTINUE11:%.*]]
3297 // CHECK8:       .cancel.exit10:
3298 // CHECK8-NEXT:    br label [[CANCEL_EXIT19:%.*]]
3299 // CHECK8:       cancel.exit:
3300 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
3301 // CHECK8-NEXT:    br label [[CANCEL_CONT]]
3302 // CHECK8:       .cancel.continue11:
3303 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
3304 // CHECK8:       .omp.sections.case12:
3305 // CHECK8-NEXT:    [[TMP20:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 3)
3306 // CHECK8-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
3307 // CHECK8-NEXT:    br i1 [[TMP21]], label [[DOTCANCEL_EXIT13:%.*]], label [[DOTCANCEL_CONTINUE14:%.*]]
3308 // CHECK8:       .cancel.exit13:
3309 // CHECK8-NEXT:    br label [[CANCEL_EXIT19]]
3310 // CHECK8:       .cancel.continue14:
3311 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT15]]
3312 // CHECK8:       .omp.sections.exit15:
3313 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
3314 // CHECK8:       omp.inner.for.inc16:
3315 // CHECK8-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_5]], align 4
3316 // CHECK8-NEXT:    [[INC17:%.*]] = add nsw i32 [[TMP22]], 1
3317 // CHECK8-NEXT:    store i32 [[INC17]], i32* [[DOTOMP_SECTIONS_IV_5]], align 4
3318 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND6]]
3319 // CHECK8:       omp.inner.for.end18:
3320 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
3321 // CHECK8-NEXT:    br label [[CANCEL_CONT20:%.*]]
3322 // CHECK8:       cancel.cont20:
3323 // CHECK8-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB5]], i32 [[TMP0]])
3324 // CHECK8-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3325 // CHECK8-NEXT:    store i32 [[TMP23]], i32* [[DOTCAPTURE_EXPR_]], align 4
3326 // CHECK8-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3327 // CHECK8-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
3328 // CHECK8-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3329 // CHECK8-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[DIV]], 1
3330 // CHECK8-NEXT:    store i32 [[SUB22]], i32* [[DOTCAPTURE_EXPR_21]], align 4
3331 // CHECK8-NEXT:    store i32 0, i32* [[I]], align 4
3332 // CHECK8-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3333 // CHECK8-NEXT:    [[CMP23:%.*]] = icmp slt i32 0, [[TMP25]]
3334 // CHECK8-NEXT:    br i1 [[CMP23]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3335 // CHECK8:       omp.precond.then:
3336 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3337 // CHECK8-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
3338 // CHECK8-NEXT:    store i32 [[TMP26]], i32* [[DOTOMP_UB]], align 4
3339 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3340 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3341 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3342 // CHECK8-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3343 // CHECK8-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
3344 // CHECK8-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
3345 // CHECK8-NEXT:    br i1 [[CMP25]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3346 // CHECK8:       cond.true:
3347 // CHECK8-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_21]], align 4
3348 // CHECK8-NEXT:    br label [[COND_END:%.*]]
3349 // CHECK8:       cond.false:
3350 // CHECK8-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3351 // CHECK8-NEXT:    br label [[COND_END]]
3352 // CHECK8:       cond.end:
3353 // CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE]] ], [ [[TMP30]], [[COND_FALSE]] ]
3354 // CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3355 // CHECK8-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3356 // CHECK8-NEXT:    store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
3357 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND26:%.*]]
3358 // CHECK8:       omp.inner.for.cond26:
3359 // CHECK8-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3360 // CHECK8-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3361 // CHECK8-NEXT:    [[CMP27:%.*]] = icmp sle i32 [[TMP32]], [[TMP33]]
3362 // CHECK8-NEXT:    br i1 [[CMP27]], label [[OMP_INNER_FOR_BODY28:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
3363 // CHECK8:       omp.inner.for.body28:
3364 // CHECK8-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3365 // CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP34]], 1
3366 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3367 // CHECK8-NEXT:    store i32 [[ADD]], i32* [[I24]], align 4
3368 // CHECK8-NEXT:    [[TMP35:%.*]] = load float, float* @flag, align 4
3369 // CHECK8-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP35]], 0.000000e+00
3370 // CHECK8-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3371 // CHECK8:       omp_if.then:
3372 // CHECK8-NEXT:    [[TMP36:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 2)
3373 // CHECK8-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
3374 // CHECK8-NEXT:    br i1 [[TMP37]], label [[DOTCANCEL_EXIT29:%.*]], label [[DOTCANCEL_CONTINUE30:%.*]]
3375 // CHECK8:       .cancel.exit29:
3376 // CHECK8-NEXT:    br label [[CANCEL_EXIT34:%.*]]
3377 // CHECK8:       cancel.exit19:
3378 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP0]])
3379 // CHECK8-NEXT:    br label [[CANCEL_CONT20]]
3380 // CHECK8:       .cancel.continue30:
3381 // CHECK8-NEXT:    br label [[OMP_IF_END:%.*]]
3382 // CHECK8:       omp_if.else:
3383 // CHECK8-NEXT:    br label [[OMP_IF_END]]
3384 // CHECK8:       omp_if.end:
3385 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3386 // CHECK8:       omp.body.continue:
3387 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC31:%.*]]
3388 // CHECK8:       omp.inner.for.inc31:
3389 // CHECK8-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3390 // CHECK8-NEXT:    [[ADD32:%.*]] = add nsw i32 [[TMP38]], 1
3391 // CHECK8-NEXT:    store i32 [[ADD32]], i32* [[DOTOMP_IV]], align 4
3392 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND26]]
3393 // CHECK8:       omp.inner.for.end33:
3394 // CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3395 // CHECK8:       omp.loop.exit:
3396 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
3397 // CHECK8-NEXT:    br label [[OMP_PRECOND_END]]
3398 // CHECK8:       cancel.exit34:
3399 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP0]])
3400 // CHECK8-NEXT:    br label [[CANCEL_CONT35:%.*]]
3401 // CHECK8:       omp.precond.end:
3402 // CHECK8-NEXT:    br label [[CANCEL_CONT35]]
3403 // CHECK8:       cancel.cont35:
3404 // CHECK8-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP0]])
3405 // 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*)*))
3406 // CHECK8-NEXT:    [[TMP40:%.*]] = bitcast i8* [[TMP39]] to %struct.kmp_task_t_with_privates*
3407 // CHECK8-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP40]], i32 0, i32 0
3408 // CHECK8-NEXT:    [[TMP42:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP39]])
3409 // 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*, ...)*))
3410 // 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*, ...)*))
3411 // CHECK8-NEXT:    store i32 0, i32* [[R]], align 4
3412 // 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]])
3413 // CHECK8-NEXT:    [[TMP43:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3414 // CHECK8-NEXT:    ret i32 [[TMP43]]
3415 //
3416 //
3417 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined.
3418 // 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]+]] {
3419 // CHECK8-NEXT:  entry:
3420 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3421 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3422 // CHECK8-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8***, align 8
3423 // CHECK8-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
3424 // CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3425 // CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3426 // CHECK8-NEXT:    store i8*** [[ARGV]], i8**** [[ARGV_ADDR]], align 8
3427 // CHECK8-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
3428 // CHECK8-NEXT:    [[TMP0:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
3429 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
3430 // CHECK8-NEXT:    [[TMP2:%.*]] = load float, float* @flag, align 4
3431 // CHECK8-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP2]], 0.000000e+00
3432 // CHECK8-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3433 // CHECK8:       omp_if.then:
3434 // CHECK8-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3435 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
3436 // CHECK8-NEXT:    [[TMP5:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 1)
3437 // CHECK8-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[TMP5]], 0
3438 // CHECK8-NEXT:    br i1 [[TMP6]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3439 // CHECK8:       .cancel.exit:
3440 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3441 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3442 // CHECK8-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]])
3443 // CHECK8-NEXT:    br label [[RETURN:%.*]]
3444 // CHECK8:       .cancel.continue:
3445 // CHECK8-NEXT:    br label [[OMP_IF_END:%.*]]
3446 // CHECK8:       omp_if.else:
3447 // CHECK8-NEXT:    br label [[OMP_IF_END]]
3448 // CHECK8:       omp_if.end:
3449 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP1]], align 4
3450 // CHECK8-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP10]] to i8
3451 // CHECK8-NEXT:    [[TMP11:%.*]] = load i8**, i8*** [[TMP0]], align 8
3452 // CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
3453 // CHECK8-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
3454 // CHECK8-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
3455 // CHECK8-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX1]], align 1
3456 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3457 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
3458 // CHECK8-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP14]])
3459 // CHECK8-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
3460 // CHECK8-NEXT:    br i1 [[TMP16]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
3461 // CHECK8:       .cancel.exit2:
3462 // CHECK8-NEXT:    br label [[RETURN]]
3463 // CHECK8:       .cancel.continue3:
3464 // CHECK8-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP1]], align 4
3465 // CHECK8-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP0]], align 8
3466 // CHECK8-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8*, i8** [[TMP18]], i64 0
3467 // CHECK8-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[ARRAYIDX4]], align 8
3468 // CHECK8-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i8, i8* [[TMP19]], i64 0
3469 // CHECK8-NEXT:    [[TMP20:%.*]] = load i8, i8* [[ARRAYIDX5]], align 1
3470 // CHECK8-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP20]] to i32
3471 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV6]], [[TMP17]]
3472 // CHECK8-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD]] to i8
3473 // CHECK8-NEXT:    store i8 [[CONV7]], i8* [[ARRAYIDX5]], align 1
3474 // CHECK8-NEXT:    br label [[RETURN]]
3475 // CHECK8:       return:
3476 // CHECK8-NEXT:    ret void
3477 //
3478 //
3479 // CHECK8-LABEL: define {{[^@]+}}@.omp_task_entry.
