xref: /llvm-project/clang/test/OpenMP/cancel_codegen.cpp (revision 839ac62c5085d895d3165bc5024db623a7a78813)
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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK1
5 
6 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK3
9 
10 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -verify -fopenmp -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK1
15 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple x86_64-apple-darwin13.4.0 -emit-pch -o %t.3 %s
16 // RUN: %clang_cc1 -no-opaque-pointers -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=CHECK1
17 
18 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-enable-irbuilder -triple x86_64-apple-darwin13.4.0 -emit-llvm -o - %s | FileCheck %s --check-prefix=CHECK3
19 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK3
21 
22 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple x86_64-apple-darwin13.4.0 -emit-pch -o %t.5 %s
24 // RUN: %clang_cc1 -no-opaque-pointers -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 noundef [[ARGC:%.*]], i8** noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i8*** noundef nonnull align 8 dereferenceable(8) [[ARGV:%.*]], i32* noundef 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 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[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)
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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* noundef 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* noundef [[TMP0:%.*]], i8* noundef [[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 // CHECK3-LABEL: define {{[^@]+}}@main
702 // CHECK3-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
703 // CHECK3-NEXT:  entry:
704 // CHECK3-NEXT:    [[STRUCTARG:%.*]] = alloca { i32*, i8*** }, align 8
705 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
706 // CHECK3-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
707 // CHECK3-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
708 // CHECK3-NEXT:    [[P_LASTITER:%.*]] = alloca i32, align 4
709 // CHECK3-NEXT:    [[P_LOWERBOUND:%.*]] = alloca i32, align 4
710 // CHECK3-NEXT:    [[P_UPPERBOUND:%.*]] = alloca i32, align 4
711 // CHECK3-NEXT:    [[P_STRIDE:%.*]] = alloca i32, align 4
712 // CHECK3-NEXT:    [[P_LASTITER28:%.*]] = alloca i32, align 4
713 // CHECK3-NEXT:    [[P_LOWERBOUND29:%.*]] = alloca i32, align 4
714 // CHECK3-NEXT:    [[P_UPPERBOUND30:%.*]] = alloca i32, align 4
715 // CHECK3-NEXT:    [[P_STRIDE31:%.*]] = alloca i32, align 4
716 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
717 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
718 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
719 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_34:%.*]] = alloca i32, align 4
720 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
721 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
722 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
723 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
724 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
725 // CHECK3-NEXT:    [[I36:%.*]] = alloca i32, align 4
726 // CHECK3-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
727 // CHECK3-NEXT:    [[R:%.*]] = alloca i32, align 4
728 // CHECK3-NEXT:    store i32 0, i32* [[RETVAL]], align 4
729 // CHECK3-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
730 // CHECK3-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
731 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
732 // CHECK3-NEXT:    br label [[OMP_PARALLEL:%.*]]
733 // CHECK3:       omp_parallel:
734 // CHECK3-NEXT:    [[GEP_ARGC_ADDR:%.*]] = getelementptr { i32*, i8*** }, { i32*, i8*** }* [[STRUCTARG]], i32 0, i32 0
735 // CHECK3-NEXT:    store i32* [[ARGC_ADDR]], i32** [[GEP_ARGC_ADDR]], align 8
736 // CHECK3-NEXT:    [[GEP_ARGV_ADDR:%.*]] = getelementptr { i32*, i8*** }, { i32*, i8*** }* [[STRUCTARG]], i32 0, i32 1
737 // CHECK3-NEXT:    store i8*** [[ARGV_ADDR]], i8**** [[GEP_ARGV_ADDR]], align 8
738 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, { i32*, i8*** }*)* @main..omp_par to void (i32*, i32*, ...)*), { i32*, i8*** }* [[STRUCTARG]])
739 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT:%.*]]
740 // CHECK3:       omp.par.outlined.exit:
741 // CHECK3-NEXT:    br label [[OMP_PAR_EXIT_SPLIT:%.*]]
742 // CHECK3:       omp.par.exit.split:
743 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER:%.*]]
744 // CHECK3:       omp_section_loop.preheader:
745 // CHECK3-NEXT:    store i32 0, i32* [[P_LOWERBOUND]], align 4
