1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - -fsanitize-address-use-after-scope | FileCheck %s --check-prefix=CHECK --check-prefix=LIFETIME --check-prefix=OMP45
2 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t -fopenmp-version=50 %s
3 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -fopenmp-version=50 | FileCheck %s --check-prefix=CHECK --check-prefix=OMP5
4 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
5 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK --check-prefix=OMP45
6 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp -fexceptions -fcxx-exceptions -debug-info-kind=line-tables-only -x c++ -emit-llvm %s -o - | FileCheck %s --check-prefix=TERM_DEBUG
7 // RUN: %clang_cc1 -main-file-name for_codegen.cpp %s -o - -emit-llvm -fprofile-instrument=clang -fprofile-instrument-path=for_codegen-test.profraw | FileCheck %s --check-prefix=PROF-INSTR-PATH
8 
9 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck --check-prefix SIMD-ONLY0 %s
10 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t -fopenmp-version=50 %s
11 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -fopenmp-version=50 | FileCheck --check-prefix SIMD-ONLY0 %s
12 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
13 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s
14 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp-simd -fexceptions -fcxx-exceptions -debug-info-kind=line-tables-only -x c++ -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
15 // RUN: %clang_cc1 -main-file-name for_codegen.cpp %s -o - -emit-llvm -fprofile-instrument=clang -fprofile-instrument-path=for_codegen-test.profraw | FileCheck --check-prefix SIMD-ONLY0 %s
16 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
17 //
18 // expected-no-diagnostics
19 #ifndef HEADER
20 #define HEADER
21 // PROF-INSTR-PATH: constant [25 x i8] c"for_codegen-test.profraw\00"
22 
23 // CHECK: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* }
24 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr global %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8*
25 // CHECK-DAG: [[LOOP_LOC:@.+]] = private unnamed_addr global %{{.+}} { i32 0, i32 514, i32 0, i32 0, i8*
26 // CHECK-DAG: [[I:@.+]] = global i8 1,
27 // CHECK-DAG: [[J:@.+]] = global i8 2,
28 // CHECK-DAG: [[K:@.+]] = global i8 3,
29 
30 // CHECK-LABEL: loop_with_counter_collapse
31 void loop_with_counter_collapse() {
32   // Captured initializations.
33   // CHECK: store i32 0, i32* [[I_TMP:%.+]],
34   // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]],
35   // CHECK: store i32 [[VAL]], i32* [[J_LB_MIN:%.+]],
36   // CHECK: store i32 3, i32* [[I_TMP]],
37   // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]],
38   // CHECK: store i32 [[VAL]], i32* [[J_LB_MAX:%.+]],
39   // CHECK: [[J_LB_MIN_VAL:%.+]] = load i32, i32* [[J_LB_MIN]],
40   // CHECK: [[J_LB_MAX_VAL:%.+]] = load i32, i32* [[J_LB_MAX]],
41   // CHECK: [[CMP:%.+]] = icmp slt i32 [[J_LB_MIN_VAL]], [[J_LB_MAX_VAL]]
42   // CHECK: [[BOOL:%.+]] = zext i1 [[CMP]] to i8
43   // CHECK: store i8 [[BOOL]], i8* [[J_LB_CMP:%.+]],
44   // CHECK: store i32 0, i32* [[I_TMP]],
45   // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]],
46   // CHECK: [[J_UB_MIN_VAL:%.+]] = add nsw i32 4, [[VAL]]
47   // CHECK: store i32 [[J_UB_MIN_VAL]], i32* [[J_UB_MIN:%.+]],
48   // CHECK: store i32 3, i32* [[I_TMP]],
49   // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]],
50   // CHECK: [[J_UB_MAX_VAL:%.+]] = add nsw i32 4, [[VAL]]
