1 // RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s
2 // RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
3 // RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
4 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp=libiomp5 -fexceptions -fcxx-exceptions -gline-tables-only -x c++ -emit-llvm %s -o - | FileCheck %s --check-prefix=TERM_DEBUG
5 //
6 // expected-no-diagnostics
7 #ifndef HEADER
8 #define HEADER
9 
10 // CHECK: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* }
11 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
12 void without_schedule_clause(float *a, float *b, float *c, float *d) {
13   #pragma omp parallel for
14 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %{{.+}}*)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), i8* %{{.+}})
15 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
16 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
17 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
18 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
19 // CHECK: call void @__kmpc_for_static_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 34, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 1)
20 // UB = min(UB, GlobalUB)
21 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
22 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
23 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
24 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
25 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
26 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
27 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
28 // Loop header
29 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
30 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
31 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
32 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
33   for (int i = 33; i < 32000000; i += 7) {
34 // CHECK: [[LOOP1_BODY]]
35 // Start of body: calculate i from IV:
36 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
37 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7
38 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]]
39 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
40 // ... loop body ...
41 // End of body: store into a[i]:
42 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
43     a[i] = b[i] * c[i] * d[i];
44 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
45 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
46 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
47 // CHECK-NEXT: br label %{{.+}}
48   }
49 // CHECK: [[LOOP1_END]]
50 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
51 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
52 // CHECK: ret void
53 }
54 
55 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
56 void static_not_chunked(float *a, float *b, float *c, float *d) {
57   #pragma omp parallel for schedule(static)
58 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %{{.+}}*)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), i8* %{{.+}})
59 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
60 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
61 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
62 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
63 // CHECK: call void @__kmpc_for_static_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 34, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 1)
64 // UB = min(UB, GlobalUB)
65 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
66 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
67 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
68 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
69 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
70 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
71 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
72 // Loop header
73 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
74 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
75 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
76 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
77   for (int i = 32000000; i > 33; i += -7) {
78 // CHECK: [[LOOP1_BODY]]
79 // Start of body: calculate i from IV:
80 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
81 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7
82 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]]
83 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
84 // ... loop body ...
85 // End of body: store into a[i]:
86 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
87     a[i] = b[i] * c[i] * d[i];
88 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
89 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
90 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
91 // CHECK-NEXT: br label %{{.+}}
92   }
93 // CHECK: [[LOOP1_END]]
94 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
95 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
96 // CHECK: ret void
97 }
98 
99 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
100 void static_chunked(float *a, float *b, float *c, float *d) {
101   #pragma omp parallel for schedule(static, 5)
102 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %{{.+}}*)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), i8* %{{.+}})
103 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
104 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
105 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
106 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
107 // CHECK: call void @__kmpc_for_static_init_4u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 33, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 5)
108 // UB = min(UB, GlobalUB)
109 // CHECK: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
110 // CHECK-NEXT: [[UBCMP:%.+]] = icmp ugt i32 [[UB]], 16908288
111 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
112 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 16908288, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
113 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
114 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
115 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
116 
117 // Outer loop header
118 // CHECK: [[O_IV:%.+]] = load i32, i32* [[OMP_IV]]
119 // CHECK-NEXT: [[O_UB:%.+]] = load i32, i32* [[OMP_UB]]
120 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ule i32 [[O_IV]], [[O_UB]]
121 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
122 
123 // Loop header
124 // CHECK: [[O_LOOP1_BODY]]
125 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
126 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
127 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i32 [[IV]], [[UB]]
128 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
129   for (unsigned i = 131071; i <= 2147483647; i += 127) {
130 // CHECK: [[LOOP1_BODY]]
131 // Start of body: calculate i from IV:
132 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
133 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i32 [[IV1_1]], 127
134 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 131071, [[CALC_I_1]]
135 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
136 // ... loop body ...