746 // CHECK3-NEXT:    store i32 0, i32* [[P_UPPERBOUND]], align 4
747 // CHECK3-NEXT:    store i32 1, i32* [[P_STRIDE]], align 4
748 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
749 // 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 0)
750 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[P_LOWERBOUND]], align 4
751 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[P_UPPERBOUND]], align 4
752 // CHECK3-NEXT:    [[TMP2:%.*]] = sub i32 [[TMP1]], [[TMP0]]
753 // CHECK3-NEXT:    [[TMP3:%.*]] = add i32 [[TMP2]], 1
754 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER:%.*]]
755 // CHECK3:       omp_section_loop.header:
756 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER]] ], [ [[OMP_SECTION_LOOP_NEXT:%.*]], [[OMP_SECTION_LOOP_INC:%.*]] ]
757 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_COND:%.*]]
758 // CHECK3:       omp_section_loop.cond:
759 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV]], [[TMP3]]
760 // CHECK3-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP]], label [[OMP_SECTION_LOOP_BODY:%.*]], label [[OMP_SECTION_LOOP_EXIT:%.*]]
761 // CHECK3:       omp_section_loop.body:
762 // CHECK3-NEXT:    [[TMP4:%.*]] = add i32 [[OMP_SECTION_LOOP_IV]], [[TMP0]]
763 // CHECK3-NEXT:    [[TMP5:%.*]] = mul i32 [[TMP4]], 1
764 // CHECK3-NEXT:    [[TMP6:%.*]] = add i32 [[TMP5]], 0
765 // CHECK3-NEXT:    switch i32 [[TMP6]], label [[OMP_SECTION_LOOP_BODY_SECTIONS_AFTER:%.*]] [
766 // CHECK3-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE:%.*]]
767 // CHECK3-NEXT:    ]
768 // CHECK3:       omp_section_loop.body.case:
769 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
770 // CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM10]], i32 3)
771 // CHECK3-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
772 // CHECK3-NEXT:    br i1 [[TMP8]], label [[OMP_SECTION_LOOP_BODY_CASE_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE_CNCL:%.*]]
773 // CHECK3:       omp_section_loop.body.case.split:
774 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE_SECTION_AFTER:%.*]]
775 // CHECK3:       omp_section_loop.body.case.section.after:
776 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY_SECTIONS_AFTER]]
777 // CHECK3:       omp_section_loop.body.sections.after:
778 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_INC]]
779 // CHECK3:       omp_section_loop.inc:
780 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_NEXT]] = add nuw i32 [[OMP_SECTION_LOOP_IV]], 1
781 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER]]
782 // CHECK3:       omp_section_loop.exit:
783 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
784 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
785 // CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM12]])
786 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER:%.*]]
787 // CHECK3:       omp_section_loop.after:
788 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTERSECTIONS_FINI:%.*]]
789 // CHECK3:       omp_section_loop.aftersections.fini:
790 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_PREHEADER13:%.*]]
791 // CHECK3:       omp_section_loop.preheader13:
792 // CHECK3-NEXT:    store i32 0, i32* [[P_LOWERBOUND29]], align 4
793 // CHECK3-NEXT:    store i32 1, i32* [[P_UPPERBOUND30]], align 4
794 // CHECK3-NEXT:    store i32 1, i32* [[P_STRIDE31]], align 4
795 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM32:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
796 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM32]], i32 34, i32* [[P_LASTITER28]], i32* [[P_LOWERBOUND29]], i32* [[P_UPPERBOUND30]], i32* [[P_STRIDE31]], i32 1, i32 0)
797 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[P_LOWERBOUND29]], align 4
798 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[P_UPPERBOUND30]], align 4
799 // CHECK3-NEXT:    [[TMP11:%.*]] = sub i32 [[TMP10]], [[TMP9]]
800 // CHECK3-NEXT:    [[TMP12:%.*]] = add i32 [[TMP11]], 1
801 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14:%.*]]
802 // CHECK3:       omp_section_loop.header14:
803 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_IV20:%.*]] = phi i32 [ 0, [[OMP_SECTION_LOOP_PREHEADER13]] ], [ [[OMP_SECTION_LOOP_NEXT22:%.*]], [[OMP_SECTION_LOOP_INC17:%.*]] ]
804 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_COND15:%.*]]
805 // CHECK3:       omp_section_loop.cond15:
806 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_CMP21:%.*]] = icmp ult i32 [[OMP_SECTION_LOOP_IV20]], [[TMP12]]
807 // CHECK3-NEXT:    br i1 [[OMP_SECTION_LOOP_CMP21]], label [[OMP_SECTION_LOOP_BODY16:%.*]], label [[OMP_SECTION_LOOP_EXIT18:%.*]]