51   // CHECK: store i32 [[J_UB_MAX_VAL]], i32* [[J_UB_MAX:%.+]],
52   // CHECK: [[J_UB_MIN_VAL:%.+]] = load i32, i32* [[J_UB_MIN]],
53   // CHECK: [[J_UB_MAX_VAL:%.+]] = load i32, i32* [[J_UB_MAX]],
54   // CHECK: [[CMP:%.+]] = icmp sgt i32 [[J_UB_MIN_VAL]], [[J_UB_MAX_VAL]]
55   // CHECK: [[BOOL:%.+]] = zext i1 [[CMP]] to i8
56   // CHECK: store i8 [[BOOL]], i8* [[J_UB_CMP:%.+]],
57   // CHECK: [[J_UB_CMP_VAL:%.+]] = load i8, i8* [[J_UB_CMP]],
58   // CHECK: [[BOOL:%.+]] = trunc i8 [[J_UB_CMP_VAL]] to i1
59   // CHECK: br i1 [[BOOL]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]]
60   // CHECK: [[TRUE]]:
61   // CHECK: [[J_UB_MIN_VAL:%.+]] = load i32, i32* [[J_UB_MIN]],
62   // CHECK: br label %[[EXIT:[^,]+]]
63   // CHECK: [[FALSE]]:
64   // CHECK: [[J_UB_MAX_VAL:%.+]] = load i32, i32* [[J_UB_MAX]],
65   // CHECK: br label %[[EXIT]]
66   // CHECK: [[EXIT]]:
67   // CHECK: [[J_UB_VAL:%.+]] = phi i32 [ [[J_UB_MIN_VAL]], %[[TRUE]] ], [ [[J_UB_MAX_VAL]], %[[FALSE]] ]
68   // CHECK: store i32 [[J_UB_VAL]], i32* [[J_UB:%.+]],
69   // CHECK: [[J_LB_CMP_VAL:%.+]] = load i8, i8* [[J_LB_CMP]],
70   // CHECK: [[BOOL:%.+]] = trunc i8 [[J_LB_CMP_VAL]] to i1
71   // CHECK: br i1 [[BOOL]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]]
72   // CHECK: [[TRUE]]:
73   // CHECK: [[J_LB_MIN_VAL:%.+]] = load i32, i32* [[J_LB_MIN]],
74   // CHECK: br label %[[EXIT:[^,]+]]
75   // CHECK: [[FALSE]]:
76   // CHECK: [[J_LB_MAX_VAL:%.+]] = load i32, i32* [[J_LB_MAX]],
77   // CHECK: br label %[[EXIT]]
78   // CHECK: [[EXIT]]:
79   // CHECK: [[J_LB_VAL:%.+]] = phi i32 [ [[J_LB_MIN_VAL]], %[[TRUE]] ], [ [[J_LB_MAX_VAL]], %[[FALSE]] ]
80   // CHECK: store i32 [[J_LB_VAL]], i32* [[J_LB:%.+]],
81   // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]],
82   // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]],
83   // CHECK: [[SUB:%.+]] = sub nsw i32 [[J_UB_VAL]], [[J_LB_VAL]]
84   // CHECK: [[SUB_ST:%.+]] = sub nsw i32 [[SUB]], 1
85   // CHECK: [[ADD_ST:%.+]] = add nsw i32 [[SUB_ST]], 1
86   // CHECK: [[DIV_ST:%.+]] = sdiv i32 [[ADD_ST]], 1
87   // CHECK: [[CAST:%.+]] = sext i32 [[DIV_ST]] to i64
88   // CHECK: [[MUL:%.+]] = mul nsw i64 4, [[CAST]]
89   // CHECK: [[NUM_ITERS_VAL:%.+]] = sub nsw i64 [[MUL]], 1
90   // CHECK: store i64 [[NUM_ITERS_VAL]], i64* [[NUM_ITERS:%.+]],
91 
92   // CHECK: store i64 0, i64* [[LB:%.+]],
93   // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, i64* [[NUM_ITERS]],
94   // CHECK: store i64 [[NUM_ITERS_VAL]], i64* [[UB:%.+]],
95   // CHECK: store i64 1, i64* [[STRIDE:%.+]],
96   // CHECK: store i32 0, i32* [[IS_LAST:%.+]],
97   // CHECK: call void @__kmpc_for_static_init_8(%struct.ident_t* @{{.+}}, i32 %{{.+}}, i32 34, i32* [[IS_LAST]], i64* [[LB]], i64* [[UB]], i64* [[STRIDE]], i64 1, i64 1)
98   // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]],
99   // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, i64* [[NUM_ITERS]],
100   // CHECK: [[CMP:%.+]] = icmp sgt i64 [[UB_VAL]], [[NUM_ITERS_VAL]]
101   // CHECK: br i1 [[CMP]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]]
102   // CHECK: [[TRUE]]:
103   // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, i64* [[NUM_ITERS]],
104   // CHECK: br label %[[DONE:[^,]+]]
105   // CHECK: [[FALSE]]:
106   // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]],
107   // CHECK: br label %[[DONE]]
108   // CHECK: [[DONE]]:
109   // CHECK: [[TOP:%.+]] = phi i64 [ [[NUM_ITERS_VAL]], %[[TRUE]] ], [ [[UB_VAL]], %[[FALSE]] ]