3480 // CHECK8-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
3481 // CHECK8-NEXT:  entry:
3482 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
3483 // CHECK8-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
3484 // CHECK8-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
3485 // CHECK8-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
3486 // CHECK8-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
3487 // CHECK8-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
3488 // CHECK8-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
3489 // CHECK8-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
3490 // CHECK8-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
3491 // CHECK8-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
3492 // CHECK8-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
3493 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
3494 // CHECK8-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
3495 // CHECK8-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
3496 // CHECK8-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
3497 // CHECK8-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
3498 // CHECK8-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
3499 // CHECK8-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
3500 // CHECK8-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
3501 // CHECK8-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META5:![0-9]+]])
3502 // CHECK8-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
3503 // CHECK8-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
3504 // CHECK8-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
3505 // CHECK8-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
3506 // CHECK8-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !14
3507 // CHECK8-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !14
3508 // CHECK8-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !14
3509 // CHECK8-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !14
3510 // CHECK8-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
3511 // CHECK8-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
3512 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
3513 // CHECK8-NEXT:    [[TMP12:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 4) #[[ATTR2:[0-9]+]]
3514 // CHECK8-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
3515 // CHECK8-NEXT:    br i1 [[TMP13]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
3516 // CHECK8:       .cancel.exit.i:
3517 // CHECK8-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
3518 // CHECK8-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
3519 // CHECK8:       .cancel.continue.i:
3520 // CHECK8-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
3521 // CHECK8-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
3522 // CHECK8:       .omp_outlined..1.exit:
3523 // CHECK8-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
3524 // CHECK8-NEXT:    ret i32 0
3525 //
3526 //
3527 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..2
3528 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3529 // CHECK8-NEXT:  entry:
3530 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3531 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3532 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
3533 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
3534 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
3535 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
3536 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
3537 // CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3538 // CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3539 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
3540 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3541 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
3542 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
3543 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3544 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3545 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
3546 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3547 // CHECK8-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 0
3548 // CHECK8-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 0
3549 // CHECK8-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
3550 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
3551 // CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3552 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3553 // CHECK8:       omp.inner.for.cond:
3554 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3555 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3556 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3557 // CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3558 // CHECK8:       omp.inner.for.body:
3559 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3560 // CHECK8-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
3561 // CHECK8-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
3562 // CHECK8-NEXT:    ]
3563 // CHECK8:       .omp.sections.case:
3564 // CHECK8-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
3565 // CHECK8-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
3566 // CHECK8-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3567 // CHECK8:       .cancel.exit:
3568 // CHECK8-NEXT:    br label [[CANCEL_EXIT:%.*]]
3569 // CHECK8:       .cancel.continue:
3570 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
3571 // CHECK8:       .omp.sections.exit:
3572 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3573 // CHECK8:       omp.inner.for.inc:
3574 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3575 // CHECK8-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
3576 // CHECK8-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3577 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
3578 // CHECK8:       omp.inner.for.end:
3579 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
3580 // CHECK8-NEXT:    br label [[CANCEL_CONT:%.*]]
3581 // CHECK8:       cancel.cont:
3582 // CHECK8-NEXT:    ret void
3583 // CHECK8:       cancel.exit:
3584 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
3585 // CHECK8-NEXT:    br label [[CANCEL_CONT]]
3586 //
3587 //
3588 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..3
3589 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3590 // CHECK8-NEXT:  entry:
3591 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3592 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3593 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
3594 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
3595 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
3596 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
3597 // CHECK8-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
3598 // CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3599 // CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3600 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
3601 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3602 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
3603 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
3604 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3605 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3606 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
3607 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3608 // CHECK8-NEXT:    [[TMP3:%.*]] = icmp slt i32 [[TMP2]], 1
3609 // CHECK8-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 1
3610 // CHECK8-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
3611 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
3612 // CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3613 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3614 // CHECK8:       omp.inner.for.cond:
3615 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3616 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
3617 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3618 // CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3619 // CHECK8:       omp.inner.for.body:
3620 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3621 // CHECK8-NEXT:    switch i32 [[TMP8]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
3622 // CHECK8-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
3623 // CHECK8-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE1:%.*]]
3624 // CHECK8-NEXT:    ]
3625 // CHECK8:       .omp.sections.case:
3626 // CHECK8-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
3627 // CHECK8-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
3628 // CHECK8-NEXT:    br i1 [[TMP10]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3629 // CHECK8:       .cancel.exit:
3630 // CHECK8-NEXT:    br label [[CANCEL_EXIT:%.*]]
3631 // CHECK8:       .cancel.continue:
3632 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
3633 // CHECK8:       .omp.sections.case1:
3634 // CHECK8-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 3)
3635 // CHECK8-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3636 // CHECK8-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT2:%.*]], label [[DOTCANCEL_CONTINUE3:%.*]]
3637 // CHECK8:       .cancel.exit2:
3638 // CHECK8-NEXT:    br label [[CANCEL_EXIT]]
3639 // CHECK8:       .cancel.continue3:
3640 // CHECK8-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
3641 // CHECK8:       .omp.sections.exit:
3642 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3643 // CHECK8:       omp.inner.for.inc:
3644 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
3645 // CHECK8-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP13]], 1
3646 // CHECK8-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
3647 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
3648 // CHECK8:       omp.inner.for.end:
3649 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
3650 // CHECK8-NEXT:    br label [[CANCEL_CONT:%.*]]
3651 // CHECK8:       cancel.cont:
3652 // CHECK8-NEXT:    ret void
3653 // CHECK8:       cancel.exit:
3654 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[TMP1]])
3655 // CHECK8-NEXT:    br label [[CANCEL_CONT]]
3656 //
3657 //
3658 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..4
3659 // 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]] {
3660 // CHECK8-NEXT:  entry:
3661 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3662 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3663 // CHECK8-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
3664 // CHECK8-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
3665 // CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3666 // CHECK8-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3667 // CHECK8-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3668 // CHECK8-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3669 // CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
3670 // CHECK8-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3671 // CHECK8-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3672 // CHECK8-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3673 // CHECK8-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3674 // CHECK8-NEXT:    [[R3:%.*]] = alloca i32, align 4
3675 // CHECK8-NEXT:    [[I4:%.*]] = alloca i32, align 4
3676 // CHECK8-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
3677 // CHECK8-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3678 // CHECK8-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3679 // CHECK8-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
3680 // CHECK8-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
3681 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
3682 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
3683 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
3684 // CHECK8-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
3685 // CHECK8-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3686 // CHECK8-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
3687 // CHECK8-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3688 // CHECK8-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3689 // CHECK8-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3690 // CHECK8-NEXT:    store i32 0, i32* [[I]], align 4
3691 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3692 // CHECK8-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
