1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s
2 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %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 - | FileCheck %s
4 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp -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-DAG: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* }
11 // CHECK-DAG: [[CAP_TY:%.+]] = type { i8* }
12 
13 // CHECK-LABEL: with_var_schedule
14 void with_var_schedule() {
15   double a = 5;
16 // CHECK: [[CHUNK_SIZE:%.+]] = fptosi double %{{.+}}to i8
17 // CHECK: store i8 %{{.+}}, i8* [[CHUNK:%.+]],
18 // CHECK: [[CHUNK_REF:%.+]] = getelementptr inbounds [[CAP_TY]], [[CAP_TY]]* [[CAP_ARG:%.+]], i{{.+}} 0, i{{.+}} 0
19 // CHECK: store i8* [[CHUNK]], i8** [[CHUNK_REF]],
20 // CHECK: [[BITCAST:%.+]] = bitcast [[CAP_TY]]* [[CAP_ARG]] to i8*
21 // CHECK: call void {{.+}} @__kmpc_fork_call({{.+}}, i8* [[BITCAST]])
22 
23 // CHECK: [[CHUNK_REF:%.+]] = getelementptr inbounds [[CAP_TY]], [[CAP_TY]]* %{{.+}}, i{{.+}} 0, i{{.+}} 0
24 // CHECK: [[CHUNK:%.+]] = load i8*, i8** [[CHUNK_REF]],
25 // CHECK: [[CHUNK_VAL:%.+]] = load i8, i8* [[CHUNK]],
26 // CHECK: [[CHUNK_SIZE:%.+]] = sext i8 [[CHUNK_VAL]] to i64
27 // CHECK: call void @__kmpc_for_static_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC:@[^,]+]], i32 [[GTID:%[^,]+]], i32 33, i32* [[IS_LAST:%[^,]+]], i64* [[OMP_LB:%[^,]+]], i64* [[OMP_UB:%[^,]+]], i64* [[OMP_ST:%[^,]+]], i64 1, i64 [[CHUNK_SIZE]])
28 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
29 // CHECK: __kmpc_cancel_barrier
30 #pragma omp parallel for schedule(static, char(a))
31   for (unsigned long long i = 1; i < 2; ++i) {
32   }
33 }
34 
35 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
36 void without_schedule_clause(float *a, float *b, float *c, float *d) {
37   #pragma omp parallel for
38 // 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* %{{.+}})
39 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
40 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
41 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
42 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
43 // 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)
44 // UB = min(UB, GlobalUB)
45 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
46 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
47 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
48 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
49 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
50 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
51 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
52 // Loop header
53 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
54 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
55 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
56 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
57   for (int i = 33; i < 32000000; i += 7) {
58 // CHECK: [[LOOP1_BODY]]
59 // Start of body: calculate i from IV:
60 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
61 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7
62 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]]
63 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
64 // ... loop body ...
65 // End of body: store into a[i]:
66 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
67     a[i] = b[i] * c[i] * d[i];
68 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
69 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
70 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
71 // CHECK-NEXT: br label %{{.+}}
72   }
73 // CHECK: [[LOOP1_END]]
74 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
75 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
76 // CHECK: ret void
77 }
78 
79 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
80 void static_not_chunked(float *a, float *b, float *c, float *d) {
81   #pragma omp parallel for schedule(static)
82 // 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* %{{.+}})
83 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
84 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
85 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
86 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
87 // 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)
88 // UB = min(UB, GlobalUB)
89 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
90 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
91 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
92 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
93 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
94 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
95 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
96 // Loop header
97 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
98 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
99 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
100 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
101   for (int i = 32000000; i > 33; i += -7) {
102 // CHECK: [[LOOP1_BODY]]
103 // Start of body: calculate i from IV:
104 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
105 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7
106 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]]
107 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
108 // ... loop body ...