110   // CHECK: store i64 [[TOP]], i64* [[UB]],
111   // CHECK: [[LB_VAL:%.+]] = load i64, i64* [[LB]],
112   // CHECK: store i64 [[LB_VAL]], i64* [[IV:%.+]],
113   // CHECK: br label %[[COND:[^,]+]]
114   // CHECK: [[COND]]:
115   // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]],
116   // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]],
117   // CHECK: [[CMP:%.+]] = icmp sle i64 [[IV_VAL]], [[UB_VAL]]
118   // CHECK: br i1 [[CMP]], label %[[BODY:[^,]+]], label %[[CLEANUP:[^,]+]]
119   // LIFETIME: [[CLEANUP]]:
120   // LIFETIME: br label %[[CLEANUP:[^,]+]]
121   // CHECK: [[BODY]]:
122   // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]],
123   // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]],
124   // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]],
125   // CHECK: [[SUB:%.+]] = sub nsw i32 [[J_UB_VAL]], [[J_LB_VAL]]
126   // CHECK: [[SUB_ST:%.+]] = sub nsw i32 [[SUB]], 1
127   // CHECK: [[ADD_ST:%.+]] = add nsw i32 [[SUB_ST]], 1
128   // CHECK: [[DIV_ST:%.+]] = sdiv i32 [[ADD_ST]], 1
129   // CHECK: [[MUL:%.+]] = mul nsw i32 1, [[DIV_ST]]
130   // CHECK: [[CAST:%.+]] = sext i32 [[MUL]] to i64
131   // CHECK: [[DIV:%.+]] = sdiv i64 [[IV_VAL]], [[CAST]]
132   // CHECK: [[MUL:%.+]] = mul nsw i64 [[DIV]], 1
133   // CHECK: [[ADD:%.+]] = add nsw i64 0, [[MUL]]
134   // CHECK: [[CAST:%.+]] = trunc i64 [[ADD]] to i32
135   // CHECK: store i32 [[CAST]], i32* [[I_PRIV:%.+]],
136   // CHECK: [[I_VAL:%.+]] = load i32, i32* [[I_PRIV]],
137   // CHECK: [[CONV:%.+]] = sext i32 [[I_VAL]] to i64
138   // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]],
139   // CHECK: [[IV_VAL1:%.+]] = load i64, i64* [[IV]],
140   // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]],
141   // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]],
142   // CHECK: [[SUB:%.+]] = sub nsw i32 [[J_UB_VAL]], [[J_LB_VAL]]
143   // CHECK: [[SUB_ST:%.+]] = sub nsw i32 [[SUB]], 1
144   // CHECK: [[ADD_ST:%.+]] = add nsw i32 [[SUB_ST]], 1
145   // CHECK: [[DIV_ST:%.+]] = sdiv i32 [[ADD_ST]], 1
146   // CHECK: [[MUL:%.+]] = mul nsw i32 1, [[DIV_ST]]
147   // CHECK: [[CAST:%.+]] = sext i32 [[MUL]] to i64
148   // CHECK: [[DIV:%.+]] = sdiv i64 [[IV_VAL1]], [[CAST]]
149   // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]],
150   // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]],
151   // CHECK: [[SUB:%.+]] = sub nsw i32 [[J_UB_VAL]], [[J_LB_VAL]]
152   // CHECK: [[SUB_ST:%.+]] = sub nsw i32 [[SUB]], 1
153   // CHECK: [[ADD_ST:%.+]] = add nsw i32 [[SUB_ST]], 1
154   // CHECK: [[DIV_ST:%.+]] = sdiv i32 [[ADD_ST]], 1
155   // CHECK: [[MUL:%.+]] = mul nsw i32 1, [[DIV_ST]]
156   // CHECK: [[CAST:%.+]] = sext i32 [[MUL]] to i64
157   // CHECK: [[MUL:%.+]] = mul nsw i64 [[DIV]], [[CAST]]
158   // CHECK: [[SUB:%.+]] = sub nsw i64 [[IV_VAL]], [[MUL]]
159   // CHECK: [[MUL:%.+]] = mul nsw i64 [[SUB:%.+]], 1
160   // CHECK: [[ADD:%.+]] = add nsw i64 [[CONV]], [[MUL]]
161   // CHECK: [[CAST:%.+]] = trunc i64 [[ADD]] to i32
162   // CHECK: store i32 [[CAST]], i32* [[J_PRIV:%.+]],
163 
164   // Check that the loop variable is not out of its boundaries.
165   // CHECK: [[J_VAL:%.+]] = load i32, i32* [[J_PRIV]],
166   // CHECK: [[I_VAL:%.+]] = load i32, i32* [[I_PRIV]],
167   // CHECK: [[J_COND:%.+]] = add nsw i32 4, [[I_VAL]]
168   // CHECK: [[CMP:%.+]] = icmp slt i32 [[J_VAL]], [[J_COND]]
169   // CHECK: br i1 [[CMP]], label %[[NEXT:[^,]+]], label %[[BODY_CONT:[^,]+]]
170   // CHECK: [[NEXT]]:
171 
172   // Main body is empty.
