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 -debug-info-kind=line-tables-only -x c++ -emit-llvm %s -o - | FileCheck %s --check-prefix=TERM_DEBUG
5 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -O1 -fopenmp -emit-llvm %s -o - | FileCheck %s --check-prefix=CLEANUP
6 
7 // 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
8 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
9 // 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
10 // 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
11 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -O1 -fopenmp-simd -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
12 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
13 // expected-no-diagnostics
14 
15 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck --check-prefix=OMP5 %s
16 // RUN: %clang_cc1 -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
17 // RUN: %clang_cc1 -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix=OMP5 %s
18 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=50 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
19 #ifndef HEADER
20 #define HEADER
21 
22 // CHECK-DAG: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* }
23 // CHECK-DAG: [[LOOP_LOC:@.+]] = private unnamed_addr global [[IDENT_T_TY]] { i32 0, i32 514, i32 0, i32 0, i8*
24 
25 // CHECK-LABEL: with_var_schedule
26 void with_var_schedule() {
27   double a = 5;
28 // CHECK: [[CHUNK_SIZE:%.+]] = fptosi double %{{.+}}to i8
29 // CHECK: store i8 %{{.+}}, i8* [[CHUNK:%.+]],
30 // CHECK: [[VAL:%.+]] = load i8, i8* [[CHUNK]],
31 // CHECK: store i8 [[VAL]], i8*
32 // CHECK: [[CHUNK:%.+]] = load i64, i64* %
33 // CHECK: call void {{.+}} @__kmpc_fork_call({{.+}}, i64 [[CHUNK]])
34 
35 // CHECK: [[CHUNK_VAL:%.+]] = load i8, i8* %
36 // CHECK: [[CHUNK_SIZE:%.+]] = sext i8 [[CHUNK_VAL]] to i64
37 // CHECK: call void @__kmpc_for_static_init_8u([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID:%[^,]+]], i32 33, i32* [[IS_LAST:%[^,]+]], i64* [[OMP_LB:%[^,]+]], i64* [[OMP_UB:%[^,]+]], i64* [[OMP_ST:%[^,]+]], i64 1, i64 [[CHUNK_SIZE]])
38 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
39 #pragma omp parallel for schedule(static, char(a))
40   for (unsigned long long i = 1; i < 2; ++i) {
41   }
42 }
43 
44 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
45 void without_schedule_clause(float *a, float *b, float *c, float *d) {
46   #pragma omp parallel for
47 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*),
48 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}})
49 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
50 // 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)
51 // UB = min(UB, GlobalUB)
52 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
53 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
54 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
55 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
56 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
57 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
58 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
59 // Loop header
60 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
61 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
62 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
63 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
64   for (int i = 33; i < 32000000; i += 7) {
65 // CHECK: [[LOOP1_BODY]]
66 // Start of body: calculate i from IV:
67 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
68 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7
69 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]]
70 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
71 // ... loop body ...
72 // End of body: store into a[i]:
73 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
74     a[i] = b[i] * c[i] * d[i];
75 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
76 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
77 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
78 // CHECK-NEXT: br label %{{.+}}
79   }
80 // CHECK: [[LOOP1_END]]
81 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
82 // CHECK: ret void
83 }
84 
85 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
86 void static_not_chunked(float *a, float *b, float *c, float *d) {
87   #pragma omp parallel for schedule(static)
88 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*),
89 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}})
90 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
91 // 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)
92 // UB = min(UB, GlobalUB)
93 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
94 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
95 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
96 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
97 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
98 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
99 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
100 // Loop header
101 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
102 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
103 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
104 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
105   for (int i = 32000000; i > 33; i += -7) {
106 // CHECK: [[LOOP1_BODY]]
107 // Start of body: calculate i from IV:
108 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
109 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7
110 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]]
111 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
112 // ... loop body ...