109 // End of body: store into a[i]:
110 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
111     a[i] = b[i] * c[i] * d[i];
112 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
113 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
114 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
115 // CHECK-NEXT: br label %{{.+}}
116   }
117 // CHECK: [[LOOP1_END]]
118 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
119 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
120 // CHECK: ret void
121 }
122 
123 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
124 void static_chunked(float *a, float *b, float *c, float *d) {
125   #pragma omp parallel for schedule(static, 5)
126 // 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* %{{.+}})
127 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
128 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
129 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
130 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
131 // 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)
132 // UB = min(UB, GlobalUB)
133 // CHECK: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
134 // CHECK-NEXT: [[UBCMP:%.+]] = icmp ugt i32 [[UB]], 16908288
135 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
136 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 16908288, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
137 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
138 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
139 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
140 
141 // Outer loop header
142 // CHECK: [[O_IV:%.+]] = load i32, i32* [[OMP_IV]]
143 // CHECK-NEXT: [[O_UB:%.+]] = load i32, i32* [[OMP_UB]]
144 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ule i32 [[O_IV]], [[O_UB]]
145 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
146 
147 // Loop header
148 // CHECK: [[O_LOOP1_BODY]]
149 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
150 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
151 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i32 [[IV]], [[UB]]
152 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
153   for (unsigned i = 131071; i <= 2147483647; i += 127) {
154 // CHECK: [[LOOP1_BODY]]
155 // Start of body: calculate i from IV:
156 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
157 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i32 [[IV1_1]], 127
158 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 131071, [[CALC_I_1]]
159 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
160 // ... loop body ...
161 // End of body: store into a[i]:
162 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
163     a[i] = b[i] * c[i] * d[i];
164 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
165 // CHECK-NEXT: [[ADD1_2:%.+]] = add i32 [[IV1_2]], 1
166 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
167 // CHECK-NEXT: br label %{{.+}}
168   }
169 // CHECK: [[LOOP1_END]]
170 // Update the counters, adding stride
171 // CHECK:  [[LB:%.+]] = load i32, i32* [[OMP_LB]]
172 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]]
173 // CHECK-NEXT: [[ADD_LB:%.+]] = add i32 [[LB]], [[ST]]
174 // CHECK-NEXT: store i32 [[ADD_LB]], i32* [[OMP_LB]]
175 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
176 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]]
177 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]]
178 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]]
179 
180 // CHECK: [[O_LOOP1_END]]
181 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
182 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
183 // CHECK: ret void
184 }
185 
186 // CHECK-LABEL: define {{.*void}} @{{.*}}dynamic1{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
187 void dynamic1(float *a, float *b, float *c, float *d) {
188   #pragma omp parallel for schedule(dynamic)
189 // 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* %{{.+}})
190 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
191 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
192 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
193 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
194 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 35, i64 0, i64 16908287, i64 1, i64 1)
195 //
196 // 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:%[^,]+]])
197 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
198 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
199 
200 // Loop header
201 // CHECK: [[O_LOOP1_BODY]]
202 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
203 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
204 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
205 
206 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
207 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i64 [[IV]], [[UB]]
208 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
209   for (unsigned long long i = 131071; i < 2147483647; i += 127) {
210 // CHECK: [[LOOP1_BODY]]
211 // Start of body: calculate i from IV:
212 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]]
213 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127
214 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]]
215 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]]
216 // ... loop body ...
217 // End of body: store into a[i]:
218 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
219     a[i] = b[i] * c[i] * d[i];
220 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
221 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
222 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
223 // CHECK-NEXT: br label %{{.+}}
224   }
225 // CHECK: [[LOOP1_END]]
226 // CHECK: [[O_LOOP1_END]]
227 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
228 // CHECK: ret void
229 }
230 
231 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
232 void guided7(float *a, float *b, float *c, float *d) {
233   #pragma omp parallel for schedule(guided, 7)
234 // 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* %{{.+}})
235 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
236 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
237 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
238 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
239 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 36, i64 0, i64 16908287, i64 1, i64 7)
240 //
241 // 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:%[^,]+]])
242 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
243 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
244 
245 // Loop header
246 // CHECK: [[O_LOOP1_BODY]]
247 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
248 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
249 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
250 
251 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
252 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i64 [[IV]], [[UB]]
253 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
254   for (unsigned long long i = 131071; i < 2147483647; i += 127) {
255 // CHECK: [[LOOP1_BODY]]
256 // Start of body: calculate i from IV:
257 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]]
258 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127
259 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]]
260 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]]
261 // ... loop body ...