3693 // CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3694 // CHECK8:       omp.precond.then:
3695 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3696 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3697 // CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
3698 // CHECK8-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3699 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3700 // CHECK8-NEXT:    store i32 0, i32* [[R3]], align 4
3701 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3702 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
3703 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6]], i32 [[TMP7]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3704 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3705 // CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3706 // CHECK8-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
3707 // CHECK8-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3708 // CHECK8:       cond.true:
3709 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3710 // CHECK8-NEXT:    br label [[COND_END:%.*]]
3711 // CHECK8:       cond.false:
3712 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3713 // CHECK8-NEXT:    br label [[COND_END]]
3714 // CHECK8:       cond.end:
3715 // CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
3716 // CHECK8-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3717 // CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3718 // CHECK8-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
3719 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3720 // CHECK8:       omp.inner.for.cond:
3721 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3722 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3723 // CHECK8-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
3724 // CHECK8-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3725 // CHECK8:       omp.inner.for.body:
3726 // CHECK8-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3727 // CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
3728 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3729 // CHECK8-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
3730 // CHECK8-NEXT:    [[TMP16:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3731 // CHECK8-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP16]], align 4
3732 // CHECK8-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[TMP17]], i32 2)
3733 // CHECK8-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
3734 // CHECK8-NEXT:    br i1 [[TMP19]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3735 // CHECK8:       .cancel.exit:
3736 // CHECK8-NEXT:    br label [[CANCEL_EXIT:%.*]]
3737 // CHECK8:       .cancel.continue:
3738 // CHECK8-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I4]], align 4
3739 // CHECK8-NEXT:    [[TMP21:%.*]] = load i32, i32* [[R3]], align 4
3740 // CHECK8-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP20]]
3741 // CHECK8-NEXT:    store i32 [[ADD7]], i32* [[R3]], align 4
3742 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3743 // CHECK8:       omp.body.continue:
3744 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3745 // CHECK8:       omp.inner.for.inc:
3746 // CHECK8-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3747 // CHECK8-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
3748 // CHECK8-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
3749 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
3750 // CHECK8:       omp.inner.for.end:
3751 // CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3752 // CHECK8:       omp.loop.exit:
3753 // CHECK8-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3754 // CHECK8-NEXT:    [[TMP24:%.*]] = load i32, i32* [[TMP23]], align 4
3755 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP24]])
3756 // CHECK8-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
3757 // CHECK8-NEXT:    [[TMP26:%.*]] = bitcast i32* [[R3]] to i8*
3758 // CHECK8-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
3759 // CHECK8-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3760 // CHECK8-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
3761 // CHECK8-NEXT:    [[TMP29:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
3762 // 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)
3763 // CHECK8-NEXT:    switch i32 [[TMP30]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3764 // CHECK8-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3765 // CHECK8-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3766 // CHECK8-NEXT:    ]
3767 // CHECK8:       .omp.reduction.case1:
3768 // CHECK8-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP1]], align 4
3769 // CHECK8-NEXT:    [[TMP32:%.*]] = load i32, i32* [[R3]], align 4
3770 // CHECK8-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP31]], [[TMP32]]
3771 // CHECK8-NEXT:    store i32 [[ADD9]], i32* [[TMP1]], align 4
3772 // CHECK8-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB7]], i32 [[TMP28]], [8 x i32]* @.gomp_critical_user_.reduction.var)
3773 // CHECK8-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3774 // CHECK8:       cancel.exit:
3775 // CHECK8-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3776 // CHECK8-NEXT:    [[TMP34:%.*]] = load i32, i32* [[TMP33]], align 4
3777 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB6]], i32 [[TMP34]])
3778 // CHECK8-NEXT:    br label [[CANCEL_CONT:%.*]]
3779 // CHECK8:       .omp.reduction.case2:
3780 // CHECK8-NEXT:    [[TMP35:%.*]] = load i32, i32* [[R3]], align 4
3781 // CHECK8-NEXT:    [[TMP36:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP35]] monotonic, align 4
3782 // CHECK8-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3783 // CHECK8:       .omp.reduction.default:
3784 // CHECK8-NEXT:    br label [[OMP_PRECOND_END]]
3785 // CHECK8:       omp.precond.end:
3786 // CHECK8-NEXT:    br label [[CANCEL_CONT]]
3787 // CHECK8:       cancel.cont:
3788 // CHECK8-NEXT:    ret void
3789 //
3790 //
3791 // CHECK8-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
3792 // CHECK8-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
3793 // CHECK8-NEXT:  entry:
3794 // CHECK8-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
3795 // CHECK8-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
3796 // CHECK8-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
3797 // CHECK8-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
3798 // CHECK8-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
3799 // CHECK8-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
3800 // CHECK8-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
3801 // CHECK8-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
3802 // CHECK8-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
3803 // CHECK8-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
3804 // CHECK8-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
3805 // CHECK8-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
3806 // CHECK8-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
3807 // CHECK8-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
3808 // CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
3809 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
3810 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3811 // CHECK8-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
3812 // CHECK8-NEXT:    ret void
3813 //
3814 //
3815 // CHECK9-LABEL: define {{[^@]+}}@main
3816 // CHECK9-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
3817 // CHECK9-NEXT:  entry:
3818 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
3819 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
3820 // CHECK9-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
3821 // CHECK9-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
3822 // CHECK9-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
3823 // CHECK9-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
3824 // CHECK9-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
3825 // CHECK9-NEXT:    [[P_LASTITER27:%.*]] = alloca i32, align 4
3826 // CHECK9-NEXT:    [[P_LOWERBOUND28:%.*]] = alloca i32, align 4
3827 // CHECK9-NEXT:    [[P_UPPERBOUND29:%.*]] = alloca i32, align 4
3828 // CHECK9-NEXT:    [[P_STRIDE30:%.*]] = alloca i32, align 4
3829 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3830 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3831 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3832 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
3833 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
3834 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3835 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3836 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3837 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3838 // CHECK9-NEXT:    [[I36:%.*]] = alloca i32, align 4
3839 // CHECK9-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
3840 // CHECK9-NEXT:    [[R:%.*]] = alloca i32, align 4
3841 // CHECK9-NEXT:    store i32 0, i32* [[RETVAL]], align 4
3842 // CHECK9-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3843 // CHECK9-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
3844 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
3845 // CHECK9-NEXT:    br label [[OMP_PARALLEL:%.*]]
3846 // CHECK9:       omp_parallel:
3847 // 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]])
3848 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
3849 // CHECK9:       omp.par.outlined.exit:
3850 // CHECK9-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
3851 // CHECK9:       omp.par.exit.split:
3852 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
3853 // CHECK9:       omp_section_loop.preheader:
3854 // CHECK9-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
3855 // CHECK9-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
3856 // CHECK9-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
3857 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3858 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
3859 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
3860 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
3861 // CHECK9-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
3862 // CHECK9-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
3863 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
3864 // CHECK9:       omp_section_loop.header:
3865 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
3866 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
3867 // CHECK9:       omp_section_loop.cond:
3868 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
3869 // CHECK9-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
3870 // CHECK9:       omp_section_loop.body:
3871 // CHECK9-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
3872 // CHECK9-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
3873 // CHECK9-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
3874 // CHECK9-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
3875 // CHECK9-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
3876 // CHECK9-NEXT:    ]
3877 // CHECK9:       omp_section_loop.inc:
3878 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
3879 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
3880 // CHECK9:       omp_section_loop.exit:
3881 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
3882 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3883 // CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
3884 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
3885 // CHECK9:       omp_section_loop.after:
3886 // CHECK9-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
3887 // CHECK9:       omp_sections.end:
3888 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
3889 // CHECK9:       omp_section_loop.preheader13:
3890 // CHECK9-NEXT:    store i32 0, i32* [[P_LOWERBOUND28]], align 4
3891 // CHECK9-NEXT:    store i32 1, i32* [[P_UPPERBOUND29]], align 4
3892 // CHECK9-NEXT:    store i32 1, i32* [[P_STRIDE30]], align 4
3893 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3894 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]], i32 34, i32* [[P_LASTITER27]], i32* [[P_LOWERBOUND28]], i32* [[P_UPPERBOUND29]], i32* [[P_STRIDE30]], i32 1, i32 1)
3895 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND28]], align 4
3896 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND29]], align 4