173   // CHECK: br label %[[BODY_CONT]]
174   // CHECK: [[BODY_CONT]]:
175   // CHECK: br label %[[INC:[^,]+]]
176   // CHECK: [[INC]]:
177   // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]],
178   // CHECK: [[ADD:%.+]] = add nsw i64 [[IV_VAL]], 1
179   // CHECK: store i64 [[ADD]], i64* [[IV]],
180   // CHECK: br label %[[COND]]
181   // CHECK: [[CLEANUP]]:
182   // CHECK: br label %[[EXIT:[^,]+]]
183   // CHECK: [[EXIT]]:
184   // CHECK: call void @__kmpc_for_static_fini(%struct.ident_t* @{{.+}}, i32 %{{.+}})
185   // LIFETIME: call void @llvm.lifetime.end
186   // LIFETIME: call void @llvm.lifetime.end
187   // LIFETIME: call void @llvm.lifetime.end
188   // LIFETIME: call void @llvm.lifetime.end
189   // LIFETIME: call void @llvm.lifetime.end
190   // LIFETIME: call void @llvm.lifetime.end
191   // LIFETIME: call void @llvm.lifetime.end
192   // LIFETIME: call void @llvm.lifetime.end
193   // LIFETIME: call void @llvm.lifetime.end
194   #pragma omp for collapse(2)
195   for (int i = 0; i < 4; i++) {
196     for (int j = i; j < 4 + i; j++) {
197     }
198   }
199 }
200 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
201 void without_schedule_clause(float *a, float *b, float *c, float *d) {
202 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
203   #pragma omp for nowait
204 // CHECK: call void @__kmpc_for_static_init_4([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]], i32 34, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 1)
205 // UB = min(UB, GlobalUB)
206 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
207 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
208 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
209 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
210 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
211 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
212 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
213 // Loop header
214 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
215 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
216 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
217 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
218   for (int i = 33; i < 32000000; i += 7) {
219 // CHECK: [[LOOP1_BODY]]
220 // Start of body: calculate i from IV:
221 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
222 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7
223 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]]
224 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
225 // ... loop body ...
226 // End of body: store into a[i]:
227 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
228 // CHECK-NOT: !llvm.access.group
229     a[i] = b[i] * c[i] * d[i];
230 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
231 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
232 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
233 // CHECK-NEXT: br label %{{.+}}
234   }
235 // CHECK: [[LOOP1_END]]
236 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
237 // CHECK-NOT: __kmpc_barrier
238 // CHECK: ret void
239 }
240 
241 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
242 void static_not_chunked(float *a, float *b, float *c, float *d) {
243 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
244   #pragma omp for schedule(static)
245 // CHECK: call void @__kmpc_for_static_init_4([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]], i32 34, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 1)
246 // UB = min(UB, GlobalUB)
247 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
248 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
249 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
250 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
251 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
252 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
253 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
254 // Loop header
255 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
256 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
257 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
258 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
259   for (int i = 32000000; i > 33; i += -7) {
260 // CHECK: [[LOOP1_BODY]]
261 // Start of body: calculate i from IV:
262 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
263 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7
264 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]]
265 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
266 // ... loop body ...
267 // End of body: store into a[i]:
268 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
269 // CHECK-NOT: !llvm.access.group
270     a[i] = b[i] * c[i] * d[i];
271 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
272 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
273 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
274 // CHECK-NEXT: br label %{{.+}}
275   }
276 // CHECK: [[LOOP1_END]]
277 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
278 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
279 // CHECK: ret void
280 }
281 
282 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
283 void static_chunked(float *a, float *b, float *c, float *d) {
284 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
285   #pragma omp for schedule(monotonic: static, 5)
286 // CHECK: call void @__kmpc_for_static_init_4u([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]], i32 536870945, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 5)
287 // UB = min(UB, GlobalUB)
288 // CHECK: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
289 // CHECK-NEXT: [[UBCMP:%.+]] = icmp ugt i32 [[UB]], 16908288
290 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
291 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 16908288, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
292 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
293 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
294 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
295 
296 // Outer loop header
297 // CHECK: [[O_IV:%.+]] = load i32, i32* [[OMP_IV]]
298 // CHECK-NEXT: [[O_UB:%.+]] = load i32, i32* [[OMP_UB]]
299 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ule i32 [[O_IV]], [[O_UB]]
300 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
301 
302 // Loop header
303 // CHECK: [[O_LOOP1_BODY]]
304 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
305 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
306 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i32 [[IV]], [[UB]]
307 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
308   for (unsigned i = 131071; i <= 2147483647; i += 127) {
309 // CHECK: [[LOOP1_BODY]]
310 // Start of body: calculate i from IV:
311 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
312 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i32 [[IV1_1]], 127
313 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 131071, [[CALC_I_1]]
314 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
315 // ... loop body ...