262 // End of body: store into a[i]:
263 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
264     a[i] = b[i] * c[i] * d[i];
265 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
266 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
267 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
268 // CHECK-NEXT: br label %{{.+}}
269   }
270 // CHECK: [[LOOP1_END]]
271 // CHECK: [[O_LOOP1_END]]
272 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
273 // CHECK: ret void
274 }
275 
276 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
277 void test_auto(float *a, float *b, float *c, float *d) {
278   unsigned int x = 0;
279   unsigned int y = 0;
280   #pragma omp parallel for schedule(auto) collapse(2)
281 // 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* %{{.+}})
282 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
283 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
284 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
285 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
286 // 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)
287 //
288 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
289 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
290 // 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:%[^,]+]])
291 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
292 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
293 
294 // Loop header
295 // CHECK: [[O_LOOP1_BODY]]
296 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
297 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
298 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
299 
300 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
301 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]]
302 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
303 // FIXME: When the iteration count of some nested loop is not a known constant,
304 // we should pre-calculate it, like we do for the total number of iterations!
305   for (char i = static_cast<char>(y); i <= '9'; ++i)
306     for (x = 11; x > 0; --x) {
307 // CHECK: [[LOOP1_BODY]]
308 // Start of body: indices are calculated from IV:
309 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}}
310 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}}
311 // ... loop body ...
312 // End of body: store into a[i]:
313 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
314     a[i] = b[i] * c[i] * d[i];
315 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
316 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1
317 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
318 // CHECK-NEXT: br label %{{.+}}
319   }
320 // CHECK: [[LOOP1_END]]
321 // CHECK: [[O_LOOP1_END]]
322 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
323 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
324 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
325 // CHECK: ret void
326 }
327 
328 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
329 void runtime(float *a, float *b, float *c, float *d) {
330   int x = 0;
331   #pragma omp parallel for collapse(2) schedule(runtime)
332 // 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* %{{.+}})
333 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* [[GTID_PARAM_ADDR:%.+]], i32* %{{.+}}, %{{.+}}* %{{.+}})
334 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
335 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
336 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
337 // CHECK: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 37, i32 0, i32 199, i32 1, i32 1)
338 //
339 // 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:%[^,]+]])
340 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
341 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
342 
343 // Loop header
344 // CHECK: [[O_LOOP1_BODY]]
345 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
346 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
347 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
348 
349 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
350 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
351 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
352   for (unsigned char i = '0' ; i <= '9'; ++i)
353     for (x = -10; x < 10; ++x) {
354 // CHECK: [[LOOP1_BODY]]
355 // Start of body: indices are calculated from IV:
356 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}}
357 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}}
358 // ... loop body ...
359 // End of body: store into a[i]:
360 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
361     a[i] = b[i] * c[i] * d[i];
362 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
363 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
364 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
365 // CHECK-NEXT: br label %{{.+}}
366   }
367 // CHECK: [[LOOP1_END]]
368 // CHECK: [[O_LOOP1_END]]
369 // CHECK: [[GTID_REF:%.+]] = load i32*, i32** [[GTID_REF_ADDR]],
370 // CHECK: [[GTID:%.+]] = load i32, i32* [[GTID_REF]],
371 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
372 // CHECK: ret void
373 }
374 
375 // TERM_DEBUG-LABEL: foo
376 int foo() {return 0;};
377 
378 // TERM_DEBUG-LABEL: parallel_for
379 void parallel_for(float *a) {
380 #pragma omp parallel for schedule(static, 5)
381   // TERM_DEBUG-NOT: __kmpc_global_thread_num
382   // TERM_DEBUG:     call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC_START:![0-9]+]]
383   // TERM_DEBUG:     invoke i32 {{.*}}foo{{.*}}()
384   // TERM_DEBUG:     unwind label %[[TERM_LPAD:.+]],
385   // TERM_DEBUG-NOT: __kmpc_global_thread_num
386   // TERM_DEBUG:     call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC_END:![0-9]+]]
387   // TERM_DEBUG:     call {{.+}} @__kmpc_cancel_barrier({{.+}}), !dbg [[DBG_LOC_CANCEL:![0-9]+]]
388   // TERM_DEBUG:     [[TERM_LPAD]]
389   // TERM_DEBUG:     call void @__clang_call_terminate
390   // TERM_DEBUG:     unreachable
391   for (unsigned i = 131071; i <= 2147483647; i += 127)
392     a[i] += foo();
393 }
394 // Check source line corresponds to "#pragma omp parallel for schedule(static, 5)" above:
395 // TERM_DEBUG-DAG: [[DBG_LOC_START]] = !DILocation(line: [[@LINE-4]],
396 // TERM_DEBUG-DAG: [[DBG_LOC_END]] = !DILocation(line: [[@LINE-16]],
397 // TERM_DEBUG-DAG: [[DBG_LOC_CANCEL]] = !DILocation(line: [[@LINE-17]],
398 
399 #endif // HEADER
400 
401