3897 // CHECK9-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
3898 // CHECK9-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
3899 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
3900 // CHECK9:       omp_section_loop.header14:
3901 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
3902 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
3903 // CHECK9:       omp_section_loop.cond15:
3904 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP10]]
3905 // CHECK9-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
3906 // CHECK9:       omp_section_loop.body16:
3907 // CHECK9-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP7]]
3908 // CHECK9-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
3909 // CHECK9-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
3910 // CHECK9-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC17]] [
3911 // CHECK9-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
3912 // CHECK9-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
3913 // CHECK9-NEXT:    ]
3914 // CHECK9:       omp_section_loop.inc17:
3915 // CHECK9-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
3916 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
3917 // CHECK9:       omp_section_loop.exit18:
3918 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
3919 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3920 // CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
3921 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
3922 // CHECK9:       omp_section_loop.after19:
3923 // CHECK9-NEXT:    br label [[OMP_SECTIONS_END33:%.*]]
3924 // CHECK9:       omp_sections.end33:
3925 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3926 // CHECK9-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
3927 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3928 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
3929 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3930 // CHECK9-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
3931 // CHECK9-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
3932 // CHECK9-NEXT:    store i32 0, i32* [[I]], align 4
3933 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3934 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
3935 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3936 // CHECK9:       omp.precond.then:
3937 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3938 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
3939 // CHECK9-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
3940 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3941 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3942 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
3943 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM37]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3944 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3945 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
3946 // CHECK9-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
3947 // CHECK9-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3948 // CHECK9:       cond.true:
3949 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
3950 // CHECK9-NEXT:    br label [[COND_END:%.*]]
3951 // CHECK9:       cond.false:
3952 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3953 // CHECK9-NEXT:    br label [[COND_END]]
3954 // CHECK9:       cond.end:
3955 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
3956 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3957 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3958 // CHECK9-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
3959 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3960 // CHECK9:       omp.inner.for.cond:
3961 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3962 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3963 // CHECK9-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
3964 // CHECK9-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3965 // CHECK9:       omp.inner.for.body:
3966 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3967 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
3968 // CHECK9-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
3969 // CHECK9-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
3970 // CHECK9-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
3971 // CHECK9-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
3972 // CHECK9-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3973 // CHECK9:       omp_if.then:
3974 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
3975 // CHECK9-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
3976 // CHECK9-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
3977 // CHECK9-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
3978 // CHECK9:       .cancel.exit:
3979 // CHECK9-NEXT:    br label [[CANCEL_EXIT:%.*]]
3980 // CHECK9:       omp_section_loop.body.case:
3981 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3982 // CHECK9-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
3983 // CHECK9-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
3984 // CHECK9-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
3985 // CHECK9:       omp_section_loop.body.case.split:
3986 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
3987 // CHECK9:       omp_section_loop.body.case.cncl:
3988 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
3989 // CHECK9:       omp_section_loop.body.case23:
3990 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3991 // CHECK9-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
3992 // CHECK9-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
3993 // CHECK9-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
3994 // CHECK9:       omp_section_loop.body.case23.split:
3995 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
3996 // CHECK9:       omp_section_loop.body.case23.cncl:
3997 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
3998 // CHECK9:       omp_section_loop.body.case25:
3999 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM26:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4000 // CHECK9-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM26]], i32 3)
4001 // CHECK9-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
4002 // CHECK9-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
4003 // CHECK9:       omp_section_loop.body.case25.split:
4004 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
4005 // CHECK9:       omp_section_loop.body.case25.cncl:
4006 // CHECK9-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
4007 // CHECK9:       .cancel.continue:
4008 // CHECK9-NEXT:    br label [[OMP_IF_END:%.*]]
4009 // CHECK9:       omp_if.else:
4010 // CHECK9-NEXT:    br label [[OMP_IF_END]]
4011 // CHECK9:       omp_if.end:
4012 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4013 // CHECK9:       omp.body.continue:
4014 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4015 // CHECK9:       omp.inner.for.inc:
4016 // CHECK9-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4017 // CHECK9-NEXT:    [[ADD43:%.*]] = add nsw i32 [[TMP35]], 1
4018 // CHECK9-NEXT:    store i32 [[ADD43]], i32* [[DOTOMP_IV]], align 4
4019 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
4020 // CHECK9:       omp.inner.for.end:
4021 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4022 // CHECK9:       omp.loop.exit:
4023 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
4024 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
4025 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
4026 // CHECK9:       cancel.exit:
4027 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
4028 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
4029 // CHECK9-NEXT:    br label [[CANCEL_CONT:%.*]]
4030 // CHECK9:       omp.precond.end:
4031 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
4032 // CHECK9:       cancel.cont:
4033 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4034 // CHECK9-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM46]])
4035 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
4036 // CHECK9-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], 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*)*))
4037 // CHECK9-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
4038 // CHECK9-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
4039 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM48:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
4040 // CHECK9-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM48]], i8* [[TMP36]])
4041 // 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*, ...)*))
4042 // 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*, ...)*))
4043 // CHECK9-NEXT:    store i32 0, i32* [[R]], align 4
4044 // 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]])
4045 // CHECK9-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4046 // CHECK9-NEXT:    ret i32 [[TMP40]]
4047 //
4048 //
4049 // CHECK9-LABEL: define {{[^@]+}}@main..omp_par
4050 // CHECK9-SAME: (i32* noalias [[TID_ADDR:%.*]], i32* noalias [[ZERO_ADDR:%.*]], i32* [[ARGC_ADDR:%.*]], i8*** [[ARGV_ADDR:%.*]]) #[[ATTR1:[0-9]+]] {
4051 // CHECK9-NEXT:  omp.par.entry:
4052 // CHECK9-NEXT:    [[TID_ADDR_LOCAL:%.*]] = alloca i32, align 4
4053 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[TID_ADDR]], align 4
4054 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[TID_ADDR_LOCAL]], align 4
4055 // CHECK9-NEXT:    [[TID:%.*]] = load i32, i32* [[TID_ADDR_LOCAL]], align 4
4056 // CHECK9-NEXT:    br label [[OMP_PAR_REGION:%.*]]
4057 // CHECK9:       omp.par.region:
4058 // CHECK9-NEXT:    [[TMP1:%.*]] = load float, float* @flag, align 4
4059 // CHECK9-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP1]], 0.000000e+00
4060 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[TMP13:%.*]], label [[TMP2:%.*]]
4061 // CHECK9:       2:
4062 // CHECK9-NEXT:    br label [[TMP3:%.*]]
4063 // CHECK9:       3:
4064 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4065 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP4]] to i8
4066 // CHECK9-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
4067 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 0
4068 // CHECK9-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
4069 // CHECK9-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
4070 // CHECK9-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX3]], align 1
4071 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4072 // CHECK9-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
4073 // CHECK9-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
4074 // CHECK9-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL5:%.*]]
4075 // CHECK9:       .cncl5:
4076 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
4077 // CHECK9:       .cont:
4078 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4079 // CHECK9-NEXT:    [[TMP10:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
4080 // CHECK9-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
4081 // CHECK9-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
4082 // CHECK9-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
4083 // CHECK9-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
4084 // CHECK9-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP12]] to i32
4085 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP9]]
4086 // CHECK9-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
4087 // CHECK9-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
4088 // CHECK9-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
4089 // CHECK9:       omp.par.pre_finalize:
4090 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
4091 // CHECK9:       13:
4092 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4093 // CHECK9-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
4094 // CHECK9-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 0
4095 // CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
4096 // CHECK9:       .cncl:
4097 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4098 // CHECK9-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
4099 // CHECK9-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
4100 // CHECK9:       .split:
4101 // CHECK9-NEXT:    br label [[TMP3]]
4102 // CHECK9:       omp.par.outlined.exit.exitStub:
4103 // CHECK9-NEXT:    ret void
4104 //
4105 //
4106 // CHECK9-LABEL: define {{[^@]+}}@.omp_task_entry.