808 // CHECK3:       omp_section_loop.body16:
809 // CHECK3-NEXT:    [[TMP13:%.*]] = add i32 [[OMP_SECTION_LOOP_IV20]], [[TMP9]]
810 // CHECK3-NEXT:    [[TMP14:%.*]] = mul i32 [[TMP13]], 1
811 // CHECK3-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], 0
812 // CHECK3-NEXT:    switch i32 [[TMP15]], label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER:%.*]] [
813 // CHECK3-NEXT:    i32 0, label [[OMP_SECTION_LOOP_BODY_CASE23:%.*]]
814 // CHECK3-NEXT:    i32 1, label [[OMP_SECTION_LOOP_BODY_CASE25:%.*]]
815 // CHECK3-NEXT:    ]
816 // CHECK3:       omp_section_loop.body.case23:
817 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM24:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
818 // CHECK3-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM24]], i32 3)
819 // CHECK3-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[TMP16]], 0
820 // CHECK3-NEXT:    br i1 [[TMP17]], label [[OMP_SECTION_LOOP_BODY_CASE23_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE23_CNCL:%.*]]
821 // CHECK3:       omp_section_loop.body.case23.split:
822 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE23_SECTION_AFTER:%.*]]
823 // CHECK3:       omp_section_loop.body.case23.section.after:
824 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER]]
825 // CHECK3:       omp_section_loop.body.case25:
826 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM27:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
827 // CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM27]], i32 3)
828 // CHECK3-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[TMP18]], 0
829 // CHECK3-NEXT:    br i1 [[TMP19]], label [[OMP_SECTION_LOOP_BODY_CASE25_SPLIT:%.*]], label [[OMP_SECTION_LOOP_BODY_CASE25_CNCL:%.*]]
830 // CHECK3:       omp_section_loop.body.case25.split:
831 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE25_SECTION_AFTER26:%.*]]
832 // CHECK3:       omp_section_loop.body.case25.section.after26:
833 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY_CASE25_SECTION_AFTER:%.*]]
834 // CHECK3:       omp_section_loop.body.case25.section.after:
835 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_BODY16_SECTIONS_AFTER]]
836 // CHECK3:       omp_section_loop.body16.sections.after:
837 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_INC17]]
838 // CHECK3:       omp_section_loop.inc17:
839 // CHECK3-NEXT:    [[OMP_SECTION_LOOP_NEXT22]] = add nuw i32 [[OMP_SECTION_LOOP_IV20]], 1
840 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_HEADER14]]
841 // CHECK3:       omp_section_loop.exit18:
842 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM32]])
843 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM33:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
844 // CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM33]])
845 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19:%.*]]
846 // CHECK3:       omp_section_loop.after19:
847 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_AFTER19SECTIONS_FINI:%.*]]
848 // CHECK3:       omp_section_loop.after19sections.fini:
849 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
850 // CHECK3-NEXT:    store i32 [[TMP20]], i32* [[DOTCAPTURE_EXPR_]], align 4
851 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
852 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP21]], 0
853 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
854 // CHECK3-NEXT:    [[SUB35:%.*]] = sub nsw i32 [[DIV]], 1
855 // CHECK3-NEXT:    store i32 [[SUB35]], i32* [[DOTCAPTURE_EXPR_34]], align 4
856 // CHECK3-NEXT:    store i32 0, i32* [[I]], align 4
857 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
858 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP22]]
859 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
860 // CHECK3:       omp.precond.then:
861 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
862 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
863 // CHECK3-NEXT:    store i32 [[TMP23]], i32* [[DOTOMP_UB]], align 4
864 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
865 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
866 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM37:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB6:[0-9]+]])
867 // 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)
868 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
869 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
870 // CHECK3-NEXT:    [[CMP38:%.*]] = icmp sgt i32 [[TMP24]], [[TMP25]]
871 // CHECK3-NEXT:    br i1 [[CMP38]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