316 // End of body: store into a[i]:
317 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
318 // CHECK-NOT: !llvm.access.group
319     a[i] = b[i] * c[i] * d[i];
320 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
321 // CHECK-NEXT: [[ADD1_2:%.+]] = add i32 [[IV1_2]], 1
322 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
323 // CHECK-NEXT: br label %{{.+}}
324   }
325 // CHECK: [[LOOP1_END]]
326 // Update the counters, adding stride
327 // CHECK:  [[LB:%.+]] = load i32, i32* [[OMP_LB]]
328 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]]
329 // CHECK-NEXT: [[ADD_LB:%.+]] = add i32 [[LB]], [[ST]]
330 // CHECK-NEXT: store i32 [[ADD_LB]], i32* [[OMP_LB]]
331 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
332 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]]
333 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]]
334 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]]
335 
336 // CHECK: [[O_LOOP1_END]]
337 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
338 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
339 // CHECK: ret void
340 }
341 
342 // CHECK-LABEL: define {{.*void}} @{{.*}}dynamic1{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
343 void dynamic1(float *a, float *b, float *c, float *d) {
344 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
345   #pragma omp for schedule(nonmonotonic: dynamic)
346 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741859, i64 0, i64 16908287, i64 1, i64 1)
347 //
348 // CHECK: [[HASWORK:%.+]] = call i32 @__kmpc_dispatch_next_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32* [[OMP_ISLAST:%[^,]+]], i64* [[OMP_LB:%[^,]+]], i64* [[OMP_UB:%[^,]+]], i64* [[OMP_ST:%[^,]+]])
349 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
350 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
351 
352 // Loop header
353 // CHECK: [[O_LOOP1_BODY]]
354 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
355 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
356 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
357 
358 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
359 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1
360 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]]
361 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
362   for (unsigned long long i = 131071; i < 2147483647; i += 127) {
363 // CHECK: [[LOOP1_BODY]]
364 // Start of body: calculate i from IV:
365 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]]
366 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127
367 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]]
368 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]]
369 // ... loop body ...
370 // End of body: store into a[i]:
371 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.access.group
372     a[i] = b[i] * c[i] * d[i];
373 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
374 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
375 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
376 // CHECK-NEXT: br label %{{.+}}
377   }
378 // CHECK: [[LOOP1_END]]
379 // CHECK: [[O_LOOP1_END]]
380 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
381 // CHECK: ret void
382 }
383 
384 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
385 void guided7(float *a, float *b, float *c, float *d) {
386 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
387   #pragma omp for schedule(guided, 7)
388 // OMP45: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 36, i64 0, i64 16908287, i64 1, i64 7)
389 // OMP5: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741860, i64 0, i64 16908287, i64 1, i64 7)
390 //
391 // CHECK: [[HASWORK:%.+]] = call i32 @__kmpc_dispatch_next_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32* [[OMP_ISLAST:%[^,]+]], i64* [[OMP_LB:%[^,]+]], i64* [[OMP_UB:%[^,]+]], i64* [[OMP_ST:%[^,]+]])
392 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
393 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
394 
395 // Loop header
396 // CHECK: [[O_LOOP1_BODY]]
397 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
398 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
399 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
400 
401 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
402 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1
403 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]]
404 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
405   for (unsigned long long i = 131071; i < 2147483647; i += 127) {
406 // CHECK: [[LOOP1_BODY]]
407 // Start of body: calculate i from IV:
408 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]]
409 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127
410 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]]
411 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]]
412 // ... loop body ...