4107 // CHECK9-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
4108 // CHECK9-NEXT:  entry:
4109 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
4110 // CHECK9-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
4111 // CHECK9-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
4112 // CHECK9-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
4113 // CHECK9-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
4114 // CHECK9-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
4115 // CHECK9-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
4116 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
4117 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
4118 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
4119 // CHECK9-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
4120 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
4121 // CHECK9-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
4122 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
4123 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
4124 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
4125 // CHECK9-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
4126 // CHECK9-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
4127 // CHECK9-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
4128 // CHECK9-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META5:![0-9]+]])
4129 // CHECK9-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
4130 // CHECK9-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
4131 // CHECK9-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
4132 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
4133 // CHECK9-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !14
4134 // CHECK9-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !14
4135 // CHECK9-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !14
4136 // CHECK9-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !14
4137 // CHECK9-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
4138 // CHECK9-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
4139 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM_I:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]]) #[[ATTR2:[0-9]+]]
4140 // CHECK9-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM_I]], i32 4) #[[ATTR2]]
4141 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4142 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
4143 // CHECK9:       .cancel.exit.i:
4144 // CHECK9-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
4145 // CHECK9-NEXT:    br label [[DOTOMP_OUTLINED__EXIT:%.*]]
4146 // CHECK9:       .cancel.continue.i:
4147 // CHECK9-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
4148 // CHECK9-NEXT:    br label [[DOTOMP_OUTLINED__EXIT]]
4149 // CHECK9:       .omp_outlined..exit:
4150 // CHECK9-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
4151 // CHECK9-NEXT:    ret i32 0
4152 //
4153 //
4154 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
4155 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5:[0-9]+]] {
4156 // CHECK9-NEXT:  entry:
4157 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4158 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4159 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
4160 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
4161 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
4162 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
4163 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
4164 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4165 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4166 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4167 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4168 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
4169 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
4170 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB17:[0-9]+]])
4171 // 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)
4172 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4173 // CHECK9-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 0
4174 // CHECK9-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 0
4175 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
4176 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4177 // CHECK9-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4178 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4179 // CHECK9:       omp.inner.for.cond:
4180 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4181 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4182 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
4183 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4184 // CHECK9:       omp.inner.for.body:
4185 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4186 // CHECK9-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
4187 // CHECK9-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
4188 // CHECK9-NEXT:    ]
4189 // CHECK9:       .omp.sections.case:
4190 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4191 // CHECK9-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
4192 // CHECK9-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
4193 // CHECK9-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
4194 // CHECK9:       .omp.sections.case.split:
4195 // CHECK9-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
4196 // CHECK9:       .omp.sections.case.cncl:
4197 // CHECK9-NEXT:    br label [[CANCEL_CONT:%.*]]
4198 // CHECK9:       .omp.sections.exit:
4199 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4200 // CHECK9:       omp.inner.for.inc:
4201 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4202 // CHECK9-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
4203 // CHECK9-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4204 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
4205 // CHECK9:       omp.inner.for.end:
4206 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19:[0-9]+]])
4207 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
4208 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
4209 // CHECK9:       cancel.cont:
4210 // CHECK9-NEXT:    ret void
4211 // CHECK9:       cancel.exit:
4212 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19]])
4213 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
4214 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
4215 //
4216 //
4217 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
4218 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5]] {
4219 // CHECK9-NEXT:  entry:
4220 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4221 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4222 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
4223 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
4224 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
4225 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
4226 // CHECK9-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
4227 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4228 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4229 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4230 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4231 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
4232 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
4233 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB21:[0-9]+]])
4234 // 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)
4235 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4236 // CHECK9-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 1
4237 // CHECK9-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 1
4238 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
4239 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4240 // CHECK9-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4241 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4242 // CHECK9:       omp.inner.for.cond:
4243 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4244 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4245 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
4246 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4247 // CHECK9:       omp.inner.for.body:
4248 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4249 // CHECK9-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
4250 // CHECK9-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
4251 // CHECK9-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE2:%.*]]
4252 // CHECK9-NEXT:    ]
4253 // CHECK9:       .omp.sections.case:
4254 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4255 // CHECK9-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
4256 // CHECK9-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
4257 // CHECK9-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
4258 // CHECK9:       .omp.sections.case.split:
4259 // CHECK9-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
4260 // CHECK9:       .omp.sections.case.cncl:
4261 // CHECK9-NEXT:    br label [[CANCEL_CONT:%.*]]
4262 // CHECK9:       .omp.sections.case2:
4263 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4264 // CHECK9-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
4265 // CHECK9-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
4266 // CHECK9-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
4267 // CHECK9:       .omp.sections.case2.split:
4268 // CHECK9-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
4269 // CHECK9:       .omp.sections.case2.cncl:
4270 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_END]]
4271 // CHECK9:       .omp.sections.exit:
4272 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4273 // CHECK9:       omp.inner.for.inc:
4274 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4275 // CHECK9-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
4276 // CHECK9-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4277 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
4278 // CHECK9:       omp.inner.for.end:
4279 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM5:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23:[0-9]+]])
4280 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM5]])
4281 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
4282 // CHECK9:       cancel.cont:
4283 // CHECK9-NEXT:    ret void
4284 // CHECK9:       cancel.exit:
4285 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23]])
4286 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
4287 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
4288 //
4289 //
4290 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
4291 // 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]] {
4292 // CHECK9-NEXT:  entry:
4293 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4294 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4295 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
4296 // CHECK9-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
4297 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4298 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4299 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4300 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4301 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4302 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4303 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4304 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4305 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4306 // CHECK9-NEXT:    [[R3:%.*]] = alloca i32, align 4
4307 // CHECK9-NEXT:    [[I4:%.*]] = alloca i32, align 4
4308 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
4309 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4310 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4311 // CHECK9-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
4312 // CHECK9-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
4313 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
4314 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
4315 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
4316 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
4317 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4318 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
4319 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4320 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
4321 // CHECK9-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4322 // CHECK9-NEXT:    store i32 0, i32* [[I]], align 4