113 // End of body: store into a[i]:
114 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
115     a[i] = b[i] * c[i] * d[i];
116 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
117 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
118 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
119 // CHECK-NEXT: br label %{{.+}}
120   }
121 // CHECK: [[LOOP1_END]]
122 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
123 // CHECK: ret void
124 }
125 
126 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
127 void static_chunked(float *a, float *b, float *c, float *d) {
128   #pragma omp parallel for schedule(static, 5)
129 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*),
130 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}})
131 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
132 // CHECK: call void @__kmpc_for_static_init_4u([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID:%.+]], i32 33, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 5)
133 // UB = min(UB, GlobalUB)
134 // CHECK: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
135 // CHECK-NEXT: [[UBCMP:%.+]] = icmp ugt i32 [[UB]], 16908288
136 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
137 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 16908288, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
138 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
139 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
140 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
141 
142 // Outer loop header
143 // CHECK: [[O_IV:%.+]] = load i32, i32* [[OMP_IV]]
144 // CHECK-NEXT: [[O_UB:%.+]] = load i32, i32* [[OMP_UB]]
145 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ule i32 [[O_IV]], [[O_UB]]
146 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
147 
148 // Loop header
149 // CHECK: [[O_LOOP1_BODY]]
150 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
151 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
152 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i32 [[IV]], [[UB]]
153 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
154   for (unsigned i = 131071; i <= 2147483647; i += 127) {
155 // CHECK: [[LOOP1_BODY]]
156 // Start of body: calculate i from IV:
157 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
158 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i32 [[IV1_1]], 127
159 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 131071, [[CALC_I_1]]
160 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
161 // ... loop body ...
162 // End of body: store into a[i]:
163 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
164     a[i] = b[i] * c[i] * d[i];
165 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
166 // CHECK-NEXT: [[ADD1_2:%.+]] = add i32 [[IV1_2]], 1
167 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
168 // CHECK-NEXT: br label %{{.+}}
169   }
170 // CHECK: [[LOOP1_END]]
171 // Update the counters, adding stride
172 // CHECK:  [[LB:%.+]] = load i32, i32* [[OMP_LB]]
173 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]]
174 // CHECK-NEXT: [[ADD_LB:%.+]] = add i32 [[LB]], [[ST]]
175 // CHECK-NEXT: store i32 [[ADD_LB]], i32* [[OMP_LB]]
176 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
177 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]]
178 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]]
179 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]]
180 
181 // CHECK: [[O_LOOP1_END]]
182 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], 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 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*),
190 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}})
191 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
192 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID:%.+]], i32 35, i64 0, i64 16908287, i64 1, i64 1)
193 //
194 // 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:%[^,]+]])
195 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
196 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
197 
198 // Loop header
199 // CHECK: [[O_LOOP1_BODY]]
200 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
201 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
202 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
203 
204 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
205 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1
206 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]]
207 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
208   for (unsigned long long i = 131071; i < 2147483647; i += 127) {
209 // CHECK: [[LOOP1_BODY]]
210 // Start of body: calculate i from IV:
211 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]]
212 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127
213 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]]
214 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]]
215 // ... loop body ...
216 // End of body: store into a[i]:
217 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
218     a[i] = b[i] * c[i] * d[i];
219 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
220 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
221 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
222 // CHECK-NEXT: br label %{{.+}}
223   }
224 // CHECK: [[LOOP1_END]]
225 // CHECK: [[O_LOOP1_END]]
226 // CHECK: ret void
227 }
228 
229 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
230 void guided7(float *a, float *b, float *c, float *d) {
231   #pragma omp parallel for schedule(guided, 7)
232 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*),
233 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}})
234 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
235 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID:%.+]], i32 36, i64 0, i64 16908287, i64 1, i64 7)
236 //
237 // 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:%[^,]+]])
238 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
239 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
240 
241 // Loop header
242 // CHECK: [[O_LOOP1_BODY]]
243 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
244 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
245 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
246 
247 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
248 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1
249 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]]
250 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
251   for (unsigned long long i = 131071; i < 2147483647; i += 127) {
252 // CHECK: [[LOOP1_BODY]]
253 // Start of body: calculate i from IV:
254 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]]
255 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127
256 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]]
257 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]]
258 // ... loop body ...
259 // End of body: store into a[i]:
260 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
261     a[i] = b[i] * c[i] * d[i];
262 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
263 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
264 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
265 // CHECK-NEXT: br label %{{.+}}
266   }
267 // CHECK: [[LOOP1_END]]
268 // CHECK: [[O_LOOP1_END]]
269 // CHECK: ret void
270 }
271 
272 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
273 void test_auto(float *a, float *b, float *c, float *d) {
274   unsigned int x = 0;
275   unsigned int y = 0;
276   #pragma omp parallel for schedule(auto) collapse(2)
277 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*),
278 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}})
279 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
280 // 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)
281 //
282 // 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:%[^,]+]])
283 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
284 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
285 
286 // Loop header
287 // CHECK: [[O_LOOP1_BODY]]
288 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
289 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
290 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
291 
292 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
293 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]]
294 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
295 // FIXME: When the iteration count of some nested loop is not a known constant,
296 // we should pre-calculate it, like we do for the total number of iterations!