413 // End of body: store into a[i]:
414 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.access.group
415     a[i] = b[i] * c[i] * d[i];
416 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
417 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
418 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
419 // CHECK-NEXT: br label %{{.+}}
420   }
421 // CHECK: [[LOOP1_END]]
422 // CHECK: [[O_LOOP1_END]]
423 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
424 // CHECK: ret void
425 }
426 
427 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
428 void test_auto(float *a, float *b, float *c, float *d) {
429   unsigned int x = 0;
430   unsigned int y = 0;
431 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
432   #pragma omp for schedule(auto) collapse(2)
433 // OMP45: call void @__kmpc_dispatch_init_8([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 38, i64 0, i64 [[LAST_ITER:%[^,]+]], i64 1, i64 1)
434 // OMP5: call void @__kmpc_dispatch_init_8([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741862, i64 0, i64 [[LAST_ITER:%[^,]+]], i64 1, i64 1)
435 //
436 // CHECK: [[HASWORK:%.+]] = call i32 @__kmpc_dispatch_next_8([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32* [[OMP_ISLAST:%[^,]+]], i64* [[OMP_LB:%[^,]+]], i64* [[OMP_UB:%[^,]+]], i64* [[OMP_ST:%[^,]+]])
437 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
438 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
439 
440 // Loop header
441 // CHECK: [[O_LOOP1_BODY]]
442 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
443 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
444 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
445 
446 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
447 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]]
448 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
449 // FIXME: When the iteration count of some nested loop is not a known constant,
450 // we should pre-calculate it, like we do for the total number of iterations!
451   for (char i = static_cast<char>(y); i <= '9'; ++i)
452     for (x = 11; x > 0; --x) {
453 // CHECK: [[LOOP1_BODY]]
454 // Start of body: indices are calculated from IV:
455 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}}
456 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}}
457 // ... loop body ...
458 // End of body: store into a[i]:
459 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
460 // CHECK-NOT: !llvm.access.group
461     a[i] = b[i] * c[i] * d[i];
462 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
463 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1
464 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
465 // CHECK-NEXT: br label %{{.+}}
466   }
467 // CHECK: [[LOOP1_END]]
468 // CHECK: [[O_LOOP1_END]]
469 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
470 // CHECK: ret void
471 }
472 
473 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
474 void runtime(float *a, float *b, float *c, float *d) {
475   int x = 0;
476 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
477   #pragma omp for collapse(2) schedule(runtime)
478 // OMP45: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 37, i32 0, i32 199, i32 1, i32 1)
479 // OMP5: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741861, i32 0, i32 199, i32 1, i32 1)
480 //
481 // CHECK: [[HASWORK:%.+]] = call i32 @__kmpc_dispatch_next_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32* [[OMP_ISLAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]])
482 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
483 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
484 
485 // Loop header
486 // CHECK: [[O_LOOP1_BODY]]
487 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
488 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
489 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
490 
491 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
492 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
493 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
494   for (unsigned char i = '0' ; i <= '9'; ++i)
495     for (x = -10; x < 10; ++x) {
496 // CHECK: [[LOOP1_BODY]]
497 // Start of body: indices are calculated from IV:
498 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}}
499 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}}
500 // ... loop body ...
501 // End of body: store into a[i]:
502 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
503 // CHECK-NOT: !llvm.access.group
504     a[i] = b[i] * c[i] * d[i];
505 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
506 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
507 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
508 // CHECK-NEXT: br label %{{.+}}
509   }
510 // CHECK: [[LOOP1_END]]
511 // CHECK: [[O_LOOP1_END]]
512 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
513 // CHECK: ret void
514 }
515 
516 // CHECK-LABEL: test_precond
517 void test_precond() {
518   // CHECK: [[A_ADDR:%.+]] = alloca i8,
519   // CHECK: [[I_ADDR:%.+]] = alloca i8,
520   // CHECK: [[CAP:%.+]] = alloca i8,
521   char a = 0;
522   // CHECK: store i8 0,
523   // CHECK: store i32
524   // CHECK: store i8
525   // CHECK: [[A:%.+]] = load i8, i8* [[CAP]],
526   // CHECK: [[CONV:%.+]] = sext i8 [[A]] to i32
527   // CHECK: [[CMP:%.+]] = icmp slt i32 [[CONV]], 10
528   // CHECK: br i1 [[CMP]], label %[[PRECOND_THEN:[^,]+]], label %[[PRECOND_END:[^,]+]]
529   // CHECK: [[PRECOND_THEN]]
530   // CHECK: call void @__kmpc_for_static_init_4
531 #pragma omp for
532   for(char i = a; i < 10; ++i);
533   // CHECK: call void @__kmpc_for_static_fini
534   // CHECK: [[PRECOND_END]]
535 }
536 
537 // TERM_DEBUG-LABEL: foo
538 int foo() {return 0;};
539 
540 // TERM_DEBUG-LABEL: parallel_for
541 void parallel_for(float *a) {
542 #pragma omp parallel
543 #pragma omp for schedule(static, 5)
544   // TERM_DEBUG-NOT: __kmpc_global_thread_num
545   // TERM_DEBUG:     call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC:![0-9]+]]
546   // TERM_DEBUG:     invoke i32 {{.*}}foo{{.*}}()
547   // TERM_DEBUG:     unwind label %[[TERM_LPAD:.+]],
548   // TERM_DEBUG-NOT: __kmpc_global_thread_num
549   // TERM_DEBUG:     call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC]]
550   // TERM_DEBUG:     call {{.+}} @__kmpc_barrier({{.+}}), !dbg [[DBG_LOC]]
551   // TERM_DEBUG:     [[TERM_LPAD]]
552   // TERM_DEBUG:     call void @__clang_call_terminate
553   // TERM_DEBUG:     unreachable
554   for (unsigned i = 131071; i <= 2147483647; i += 127)
555     a[i] += foo();
556 }
557 // Check source line corresponds to "#pragma omp for schedule(static, 5)" above:
558 // TERM_DEBUG: [[DBG_LOC]] = !DILocation(line: [[@LINE-15]],
559 
560 char i = 1, j = 2, k = 3;
561 // CHECK-LABEL: for_with_global_lcv
562 void for_with_global_lcv() {
563 // CHECK: alloca i8,
564 // CHECK: [[I_ADDR:%.+]] = alloca i8,
565 // CHECK: alloca i8,
566 // CHECK: [[J_ADDR:%.+]] = alloca i8,
567 
568 // CHECK: call void @__kmpc_for_static_init_4(
569 // CHECK-NOT: [[I]]
570 // CHECK: store i8 %{{.+}}, i8* [[I_ADDR]]
571 // CHECK-NOT: [[I]]
572 // CHECK: [[I_VAL:%.+]] = load i8, i8* [[I_ADDR]],
573 // CHECK-NOT: [[I]]
574 // CHECK: store i8 [[I_VAL]], i8* [[K]]
575 // CHECK-NOT: [[I]]
576 // CHECK: call void @__kmpc_for_static_fini(
577 // CHECK: call void @__kmpc_barrier(
578 #pragma omp for
579   for (i = 0; i < 2; ++i) {
580     k = i;
581   }
582 // CHECK: call void @__kmpc_for_static_init_4(
583 // CHECK-NOT: [[J]]
584 // CHECK: store i8 %{{.+}}, i8* [[J_ADDR]]
585 // CHECK-NOT: [[J]]
586 // CHECK: [[J_VAL:%.+]] = load i8, i8* [[J_ADDR]],
587 // CHECK-NOT: [[J]]
588 // CHECK: store i8 [[J_VAL]], i8* [[K]]
589 // CHECK-NOT: [[J]]
590 // CHECK: call void @__kmpc_for_static_fini(
591 #pragma omp for collapse(2)
592   for (int i = 0; i < 2; ++i)
593   for (j = 0; j < 2; ++j) {
594     k = i;
595     k = j;
596   }
597   char &cnt = i;
598 #pragma omp for
599   for (cnt = 0; cnt < 2; ++cnt)
600     k = cnt;
601 }
602 
603 // CHECK-LABEL: for_with_references
604 void for_with_references() {
605 // CHECK: [[I:%.+]] = alloca i8,
606 // CHECK: [[CNT:%.+]] = alloca i8*,
607 // CHECK: [[CNT_PRIV:%.+]] = alloca i8,
608 // CHECK: call void @__kmpc_for_static_init_8(
609 // CHECK-NOT: load i8, i8* [[CNT]],
610 // CHECK: call void @__kmpc_for_static_fini(
611   char i = 0;
612   char &cnt = i;
613 #pragma omp for collapse(2)
614   for (cnt = 0; cnt < 2; ++cnt)
615     for (int j = cnt; j < 4 + cnt; j++)
616     k = cnt;
617 }
618 
619 // CHECK-LABEL: for_with_references_dep_cond
620 void for_with_references_dep_cond() {
621 // CHECK: [[I:%.+]] = alloca i8,
622 // CHECK: [[CNT:%.+]] = alloca i8*,
623 // CHECK: [[CNT_PRIV:%.+]] = alloca i8,
624 // CHECK: call void @__kmpc_for_static_init_8(
625 // CHECK-NOT: load i8, i8* [[CNT]],