4323 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4324 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
4325 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4326 // CHECK9:       omp.precond.then:
4327 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4328 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4329 // CHECK9-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
4330 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4331 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4332 // CHECK9-NEXT:    store i32 0, i32* [[R3]], align 4
4333 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB25:[0-9]+]])
4334 // 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)
4335 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4336 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4337 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
4338 // CHECK9-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4339 // CHECK9:       cond.true:
4340 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4341 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4342 // CHECK9:       cond.false:
4343 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4344 // CHECK9-NEXT:    br label [[COND_END]]
4345 // CHECK9:       cond.end:
4346 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
4347 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4348 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4349 // CHECK9-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
4350 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4351 // CHECK9:       omp.inner.for.cond:
4352 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4353 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4354 // CHECK9-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
4355 // CHECK9-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4356 // CHECK9:       omp.inner.for.body:
4357 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4358 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1
4359 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4360 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
4361 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM7:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB27:[0-9]+]])
4362 // CHECK9-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM7]], i32 2)
4363 // CHECK9-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
4364 // CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
4365 // CHECK9:       .cancel.exit:
4366 // CHECK9-NEXT:    br label [[CANCEL_EXIT:%.*]]
4367 // CHECK9:       .cancel.continue:
4368 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I4]], align 4
4369 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[R3]], align 4
4370 // CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP16]]
4371 // CHECK9-NEXT:    store i32 [[ADD8]], i32* [[R3]], align 4
4372 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4373 // CHECK9:       omp.body.continue:
4374 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4375 // CHECK9:       omp.inner.for.inc:
4376 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4377 // CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], 1
4378 // CHECK9-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
4379 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
4380 // CHECK9:       omp.inner.for.end:
4381 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4382 // CHECK9:       omp.loop.exit:
4383 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29:[0-9]+]])
4384 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
4385 // CHECK9-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
4386 // CHECK9-NEXT:    [[TMP20:%.*]] = bitcast i32* [[R3]] to i8*
4387 // CHECK9-NEXT:    store i8* [[TMP20]], i8** [[TMP19]], align 8
4388 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
4389 // CHECK9-NEXT:    [[TMP21:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
4390 // 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)
4391 // CHECK9-NEXT:    switch i32 [[TMP22]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
4392 // CHECK9-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
4393 // CHECK9-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
4394 // CHECK9-NEXT:    ]
4395 // CHECK9:       .omp.reduction.case1:
4396 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP1]], align 4
4397 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[R3]], align 4
4398 // CHECK9-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
4399 // CHECK9-NEXT:    store i32 [[ADD13]], i32* [[TMP1]], align 4
4400 // 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)
4401 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
4402 // CHECK9:       cancel.exit:
4403 // CHECK9-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
4404 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
4405 // CHECK9-NEXT:    br label [[CANCEL_CONT:%.*]]
4406 // CHECK9:       .omp.reduction.case2:
4407 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, i32* [[R3]], align 4
4408 // CHECK9-NEXT:    [[TMP26:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP25]] monotonic, align 4
4409 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
4410 // CHECK9:       .omp.reduction.default:
4411 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
4412 // CHECK9:       omp.precond.end:
4413 // CHECK9-NEXT:    br label [[CANCEL_CONT]]
4414 // CHECK9:       cancel.cont:
4415 // CHECK9-NEXT:    ret void
4416 //
4417 //
4418 // CHECK9-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
4419 // CHECK9-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
4420 // CHECK9-NEXT:  entry:
4421 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
4422 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
4423 // CHECK9-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
4424 // CHECK9-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
4425 // CHECK9-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
4426 // CHECK9-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
4427 // CHECK9-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
4428 // CHECK9-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
4429 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
4430 // CHECK9-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
4431 // CHECK9-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
4432 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
4433 // CHECK9-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
4434 // CHECK9-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
4435 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
4436 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
4437 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4438 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
4439 // CHECK9-NEXT:    ret void
4440 //
4441 //
4442 // CHECK10-LABEL: define {{[^@]+}}@main
4443 // CHECK10-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
4444 // CHECK10-NEXT:  entry:
4445 // CHECK10-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
4446 // CHECK10-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
4447 // CHECK10-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
4448 // CHECK10-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
4449 // CHECK10-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
4450 // CHECK10-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
4451 // CHECK10-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
4452 // CHECK10-NEXT:    [[P_LASTITER27:%.*]] = alloca i32, align 4
4453 // CHECK10-NEXT:    [[P_LOWERBOUND28:%.*]] = alloca i32, align 4
4454 // CHECK10-NEXT:    [[P_UPPERBOUND29:%.*]] = alloca i32, align 4
4455 // CHECK10-NEXT:    [[P_STRIDE30:%.*]] = alloca i32, align 4
4456 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4457 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4458 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4459 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
4460 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
4461 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4462 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4463 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4464 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4465 // CHECK10-NEXT:    [[I36:%.*]] = alloca i32, align 4
4466 // CHECK10-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
4467 // CHECK10-NEXT:    [[R:%.*]] = alloca i32, align 4
4468 // CHECK10-NEXT:    store i32 0, i32* [[RETVAL]], align 4
4469 // CHECK10-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4470 // CHECK10-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
4471 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
4472 // CHECK10-NEXT:    br label [[OMP_PARALLEL:%.*]]
4473 // CHECK10:       omp_parallel:
4474 // 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]])
4475 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
4476 // CHECK10:       omp.par.outlined.exit:
4477 // CHECK10-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
4478 // CHECK10:       omp.par.exit.split:
4479 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
4480 // CHECK10:       omp_section_loop.preheader:
4481 // CHECK10-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
4482 // CHECK10-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
4483 // CHECK10-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
4484 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4485 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]], i32 34, i32* [[P_LASTITER]], i32* [[P_LOWERBOUND]], i32* [[P_UPPERBOUND]], i32* [[P_STRIDE]], i32 1, i32 1)
4486 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
4487 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
4488 // CHECK10-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
4489 // CHECK10-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
4490 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
4491 // CHECK10:       omp_section_loop.header:
4492 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
4493 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
4494 // CHECK10:       omp_section_loop.cond:
4495 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
4496 // CHECK10-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
4497 // CHECK10:       omp_section_loop.body:
4498 // CHECK10-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
4499 // CHECK10-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
4500 // CHECK10-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
4501 // CHECK10-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_INC]] [
4502 // CHECK10-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
4503 // CHECK10-NEXT:    ]
4504 // CHECK10:       omp_section_loop.inc:
4505 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
4506 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
4507 // CHECK10:       omp_section_loop.exit:
4508 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
4509 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4510 // CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
4511 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
4512 // CHECK10:       omp_section_loop.after:
4513 // CHECK10-NEXT:    br label [[OMP_SECTIONS_END:%.*]]
4514 // CHECK10:       omp_sections.end:
4515 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
4516 // CHECK10:       omp_section_loop.preheader13:
4517 // CHECK10-NEXT:    store i32 0, i32* [[P_LOWERBOUND28]], align 4
4518 // CHECK10-NEXT:    store i32 1, i32* [[P_UPPERBOUND29]], align 4
4519 // CHECK10-NEXT:    store i32 1, i32* [[P_STRIDE30]], align 4
4520 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM31:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4521 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]], i32 34, i32* [[P_LASTITER27]], i32* [[P_LOWERBOUND28]], i32* [[P_UPPERBOUND29]], i32* [[P_STRIDE30]], i32 1, i32 1)
4522 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[P_LOWERBOUND28]], align 4
4523 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[P_UPPERBOUND29]], align 4
4524 // CHECK10-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], [[TMP7]]
4525 // CHECK10-NEXT:    [[TMP10:%.*]] = add i32 [[TMP9]], 1
4526 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
4527 // CHECK10:       omp_section_loop.header14:
4528 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
4529 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
4530 // CHECK10:       omp_section_loop.cond15:
4531 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP10]]