137 // End of body: store into a[i]:
138 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
139     a[i] = b[i] * c[i] * d[i];
140 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
141 // CHECK-NEXT: [[ADD1_2:%.+]] = add i32 [[IV1_2]], 1
142 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
143 // CHECK-NEXT: br label %{{.+}}
144   }
145 // CHECK: [[LOOP1_END]]
146 // Update the counters, adding stride
147 // CHECK:  [[LB:%.+]] = load i32, i32* [[OMP_LB]]
148 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]]
149 // CHECK-NEXT: [[ADD_LB:%.+]] = add i32 [[LB]], [[ST]]
150 // CHECK-NEXT: store i32 [[ADD_LB]], i32* [[OMP_LB]]
151 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
152 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]]
153 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]]
154 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]]
155 
156 // CHECK: [[O_LOOP1_END]]
157 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
158 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
159 // CHECK: ret void
160 }
161 
162 // CHECK-LABEL: define {{.*void}} @{{.*}}dynamic1{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
163 void dynamic1(float *a, float *b, float *c, float *d) {
164   #pragma omp parallel for schedule(dynamic)
165 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %{{.+}}*)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), i8* %{{.+}})
166 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
167 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
168 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
169 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
170 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 35, i64 0, i64 16908287, i64 1, i64 1)
171 //
172 // 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:%[^,]+]])
173 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
174 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
175 
176 // Loop header
177 // CHECK: [[O_LOOP1_BODY]]
178 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
179 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
180 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
181 
182 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
183 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i64 [[IV]], [[UB]]
184 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
185   for (unsigned long long i = 131071; i < 2147483647; i += 127) {
186 // CHECK: [[LOOP1_BODY]]
187 // Start of body: calculate i from IV:
188 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]]
189 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127
190 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]]
191 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]]
192 // ... loop body ...
193 // End of body: store into a[i]:
194 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
195     a[i] = b[i] * c[i] * d[i];
196 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
197 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
198 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
199 // CHECK-NEXT: br label %{{.+}}
200   }
201 // CHECK: [[LOOP1_END]]
202 // CHECK: [[O_LOOP1_END]]
203 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
204 // CHECK: ret void
205 }
206 
207 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
208 void guided7(float *a, float *b, float *c, float *d) {
209   #pragma omp parallel for schedule(guided, 7)
210 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %{{.+}}*)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), i8* %{{.+}})
211 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
212 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
213 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
214 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
215 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 36, i64 0, i64 16908287, i64 1, i64 7)
216 //
217 // 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:%[^,]+]])
218 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
219 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
220 
221 // Loop header
222 // CHECK: [[O_LOOP1_BODY]]
223 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
224 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
225 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
226 
227 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
228 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i64 [[IV]], [[UB]]
229 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
230   for (unsigned long long i = 131071; i < 2147483647; i += 127) {
231 // CHECK: [[LOOP1_BODY]]
232 // Start of body: calculate i from IV:
233 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]]
234 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127
235 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]]
236 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]]
237 // ... loop body ...
238 // End of body: store into a[i]:
239 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
240     a[i] = b[i] * c[i] * d[i];
241 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
242 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
243 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
244 // CHECK-NEXT: br label %{{.+}}
245   }
246 // CHECK: [[LOOP1_END]]
247 // CHECK: [[O_LOOP1_END]]
248 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
249 // CHECK: ret void
250 }
251 
252 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
253 void test_auto(float *a, float *b, float *c, float *d) {
254   unsigned int x = 0;
255   unsigned int y = 0;
256   #pragma omp parallel for schedule(auto) collapse(2)
257 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %{{.+}}*)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), i8* %{{.+}})
258 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
259 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
260 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
261 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
262 // CHECK: call void @__kmpc_dispatch_init_8([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 38, i64 0, i64 [[LAST_ITER:%[^,]+]], i64 1, i64 1)
263 //
264 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
265 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
266 // 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:%[^,]+]])
267 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
268 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
269 
270 // Loop header
271 // CHECK: [[O_LOOP1_BODY]]
272 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
273 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
274 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
275 
276 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
277 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]]
278 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
279 // FIXME: When the iteration count of some nested loop is not a known constant,
280 // we should pre-calculate it, like we do for the total number of iterations!