872 // CHECK3:       cond.true:
873 // CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_34]], align 4
874 // CHECK3-NEXT:    br label [[COND_END:%.*]]
875 // CHECK3:       cond.false:
876 // CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
877 // CHECK3-NEXT:    br label [[COND_END]]
878 // CHECK3:       cond.end:
879 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP26]], [[COND_TRUE]] ], [ [[TMP27]], [[COND_FALSE]] ]
880 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
881 // CHECK3-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
882 // CHECK3-NEXT:    store i32 [[TMP28]], i32* [[DOTOMP_IV]], align 4
883 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
884 // CHECK3:       omp.inner.for.cond:
885 // CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
886 // CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
887 // CHECK3-NEXT:    [[CMP39:%.*]] = icmp sle i32 [[TMP29]], [[TMP30]]
888 // CHECK3-NEXT:    br i1 [[CMP39]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
889 // CHECK3:       omp.inner.for.body:
890 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
891 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP31]], 1
892 // CHECK3-NEXT:    [[ADD40:%.*]] = add nsw i32 0, [[MUL]]
893 // CHECK3-NEXT:    store i32 [[ADD40]], i32* [[I36]], align 4
894 // CHECK3-NEXT:    [[TMP32:%.*]] = load float, float* @flag, align 4
895 // CHECK3-NEXT:    [[TOBOOL41:%.*]] = fcmp une float [[TMP32]], 0.000000e+00
896 // CHECK3-NEXT:    br i1 [[TOBOOL41]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
897 // CHECK3:       omp_if.then:
898 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM42:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB8:[0-9]+]])
899 // CHECK3-NEXT:    [[TMP33:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM42]], i32 2)
900 // CHECK3-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
901 // CHECK3-NEXT:    br i1 [[TMP34]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
902 // CHECK3:       .cancel.exit:
903 // CHECK3-NEXT:    br label [[CANCEL_EXIT:%.*]]
904 // CHECK3:       omp_section_loop.body.case.cncl:
905 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT]]
906 // CHECK3:       omp_section_loop.body.case23.cncl:
907 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
908 // CHECK3:       omp_section_loop.body.case25.cncl:
909 // CHECK3-NEXT:    br label [[OMP_SECTION_LOOP_EXIT18]]
910 // CHECK3:       .cancel.continue:
911 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
912 // CHECK3:       omp_if.else:
913 // CHECK3-NEXT:    br label [[OMP_IF_END]]
914 // CHECK3:       omp_if.end:
915 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
916 // CHECK3:       omp.body.continue:
917 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
918 // CHECK3:       omp.inner.for.inc:
919 // CHECK3-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
920 // CHECK3-NEXT:    [[ADD43:%.*]] = add nsw i32 [[TMP35]], 1
921 // CHECK3-NEXT:    store i32 [[ADD43]], i32* [[DOTOMP_IV]], align 4
922 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
923 // CHECK3:       omp.inner.for.end:
924 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
925 // CHECK3:       omp.loop.exit:
926 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM45:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10:[0-9]+]])
927 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM45]])
928 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
929 // CHECK3:       cancel.exit:
930 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM44:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB10]])
931 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM44]])
932 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
933 // CHECK3:       omp.precond.end:
934 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
935 // CHECK3:       cancel.cont:
936 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM46:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
937 // CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM46]])
938 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM47:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14:[0-9]+]])
939 // 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*)*))
940 // CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
941 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
942 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM48:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB14]])
943 // CHECK3-NEXT:    [[TMP39:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM48]], i8* [[TMP36]])