4532 // CHECK10-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
4533 // CHECK10:       omp_section_loop.body16:
4534 // CHECK10-NEXT:    [[TMP11:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP7]]
4535 // CHECK10-NEXT:    [[TMP12:%.*]] = mul i32 [[TMP11]], 1
4536 // CHECK10-NEXT:    [[TMP13:%.*]] = add i32 [[TMP12]], 0
4537 // CHECK10-NEXT:    switch i32 [[TMP13]], label [[OMP_SECTION_LOOP_INC17]] [
4538 // CHECK10-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
4539 // CHECK10-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
4540 // CHECK10-NEXT:    ]
4541 // CHECK10:       omp_section_loop.inc17:
4542 // CHECK10-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
4543 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
4544 // CHECK10:       omp_section_loop.exit18:
4545 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM31]])
4546 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4547 // CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
4548 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
4549 // CHECK10:       omp_section_loop.after19:
4550 // CHECK10-NEXT:    br label [[OMP_SECTIONS_END33:%.*]]
4551 // CHECK10:       omp_sections.end33:
4552 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4553 // CHECK10-NEXT:    store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_]], align 4
4554 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4555 // CHECK10-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP15]], 0
4556 // CHECK10-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4557 // CHECK10-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
4558 // CHECK10-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
4559 // CHECK10-NEXT:    store i32 0, i32* [[I]], align 4
4560 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4561 // CHECK10-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP16]]
4562 // CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4563 // CHECK10:       omp.precond.then:
4564 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4565 // CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
4566 // CHECK10-NEXT:    store i32 [[TMP17]], i32* [[DOTOMP_UB]], align 4
4567 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4568 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4569 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
4570 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM37]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4571 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4572 // CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
4573 // CHECK10-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP18]], [[TMP19]]
4574 // CHECK10-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4575 // CHECK10:       cond.true:
4576 // CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
4577 // CHECK10-NEXT:    br label [[COND_END:%.*]]
4578 // CHECK10:       cond.false:
4579 // CHECK10-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4580 // CHECK10-NEXT:    br label [[COND_END]]
4581 // CHECK10:       cond.end:
4582 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP20]], [[COND_TRUE]] ], [ [[TMP21]], [[COND_FALSE]] ]
4583 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4584 // CHECK10-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4585 // CHECK10-NEXT:    store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
4586 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4587 // CHECK10:       omp.inner.for.cond:
4588 // CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4589 // CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4590 // CHECK10-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP23]], [[TMP24]]
4591 // CHECK10-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4592 // CHECK10:       omp.inner.for.body:
4593 // CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4594 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP25]], 1
4595 // CHECK10-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
4596 // CHECK10-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
4597 // CHECK10-NEXT:    [[TMP26:%.*]] = load float, float* @flag, align 4
4598 // CHECK10-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
4599 // CHECK10-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4600 // CHECK10:       omp_if.then:
4601 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
4602 // CHECK10-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
4603 // CHECK10-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
4604 // CHECK10-NEXT:    br i1 [[TMP28]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
4605 // CHECK10:       .cancel.exit:
4606 // CHECK10-NEXT:    br label [[CANCEL_EXIT:%.*]]
4607 // CHECK10:       omp_section_loop.body.case:
4608 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4609 // CHECK10-NEXT:    [[TMP29:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
4610 // CHECK10-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP29]], 0
4611 // CHECK10-NEXT:    br i1 [[TMP30]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
4612 // CHECK10:       omp_section_loop.body.case.split:
4613 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
4614 // CHECK10:       omp_section_loop.body.case.cncl:
4615 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
4616 // CHECK10:       omp_section_loop.body.case23:
4617 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4618 // CHECK10-NEXT:    [[TMP31:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
4619 // CHECK10-NEXT:    [[TMP32:%.*]] = icmp eq i32 [[TMP31]], 0
4620 // CHECK10-NEXT:    br i1 [[TMP32]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
4621 // CHECK10:       omp_section_loop.body.case23.split:
4622 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
4623 // CHECK10:       omp_section_loop.body.case23.cncl:
4624 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
4625 // CHECK10:       omp_section_loop.body.case25:
4626 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM26:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4627 // CHECK10-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM26]], i32 3)
4628 // CHECK10-NEXT:    [[TMP34:%.*]] = icmp eq i32 [[TMP33]], 0
4629 // CHECK10-NEXT:    br i1 [[TMP34]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
4630 // CHECK10:       omp_section_loop.body.case25.split:
4631 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
4632 // CHECK10:       omp_section_loop.body.case25.cncl:
4633 // CHECK10-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
4634 // CHECK10:       .cancel.continue:
4635 // CHECK10-NEXT:    br label [[OMP_IF_END:%.*]]
4636 // CHECK10:       omp_if.else:
4637 // CHECK10-NEXT:    br label [[OMP_IF_END]]
4638 // CHECK10:       omp_if.end:
4639 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4640 // CHECK10:       omp.body.continue:
4641 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4642 // CHECK10:       omp.inner.for.inc:
4643 // CHECK10-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4644 // CHECK10-NEXT:    [[ADD43:%.*]] = add nsw i32 [[TMP35]], 1
4645 // CHECK10-NEXT:    store i32 [[ADD43]], i32* [[DOTOMP_IV]], align 4
4646 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
4647 // CHECK10:       omp.inner.for.end:
4648 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4649 // CHECK10:       omp.loop.exit:
4650 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
4651 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
4652 // CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
4653 // CHECK10:       cancel.exit:
4654 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
4655 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
4656 // CHECK10-NEXT:    br label [[CANCEL_CONT:%.*]]
4657 // CHECK10:       omp.precond.end:
4658 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
4659 // CHECK10:       cancel.cont:
4660 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4661 // CHECK10-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM46]])
4662 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
4663 // CHECK10-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM47]], 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*)*))
4664 // CHECK10-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
4665 // CHECK10-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
4666 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM48:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
4667 // CHECK10-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM48]], i8* [[TMP36]])
4668 // 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*, ...)*))
4669 // 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*, ...)*))
4670 // CHECK10-NEXT:    store i32 0, i32* [[R]], align 4
4671 // 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]])
4672 // CHECK10-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4673 // CHECK10-NEXT:    ret i32 [[TMP40]]
4674 //
4675 //
4676 // CHECK10-LABEL: define {{[^@]+}}@main..omp_par
4677 // CHECK10-SAME: (i32* noalias [[TID_ADDR:%.*]], i32* noalias [[ZERO_ADDR:%.*]], i32* [[ARGC_ADDR:%.*]], i8*** [[ARGV_ADDR:%.*]]) #[[ATTR1:[0-9]+]] {
4678 // CHECK10-NEXT:  omp.par.entry:
4679 // CHECK10-NEXT:    [[TID_ADDR_LOCAL:%.*]] = alloca i32, align 4
4680 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[TID_ADDR]], align 4
4681 // CHECK10-NEXT:    store i32 [[TMP0]], i32* [[TID_ADDR_LOCAL]], align 4
4682 // CHECK10-NEXT:    [[TID:%.*]] = load i32, i32* [[TID_ADDR_LOCAL]], align 4
4683 // CHECK10-NEXT:    br label [[OMP_PAR_REGION:%.*]]
4684 // CHECK10:       omp.par.region:
4685 // CHECK10-NEXT:    [[TMP1:%.*]] = load float, float* @flag, align 4
4686 // CHECK10-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP1]], 0.000000e+00
4687 // CHECK10-NEXT:    br i1 [[TOBOOL]], label [[TMP13:%.*]], label [[TMP2:%.*]]
4688 // CHECK10:       2:
4689 // CHECK10-NEXT:    br label [[TMP3:%.*]]
4690 // CHECK10:       3:
4691 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4692 // CHECK10-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP4]] to i8
4693 // CHECK10-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
4694 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 0
4695 // CHECK10-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
4696 // CHECK10-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 0
4697 // CHECK10-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX3]], align 1
4698 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4699 // CHECK10-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
4700 // CHECK10-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
4701 // CHECK10-NEXT:    br i1 [[TMP8]], label [[DOTCONT:%.*]], label [[DOTCNCL5:%.*]]
4702 // CHECK10:       .cncl5:
4703 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
4704 // CHECK10:       .cont:
4705 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4706 // CHECK10-NEXT:    [[TMP10:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
4707 // CHECK10-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP10]], i64 0
4708 // CHECK10-NEXT:    [[TMP11:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
4709 // CHECK10-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP11]], i64 0
4710 // CHECK10-NEXT:    [[TMP12:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
4711 // CHECK10-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP12]] to i32
4712 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP9]]
4713 // CHECK10-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
4714 // CHECK10-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
4715 // CHECK10-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
4716 // CHECK10:       omp.par.pre_finalize:
4717 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
4718 // CHECK10:       13:
4719 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4720 // CHECK10-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
4721 // CHECK10-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[TMP14]], 0
4722 // CHECK10-NEXT:    br i1 [[TMP15]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
4723 // CHECK10:       .cncl:
4724 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4725 // CHECK10-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
4726 // CHECK10-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
4727 // CHECK10:       .split:
4728 // CHECK10-NEXT:    br label [[TMP3]]
4729 // CHECK10:       omp.par.outlined.exit.exitStub:
4730 // CHECK10-NEXT:    ret void
4731 //
4732 //
4733 // CHECK10-LABEL: define {{[^@]+}}@.omp_task_entry.