281   for (char i = static_cast<char>(y); i <= '9'; ++i)
282     for (x = 11; x > 0; --x) {
283 // CHECK: [[LOOP1_BODY]]
284 // Start of body: indices are calculated from IV:
285 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}}
286 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}}
287 // ... loop body ...
288 // End of body: store into a[i]:
289 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
290     a[i] = b[i] * c[i] * d[i];
291 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
292 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1
293 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
294 // CHECK-NEXT: br label %{{.+}}
295   }
296 // CHECK: [[LOOP1_END]]
297 // CHECK: [[O_LOOP1_END]]
298 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
299 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
300 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
301 // CHECK: ret void
302 }
303 
304 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
305 void runtime(float *a, float *b, float *c, float *d) {
306   int x = 0;
307   #pragma omp parallel for collapse(2) schedule(runtime)
308 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %{{.+}}*)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*), i8* %{{.+}})
309 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
310 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
311 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
312 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
313 // CHECK: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 37, i32 0, i32 199, i32 1, i32 1)
314 //
315 // 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:%[^,]+]])
316 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
317 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
318 
319 // Loop header
320 // CHECK: [[O_LOOP1_BODY]]
321 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
322 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
323 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
324 
325 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
326 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
327 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
328   for (unsigned char i = '0' ; i <= '9'; ++i)
329     for (x = -10; x < 10; ++x) {
330 // CHECK: [[LOOP1_BODY]]
331 // Start of body: indices are calculated from IV:
332 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}}
333 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}}
334 // ... loop body ...
335 // End of body: store into a[i]:
336 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
337     a[i] = b[i] * c[i] * d[i];
338 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
339 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
340 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
341 // CHECK-NEXT: br label %{{.+}}
342   }
343 // CHECK: [[LOOP1_END]]
344 // CHECK: [[O_LOOP1_END]]
345 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
346 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
347 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
348 // CHECK: ret void
349 }
350 
351 // TERM_DEBUG-LABEL: foo
352 int foo() {return 0;};
353 
354 // TERM_DEBUG-LABEL: parallel_for
355 void parallel_for(float *a) {
356 #pragma omp parallel for schedule(static, 5)
357   // TERM_DEBUG-NOT: __kmpc_global_thread_num
358   // TERM_DEBUG:     call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC_START:![0-9]+]]
359   // TERM_DEBUG:     invoke i32 {{.*}}foo{{.*}}()
360   // TERM_DEBUG:     unwind label %[[TERM_LPAD:.+]],
361   // TERM_DEBUG-NOT: __kmpc_global_thread_num
362   // TERM_DEBUG:     call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC_END:![0-9]+]]
363   // TERM_DEBUG:     call {{.+}} @__kmpc_cancel_barrier({{.+}}), !dbg [[DBG_LOC_CANCEL:![0-9]+]]
364   // TERM_DEBUG:     [[TERM_LPAD]]
365   // TERM_DEBUG:     call void @__clang_call_terminate
366   // TERM_DEBUG:     unreachable
367   for (unsigned i = 131071; i <= 2147483647; i += 127)
368     a[i] += foo();
369 }
370 // Check source line corresponds to "#pragma omp parallel for schedule(static, 5)" above:
371 // TERM_DEBUG-DAG: [[DBG_LOC_START]] = !DILocation(line: [[@LINE-4]],
372 // TERM_DEBUG-DAG: [[DBG_LOC_END]] = !DILocation(line: [[@LINE-16]],
373 // TERM_DEBUG-DAG: [[DBG_LOC_CANCEL]] = !DILocation(line: [[@LINE-17]],
374 
375 #endif // HEADER
376 
377