944 // 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*, ...)*))
945 // 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*, ...)*))
946 // CHECK3-NEXT:    store i32 0, i32* [[R]], align 4
947 // 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]])
948 // CHECK3-NEXT:    [[TMP40:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
949 // CHECK3-NEXT:    ret i32 [[TMP40]]
950 //
951 //
952 // CHECK3-LABEL: define {{[^@]+}}@main..omp_par
953 // CHECK3-SAME: (i32* noalias [[TID_ADDR:%.*]], i32* noalias [[ZERO_ADDR:%.*]], { i32*, i8*** }* [[TMP0:%.*]]) #[[ATTR1:[0-9]+]] {
954 // CHECK3-NEXT:  omp.par.entry:
955 // CHECK3-NEXT:    [[GEP_ARGC_ADDR:%.*]] = getelementptr { i32*, i8*** }, { i32*, i8*** }* [[TMP0]], i32 0, i32 0
956 // CHECK3-NEXT:    [[LOADGEP_ARGC_ADDR:%.*]] = load i32*, i32** [[GEP_ARGC_ADDR]], align 8
957 // CHECK3-NEXT:    [[GEP_ARGV_ADDR:%.*]] = getelementptr { i32*, i8*** }, { i32*, i8*** }* [[TMP0]], i32 0, i32 1
958 // CHECK3-NEXT:    [[LOADGEP_ARGV_ADDR:%.*]] = load i8***, i8**** [[GEP_ARGV_ADDR]], align 8
959 // CHECK3-NEXT:    [[TID_ADDR_LOCAL:%.*]] = alloca i32, align 4
960 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TID_ADDR]], align 4
961 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TID_ADDR_LOCAL]], align 4
962 // CHECK3-NEXT:    [[TID:%.*]] = load i32, i32* [[TID_ADDR_LOCAL]], align 4
963 // CHECK3-NEXT:    br label [[OMP_PAR_REGION:%.*]]
964 // CHECK3:       omp.par.region:
965 // CHECK3-NEXT:    [[TMP2:%.*]] = load float, float* @flag, align 4
966 // CHECK3-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP2]], 0.000000e+00
967 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[TMP14:%.*]], label [[TMP3:%.*]]
968 // CHECK3:       3:
969 // CHECK3-NEXT:    br label [[TMP4:%.*]]
970 // CHECK3:       4:
971 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[LOADGEP_ARGC_ADDR]], align 4
972 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP5]] to i8
973 // CHECK3-NEXT:    [[TMP6:%.*]] = load i8**, i8*** [[LOADGEP_ARGV_ADDR]], align 8
974 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP6]], i64 0
975 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
976 // CHECK3-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP7]], i64 0
977 // CHECK3-NEXT:    store i8 [[CONV]], i8* [[ARRAYIDX3]], align 1
978 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
979 // CHECK3-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
980 // CHECK3-NEXT:    [[TMP9:%.*]] = icmp eq i32 [[TMP8]], 0
981 // CHECK3-NEXT:    br i1 [[TMP9]], label [[DOTCONT:%.*]], label [[DOTCNCL5:%.*]]
982 // CHECK3:       .cncl5:
983 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB:%.*]]
984 // CHECK3:       .cont:
985 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[LOADGEP_ARGC_ADDR]], align 4
986 // CHECK3-NEXT:    [[TMP11:%.*]] = load i8**, i8*** [[LOADGEP_ARGV_ADDR]], align 8
987 // CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP11]], i64 0
988 // CHECK3-NEXT:    [[TMP12:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
989 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP12]], i64 0
990 // CHECK3-NEXT:    [[TMP13:%.*]] = load i8, i8* [[ARRAYIDX7]], align 1
991 // CHECK3-NEXT:    [[CONV8:%.*]] = sext i8 [[TMP13]] to i32
992 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV8]], [[TMP10]]
993 // CHECK3-NEXT:    [[CONV9:%.*]] = trunc i32 [[ADD]] to i8
994 // CHECK3-NEXT:    store i8 [[CONV9]], i8* [[ARRAYIDX7]], align 1
995 // CHECK3-NEXT:    br label [[OMP_PAR_REGION_PARALLEL_AFTER:%.*]]
996 // CHECK3:       omp.par.region.parallel.after:
997 // CHECK3-NEXT:    br label [[OMP_PAR_PRE_FINALIZE:%.*]]
998 // CHECK3:       omp.par.pre_finalize:
999 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
1000 // CHECK3:       14:
1001 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1002 // CHECK3-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 1)
1003 // CHECK3-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[TMP15]], 0
1004 // CHECK3-NEXT:    br i1 [[TMP16]], label [[DOTSPLIT:%.*]], label [[DOTCNCL:%.*]]
1005 // CHECK3:       .cncl:
1006 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1007 // CHECK3-NEXT:    [[TMP17:%.*]] = call i32 @__kmpc_cancel_barrier(%struct.ident_t* @[[GLOB2]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
1008 // CHECK3-NEXT:    br label [[OMP_PAR_OUTLINED_EXIT_EXITSTUB]]
1009 // CHECK3:       .split:
1010 // CHECK3-NEXT:    br label [[TMP4]]
1011 // CHECK3:       omp.par.outlined.exit.exitStub:
1012 // CHECK3-NEXT:    ret void
1013 //
1014 //
1015 // CHECK3-LABEL: define {{[^@]+}}@.omp_task_entry.