4734 // CHECK10-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
4735 // CHECK10-NEXT:  entry:
4736 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
4737 // CHECK10-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
4738 // CHECK10-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
4739 // CHECK10-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
4740 // CHECK10-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
4741 // CHECK10-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
4742 // CHECK10-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
4743 // CHECK10-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
4744 // CHECK10-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
4745 // CHECK10-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
4746 // CHECK10-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
4747 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
4748 // CHECK10-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
4749 // CHECK10-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
4750 // CHECK10-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
4751 // CHECK10-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
4752 // CHECK10-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
4753 // CHECK10-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
4754 // CHECK10-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
4755 // CHECK10-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META5:![0-9]+]])
4756 // CHECK10-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
4757 // CHECK10-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
4758 // CHECK10-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
4759 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
4760 // CHECK10-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !14
4761 // CHECK10-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !14
4762 // CHECK10-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !14
4763 // CHECK10-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !14
4764 // CHECK10-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
4765 // CHECK10-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
4766 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM_I:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]]) #[[ATTR2:[0-9]+]]
4767 // CHECK10-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM_I]], i32 4) #[[ATTR2]]
4768 // CHECK10-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4769 // CHECK10-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
4770 // CHECK10:       .cancel.exit.i:
4771 // CHECK10-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
4772 // CHECK10-NEXT:    br label [[DOTOMP_OUTLINED__EXIT:%.*]]
4773 // CHECK10:       .cancel.continue.i:
4774 // CHECK10-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
4775 // CHECK10-NEXT:    br label [[DOTOMP_OUTLINED__EXIT]]
4776 // CHECK10:       .omp_outlined..exit:
4777 // CHECK10-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
4778 // CHECK10-NEXT:    ret i32 0
4779 //
4780 //
4781 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
4782 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5:[0-9]+]] {
4783 // CHECK10-NEXT:  entry:
4784 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4785 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4786 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
4787 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
4788 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
4789 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
4790 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
4791 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4792 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4793 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4794 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4795 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
4796 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
4797 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB17:[0-9]+]])
4798 // 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)
4799 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4800 // CHECK10-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 0
4801 // CHECK10-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 0
4802 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
4803 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4804 // CHECK10-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4805 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4806 // CHECK10:       omp.inner.for.cond:
4807 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4808 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4809 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
4810 // CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4811 // CHECK10:       omp.inner.for.body:
4812 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4813 // CHECK10-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
4814 // CHECK10-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
4815 // CHECK10-NEXT:    ]
4816 // CHECK10:       .omp.sections.case:
4817 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4818 // CHECK10-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
4819 // CHECK10-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
4820 // CHECK10-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
4821 // CHECK10:       .omp.sections.case.split:
4822 // CHECK10-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
4823 // CHECK10:       .omp.sections.case.cncl:
4824 // CHECK10-NEXT:    br label [[CANCEL_CONT:%.*]]
4825 // CHECK10:       .omp.sections.exit:
4826 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4827 // CHECK10:       omp.inner.for.inc:
4828 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4829 // CHECK10-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
4830 // CHECK10-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4831 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
4832 // CHECK10:       omp.inner.for.end:
4833 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19:[0-9]+]])
4834 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
4835 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
4836 // CHECK10:       cancel.cont:
4837 // CHECK10-NEXT:    ret void
4838 // CHECK10:       cancel.exit:
4839 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19]])
4840 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
4841 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
4842 //
4843 //
4844 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
4845 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR5]] {
4846 // CHECK10-NEXT:  entry:
4847 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4848 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4849 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
4850 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
4851 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
4852 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
4853 // CHECK10-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
4854 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4855 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4856 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4857 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4858 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
4859 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
4860 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB21:[0-9]+]])
4861 // 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)
4862 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4863 // CHECK10-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 1
4864 // CHECK10-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 1
4865 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
4866 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
4867 // CHECK10-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4868 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4869 // CHECK10:       omp.inner.for.cond:
4870 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4871 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
4872 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
4873 // CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4874 // CHECK10:       omp.inner.for.body:
4875 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4876 // CHECK10-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
4877 // CHECK10-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
4878 // CHECK10-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE2:%.*]]
4879 // CHECK10-NEXT:    ]
4880 // CHECK10:       .omp.sections.case:
4881 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4882 // CHECK10-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
4883 // CHECK10-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
4884 // CHECK10-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
4885 // CHECK10:       .omp.sections.case.split:
4886 // CHECK10-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
4887 // CHECK10:       .omp.sections.case.cncl:
4888 // CHECK10-NEXT:    br label [[CANCEL_CONT:%.*]]
4889 // CHECK10:       .omp.sections.case2:
4890 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4891 // CHECK10-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
4892 // CHECK10-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
4893 // CHECK10-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
4894 // CHECK10:       .omp.sections.case2.split:
4895 // CHECK10-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
4896 // CHECK10:       .omp.sections.case2.cncl:
4897 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_END]]
4898 // CHECK10:       .omp.sections.exit:
4899 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4900 // CHECK10:       omp.inner.for.inc:
4901 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
4902 // CHECK10-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
4903 // CHECK10-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
4904 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
4905 // CHECK10:       omp.inner.for.end:
4906 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM5:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23:[0-9]+]])
4907 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM5]])
4908 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
4909 // CHECK10:       cancel.cont:
4910 // CHECK10-NEXT:    ret void
4911 // CHECK10:       cancel.exit:
4912 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23]])
4913 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
4914 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
4915 //
4916 //
4917 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
4918 // 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]] {
4919 // CHECK10-NEXT:  entry:
4920 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4921 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4922 // CHECK10-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
4923 // CHECK10-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
4924 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4925 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4926 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4927 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4928 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
4929 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4930 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4931 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4932 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4933 // CHECK10-NEXT:    [[R3:%.*]] = alloca i32, align 4
4934 // CHECK10-NEXT:    [[I4:%.*]] = alloca i32, align 4
4935 // CHECK10-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
4936 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4937 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4938 // CHECK10-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
4939 // CHECK10-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
4940 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
4941 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
4942 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
4943 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
4944 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4945 // CHECK10-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
4946 // CHECK10-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4947 // CHECK10-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
4948 // CHECK10-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4949 // CHECK10-NEXT:    store i32 0, i32* [[I]], align 4
4950 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4951 // CHECK10-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
4952 // CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4953 // CHECK10:       omp.precond.then:
4954 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4955 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4956 // CHECK10-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
4957 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4958 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4959 // CHECK10-NEXT:    store i32 0, i32* [[R3]], align 4
4960 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB25:[0-9]+]])
4961 // 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)
4962 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4963 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4964 // CHECK10-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
4965 // CHECK10-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4966 // CHECK10:       cond.true:
4967 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4968 // CHECK10-NEXT:    br label [[COND_END:%.*]]
4969 // CHECK10:       cond.false:
4970 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4971 // CHECK10-NEXT:    br label [[COND_END]]
4972 // CHECK10:       cond.end:
4973 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
4974 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4975 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4976 // CHECK10-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
4977 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4978 // CHECK10:       omp.inner.for.cond:
4979 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4980 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4981 // CHECK10-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
4982 // CHECK10-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4983 // CHECK10:       omp.inner.for.body:
4984 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4985 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1
4986 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4987 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
4988 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM7:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB27:[0-9]+]])
4989 // CHECK10-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM7]], i32 2)
4990 // CHECK10-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
4991 // CHECK10-NEXT:    br i1 [[TMP15]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
4992 // CHECK10:       .cancel.exit:
4993 // CHECK10-NEXT:    br label [[CANCEL_EXIT:%.*]]
4994 // CHECK10:       .cancel.continue:
4995 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I4]], align 4
4996 // CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[R3]], align 4
4997 // CHECK10-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP16]]
4998 // CHECK10-NEXT:    store i32 [[ADD8]], i32* [[R3]], align 4
4999 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5000 // CHECK10:       omp.body.continue:
5001 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5002 // CHECK10:       omp.inner.for.inc:
5003 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5004 // CHECK10-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], 1
5005 // CHECK10-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
5006 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
5007 // CHECK10:       omp.inner.for.end:
5008 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5009 // CHECK10:       omp.loop.exit:
5010 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29:[0-9]+]])
5011 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
5012 // CHECK10-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
5013 // CHECK10-NEXT:    [[TMP20:%.*]] = bitcast i32* [[R3]] to i8*
5014 // CHECK10-NEXT:    store i8* [[TMP20]], i8** [[TMP19]], align 8
5015 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
5016 // CHECK10-NEXT:    [[TMP21:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
5017 // 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)
5018 // CHECK10-NEXT:    switch i32 [[TMP22]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
5019 // CHECK10-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
5020 // CHECK10-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
5021 // CHECK10-NEXT:    ]
5022 // CHECK10:       .omp.reduction.case1:
5023 // CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP1]], align 4
5024 // CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[R3]], align 4
5025 // CHECK10-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
5026 // CHECK10-NEXT:    store i32 [[ADD13]], i32* [[TMP1]], align 4
5027 // 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)
5028 // CHECK10-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5029 // CHECK10:       cancel.exit:
5030 // CHECK10-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
5031 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
5032 // CHECK10-NEXT:    br label [[CANCEL_CONT:%.*]]
5033 // CHECK10:       .omp.reduction.case2:
5034 // CHECK10-NEXT:    [[TMP25:%.*]] = load i32, i32* [[R3]], align 4
5035 // CHECK10-NEXT:    [[TMP26:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP25]] monotonic, align 4
5036 // CHECK10-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5037 // CHECK10:       .omp.reduction.default:
5038 // CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
5039 // CHECK10:       omp.precond.end:
5040 // CHECK10-NEXT:    br label [[CANCEL_CONT]]
5041 // CHECK10:       cancel.cont:
5042 // CHECK10-NEXT:    ret void
5043 //
5044 //
5045 // CHECK10-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
5046 // CHECK10-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4]] {
5047 // CHECK10-NEXT:  entry:
5048 // CHECK10-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
5049 // CHECK10-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
5050 // CHECK10-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
5051 // CHECK10-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
5052 // CHECK10-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
5053 // CHECK10-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
5054 // CHECK10-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
5055 // CHECK10-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
5056 // CHECK10-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
5057 // CHECK10-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
5058 // CHECK10-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
5059 // CHECK10-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
5060 // CHECK10-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
5061 // CHECK10-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
5062 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
5063 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
5064 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5065 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
5066 // CHECK10-NEXT:    ret void
5067 //
5068