626 // CHECK: call void @__kmpc_for_static_fini(
627   char i = 0;
628   char &cnt = i;
629 #pragma omp for collapse(2)
630   for (cnt = 0; cnt < 2; ++cnt)
631     for (int j = 0; j < 4 + cnt; j++)
632     k = cnt;
633 }
634 
635 struct Bool {
636   Bool(bool b) : b(b) {}
637   operator bool() const { return b; }
638   const bool b;
639 };
640 
641 template <typename T>
642 struct It {
643   It() : p(0) {}
644   It(const It &, int = 0) ;
645   template <typename U>
646   It(U &, int = 0) ;
647   It &operator=(const It &);
648   It &operator=(It &);
649   ~It() {}
650 
651   It(T *p) : p(p) {}
652 
653   operator T *&() { return p; }
654   operator T *() const { return p; }
655   T *operator->() const { return p; }
656 
657   It &operator++() { ++p; return *this; }
658   It &operator--() { --p; return *this; }
659   It &operator+=(unsigned n) { p += n; return *this; }
660   It &operator-=(unsigned n) { p -= n; return *this; }
661 
662   T *p;
663 };
664 
665 template <typename T>
666 It<T> operator+(It<T> a, typename It<T>::difference_type n) { return a.p + n; }
667 
668 template <typename T>
669 It<T> operator+(typename It<T>::difference_type n, It<T> a) { return a.p + n; }
670 
671 template <typename T>
672 It<T> operator-(It<T> a, typename It<T>::difference_type n) { return a.p - n; }
673 
674 typedef Bool BoolType;
675 
676 template <typename T>
677 BoolType operator<(It<T> a, It<T> b) { return a.p < b.p; }
678 
679 void loop_with_It(It<char> begin, It<char> end) {
680 #pragma omp for
681   for (It<char> it = begin; it < end; ++it) {
682     *it = 0;
683   }
684 }
685 
686 // CHECK-LABEL: loop_with_It
687 // CHECK: call i32 @__kmpc_global_thread_num(
688 // CHECK: call void @__kmpc_for_static_init_8(
689 // CHECK: call void @__kmpc_for_static_fini(
690 
691 void loop_with_It_plus(It<char> begin, It<char> end) {
692 #pragma omp for
693   for (It<char> it = begin; it < end; it+=1u) {
694     *it = 0;
695   }
696 }
697 
698 // CHECK-LABEL: loop_with_It_plus
699 // CHECK: call i32 @__kmpc_global_thread_num(
700 // CHECK: call void @__kmpc_for_static_init_8(
701 // CHECK: call void @__kmpc_for_static_fini(
702 
703 void loop_with_stmt_expr() {
704 #pragma omp for collapse(2)
705   for (int i = __extension__({float b = 0;b; }); i < __extension__({double c = 1;c; }); i += __extension__({char d = 1; d; }))
706     for (int j = i; j < 4 + i; j++)
707     ;
708 }
709 // CHECK-LABEL: loop_with_stmt_expr
710 // CHECK: call i32 @__kmpc_global_thread_num(
711 // CHECK: call void @__kmpc_for_static_init_8(
712 // CHECK: call void @__kmpc_for_static_fini(
713 
714 
715 // CHECK-LABEL: fint
716 // CHECK: call {{.*}}i32 {{.*}}ftemplate
717 // CHECK: ret i32
718 
719 // CHECK: load i16, i16*
720 // CHECK: store i16 %
721 // CHECK: call void {{.+}}@__kmpc_fork_call(
722 // CHECK: call void @__kmpc_for_static_init_4(
723 template <typename T>
724 T ftemplate() {
725   short aa = 0;
726 
727 #pragma omp parallel for schedule(static, aa)
728   for (int i = 0; i < 100; i++) {
729   }
730   return T();
731 }
732 
733 int fint(void) { return ftemplate<int>(); }
734 
735 // Check for imperfectly loop nests codegen.
736 #if _OPENMP == 201811
737 void first();
738 void last();
739 void inner_f();
740 void inner_l();
741 void body_f();
742 
743 // OMP5-LABEL: imperfectly_nested_loop
744 void imperfectly_nested_loop() {
745   // OMP5: call void @__kmpc_for_static_init_4(
746 #pragma omp for collapse(3)
747   for (int i = 0; i < 10; ++i) {
748     {
749       int a, d;
750       // OMP5: invoke void @{{.+}}first{{.+}}()
751       first();
752       // OMP5: load i32
753       // OMP5: store i32
754       a = d;
755       for (int j = 0; j < 10; ++j) {
756         int a, d;
757         // OMP5: invoke void @{{.+}}inner_f{{.+}}()
758         inner_f();
759         // OMP5: load i32
760         // OMP5: store i32
761         a = d;
762         for (int k = 0; k < 10; ++k) {
763           int a, d;
764           // OMP5: invoke void @{{.+}}body_f{{.+}}()
765           body_f();
766           // OMP5: load i32
767           // OMP5: store i32
768           a = d;
769         }
770         // OMP5: invoke void @{{.+}}inner_l{{.+}}()
771         inner_l();
772       }
773       // OMP5: invoke void @{{.+}}last{{.+}}()
774       last();
775     }
776   }
777   // OMP5: call void @__kmpc_for_static_fini(
778 }
779 #endif
780 
781 #endif // HEADER
782