1016 // CHECK3-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
1017 // CHECK3-NEXT:  entry:
1018 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
1019 // CHECK3-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
1020 // CHECK3-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
1021 // CHECK3-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
1022 // CHECK3-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
1023 // CHECK3-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
1024 // CHECK3-NEXT:    [[CLEANUP_DEST_SLOT_I:%.*]] = alloca i32, align 4
1025 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
1026 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
1027 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
1028 // CHECK3-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
1029 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
1030 // CHECK3-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
1031 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
1032 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
1033 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
1034 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1035 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
1036 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
1037 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META5:![0-9]+]])
1038 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
1039 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
1040 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
1041 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !14
1042 // CHECK3-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !14
1043 // CHECK3-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !14
1044 // CHECK3-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !14
1045 // CHECK3-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !14
1046 // CHECK3-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
1047 // CHECK3-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !14
1048 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM_I:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB12:[0-9]+]])
1049 // CHECK3-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM_I]], i32 4)
1050 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1051 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTCANCEL_EXIT_I:%.*]], label [[DOTCANCEL_CONTINUE_I:%.*]]
1052 // CHECK3:       .cancel.exit.i:
1053 // CHECK3-NEXT:    store i32 1, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
1054 // CHECK3-NEXT:    br label [[DOTOMP_OUTLINED__EXIT:%.*]]
1055 // CHECK3:       .cancel.continue.i:
1056 // CHECK3-NEXT:    store i32 0, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
1057 // CHECK3-NEXT:    br label [[DOTOMP_OUTLINED__EXIT]]
1058 // CHECK3:       .omp_outlined..exit:
1059 // CHECK3-NEXT:    [[CLEANUP_DEST_I:%.*]] = load i32, i32* [[CLEANUP_DEST_SLOT_I]], align 4, !noalias !14
1060 // CHECK3-NEXT:    ret i32 0
1061 //
1062 //
1063 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
1064 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR5:[0-9]+]] {
1065 // CHECK3-NEXT:  entry:
1066 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1067 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1068 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
1069 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
1070 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
1071 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
1072 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
1073 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1074 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1075 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1076 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1077 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
1078 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
1079 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB17:[0-9]+]])
1080 // 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)
1081 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1082 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 0
1083 // CHECK3-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 0
1084 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
1085 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1086 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1087 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1088 // CHECK3:       omp.inner.for.cond:
1089 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1090 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1091 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
1092 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1093 // CHECK3:       omp.inner.for.body:
1094 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1095 // CHECK3-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
1096 // CHECK3-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
1097 // CHECK3-NEXT:    ]
1098 // CHECK3:       .omp.sections.case:
1099 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1100 // CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
1101 // CHECK3-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
1102 // CHECK3-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
1103 // CHECK3:       .omp.sections.case.split:
1104 // CHECK3-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1105 // CHECK3:       .omp.sections.case.cncl:
1106 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
1107 // CHECK3:       .omp.sections.exit:
1108 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1109 // CHECK3:       omp.inner.for.inc:
1110 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1111 // CHECK3-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP9]], 1
1112 // CHECK3-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1113 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1114 // CHECK3:       omp.inner.for.end:
1115 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19:[0-9]+]])
1116 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM3]])
1117 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1118 // CHECK3:       cancel.cont:
1119 // CHECK3-NEXT:    ret void
1120 // CHECK3:       cancel.exit:
1121 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB19]])
1122 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM2]])
1123 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1124 //
1125 //
1126 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
1127 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR5]] {
1128 // CHECK3-NEXT:  entry:
1129 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1130 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1131 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
1132 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
1133 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
1134 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
1135 // CHECK3-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
1136 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1137 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1138 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1139 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1140 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
1141 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
1142 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB21:[0-9]+]])
1143 // 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)
1144 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1145 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 1
1146 // CHECK3-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 [[TMP0]], i32 1
1147 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
1148 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
1149 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1150 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1151 // CHECK3:       omp.inner.for.cond:
1152 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1153 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
1154 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
1155 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1156 // CHECK3:       omp.inner.for.body:
1157 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1158 // CHECK3-NEXT:    switch i32 [[TMP6]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
1159 // CHECK3-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
1160 // CHECK3-NEXT:    i32 1, label [[DOTOMP_SECTIONS_CASE2:%.*]]
1161 // CHECK3-NEXT:    ]
1162 // CHECK3:       .omp.sections.case:
1163 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1164 // CHECK3-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM1]], i32 3)
1165 // CHECK3-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0
1166 // CHECK3-NEXT:    br i1 [[TMP8]], label [[DOTOMP_SECTIONS_CASE_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE_CNCL:%.*]]
1167 // CHECK3:       .omp.sections.case.split:
1168 // CHECK3-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1169 // CHECK3:       .omp.sections.case.cncl:
1170 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
1171 // CHECK3:       .omp.sections.case2:
1172 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM3:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
1173 // CHECK3-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM3]], i32 3)
1174 // CHECK3-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[TMP9]], 0
1175 // CHECK3-NEXT:    br i1 [[TMP10]], label [[DOTOMP_SECTIONS_CASE2_SPLIT:%.*]], label [[DOTOMP_SECTIONS_CASE2_CNCL:%.*]]
1176 // CHECK3:       .omp.sections.case2.split:
1177 // CHECK3-NEXT:    br label [[DOTOMP_SECTIONS_CASE2_SECTION_AFTER:%.*]]
1178 // CHECK3:       .omp.sections.case2.section.after:
1179 // CHECK3-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
1180 // CHECK3:       .omp.sections.case2.cncl:
1181 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END]]
1182 // CHECK3:       .omp.sections.exit:
1183 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1184 // CHECK3:       omp.inner.for.inc:
1185 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
1186 // CHECK3-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP11]], 1
1187 // CHECK3-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
1188 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1189 // CHECK3:       omp.inner.for.end:
1190 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM5:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23:[0-9]+]])
1191 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM5]])
1192 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1193 // CHECK3:       cancel.cont:
1194 // CHECK3-NEXT:    ret void
1195 // CHECK3:       cancel.exit:
1196 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM4:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB23]])
1197 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB15]], i32 [[OMP_GLOBAL_THREAD_NUM4]])
1198 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1199 //
1200 //
1201 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
1202 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[R:%.*]]) #[[ATTR5]] {
1203 // CHECK3-NEXT:  entry:
1204 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1205 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1206 // CHECK3-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
1207 // CHECK3-NEXT:    [[R_ADDR:%.*]] = alloca i32*, align 8
1208 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1209 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1210 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1211 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1212 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1213 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1214 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1215 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1216 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1217 // CHECK3-NEXT:    [[R3:%.*]] = alloca i32, align 4
1218 // CHECK3-NEXT:    [[I4:%.*]] = alloca i32, align 4
1219 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
1220 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1221 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1222 // CHECK3-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
1223 // CHECK3-NEXT:    store i32* [[R]], i32** [[R_ADDR]], align 8
1224 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
1225 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[R_ADDR]], align 8
1226 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
1227 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
1228 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1229 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1230 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1231 // CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1232 // CHECK3-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1233 // CHECK3-NEXT:    store i32 0, i32* [[I]], align 4
1234 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1235 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1236 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1237 // CHECK3:       omp.precond.then:
1238 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1239 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1240 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
1241 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1242 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1243 // CHECK3-NEXT:    store i32 0, i32* [[R3]], align 4
1244 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB25:[0-9]+]])
1245 // 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)
1246 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1247 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1248 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
1249 // CHECK3-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1250 // CHECK3:       cond.true:
1251 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1252 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1253 // CHECK3:       cond.false:
1254 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1255 // CHECK3-NEXT:    br label [[COND_END]]
1256 // CHECK3:       cond.end:
1257 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
1258 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1259 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1260 // CHECK3-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
1261 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1262 // CHECK3:       omp.inner.for.cond:
1263 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1264 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1265 // CHECK3-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
1266 // CHECK3-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1267 // CHECK3:       omp.inner.for.body:
1268 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1269 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1
1270 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1271 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I4]], align 4
1272 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM7:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB27:[0-9]+]])
1273 // CHECK3-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM7]], i32 2)
1274 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
1275 // CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]]
1276 // CHECK3:       .cancel.exit:
1277 // CHECK3-NEXT:    br label [[CANCEL_EXIT:%.*]]
1278 // CHECK3:       .cancel.continue:
1279 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[I4]], align 4
1280 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[R3]], align 4
1281 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP16]]
1282 // CHECK3-NEXT:    store i32 [[ADD8]], i32* [[R3]], align 4
1283 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1284 // CHECK3:       omp.body.continue:
1285 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1286 // CHECK3:       omp.inner.for.inc:
1287 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1288 // CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP18]], 1
1289 // CHECK3-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
1290 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1291 // CHECK3:       omp.inner.for.end:
1292 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1293 // CHECK3:       omp.loop.exit:
1294 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM11:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29:[0-9]+]])
1295 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM11]])
1296 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1297 // CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i32* [[R3]] to i8*
1298 // CHECK3-NEXT:    store i8* [[TMP20]], i8** [[TMP19]], align 8
1299 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM12:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
1300 // CHECK3-NEXT:    [[TMP21:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
1301 // 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)
1302 // CHECK3-NEXT:    switch i32 [[TMP22]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1303 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1304 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1305 // CHECK3-NEXT:    ]
1306 // CHECK3:       .omp.reduction.case1:
1307 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[TMP1]], align 4
1308 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[R3]], align 4
1309 // CHECK3-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
1310 // CHECK3-NEXT:    store i32 [[ADD13]], i32* [[TMP1]], align 4
1311 // 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)
1312 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1313 // CHECK3:       cancel.exit:
1314 // CHECK3-NEXT:    [[OMP_GLOBAL_THREAD_NUM10:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB29]])
1315 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB4]], i32 [[OMP_GLOBAL_THREAD_NUM10]])
1316 // CHECK3-NEXT:    br label [[CANCEL_CONT:%.*]]
1317 // CHECK3:       .omp.reduction.case2:
1318 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[R3]], align 4
1319 // CHECK3-NEXT:    [[TMP26:%.*]] = atomicrmw add i32* [[TMP1]], i32 [[TMP25]] monotonic, align 4
1320 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1321 // CHECK3:       .omp.reduction.default:
1322 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
1323 // CHECK3:       omp.precond.end:
1324 // CHECK3-NEXT:    br label [[CANCEL_CONT]]
1325 // CHECK3:       cancel.cont:
1326 // CHECK3-NEXT:    ret void
1327 //
1328 //
1329 // CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
1330 // CHECK3-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR4]] {
1331 // CHECK3-NEXT:  entry:
1332 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
1333 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
1334 // CHECK3-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1335 // CHECK3-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1336 // CHECK3-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1337 // CHECK3-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
1338 // CHECK3-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1339 // CHECK3-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
1340 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
1341 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1342 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1343 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
1344 // CHECK3-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1345 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1346 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
1347 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
1348 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1349 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
1350 // CHECK3-NEXT:    ret void
1351 //
1352