297   for (char i = static_cast<char>(y); i <= '9'; ++i)
298     for (x = 11; x > 0; --x) {
299 // CHECK: [[LOOP1_BODY]]
300 // Start of body: indices are calculated from IV:
301 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}}
302 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}}
303 // ... loop body ...
304 // End of body: store into a[i]:
305 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
306     a[i] = b[i] * c[i] * d[i];
307 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
308 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1
309 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
310 // CHECK-NEXT: br label %{{.+}}
311   }
312 // CHECK: [[LOOP1_END]]
313 // CHECK: [[O_LOOP1_END]]
314 // CHECK: ret void
315 }
316 
317 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
318 void runtime(float *a, float *b, float *c, float *d) {
319   int x = 0;
320   #pragma omp parallel for collapse(2) schedule(runtime)
321 // CHECK: call void ([[IDENT_T_TY]]*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**, float**, float**, float**)* [[OMP_PARALLEL_FUNC:@.+]] to void (i32*, i32*, ...)*),
322 // CHECK: define internal void [[OMP_PARALLEL_FUNC]](i32* noalias [[GTID_PARAM_ADDR:%.+]], i32* noalias %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}}, float** dereferenceable(8) %{{.+}})
323 // CHECK: store i32* [[GTID_PARAM_ADDR]], i32** [[GTID_REF_ADDR:%.+]],
324 // CHECK: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID:%.+]], i32 37, i32 0, i32 199, i32 1, i32 1)
325 //
326 // 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:%[^,]+]])
327 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
328 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
329 
330 // Loop header
331 // CHECK: [[O_LOOP1_BODY]]
332 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
333 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
334 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
335 
336 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
337 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
338 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
339   for (unsigned char i = '0' ; i <= '9'; ++i)
340     for (x = -10; x < 10; ++x) {
341 // CHECK: [[LOOP1_BODY]]
342 // Start of body: indices are calculated from IV:
343 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}}
344 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}}
345 // ... loop body ...
346 // End of body: store into a[i]:
347 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
348     a[i] = b[i] * c[i] * d[i];
349 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
350 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
351 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
352 // CHECK-NEXT: br label %{{.+}}
353   }
354 // CHECK: [[LOOP1_END]]
355 // CHECK: [[O_LOOP1_END]]
356 // CHECK: ret void
357 }
358 
359 // TERM_DEBUG-LABEL: foo
360 int foo() {return 0;};
361 
362 // TERM_DEBUG-LABEL: parallel_for
363 // CLEANUP: parallel_for
364 void parallel_for(float *a, const int n) {
365   float arr[n];
366 #pragma omp parallel for schedule(static, 5) private(arr) default(none) firstprivate(n) shared(a)
367   // TERM_DEBUG-NOT: __kmpc_global_thread_num
368   // TERM_DEBUG:     call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC_START:![0-9]+]]
369   // TERM_DEBUG:     invoke i32 {{.*}}foo{{.*}}()
370   // TERM_DEBUG:     unwind label %[[TERM_LPAD:.+]],
371   // TERM_DEBUG-NOT: __kmpc_global_thread_num
372   // TERM_DEBUG:     call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC_END:![0-9]+]]
373   // TERM_DEBUG:     [[TERM_LPAD]]
374   // TERM_DEBUG:     call void @__clang_call_terminate
375   // TERM_DEBUG:     unreachable
376   // CLEANUP-NOT: __kmpc_global_thread_num
377   // CLEANUP:     call void @__kmpc_for_static_init_4u({{.+}})
378   // CLEANUP:     call void @__kmpc_for_static_fini({{.+}})
379   for (unsigned i = 131071; i <= 2147483647; i += 127)
380     a[i] += foo() + arr[i] + n;
381 }
382 // Check source line corresponds to "#pragma omp parallel for schedule(static, 5)" above:
383 // TERM_DEBUG-DAG: [[DBG_LOC_START]] = !DILocation(line: [[@LINE-4]],
384 // TERM_DEBUG-DAG: [[DBG_LOC_END]] = !DILocation(line: [[@LINE-18]],
385 
386 #ifdef OMP5
387 // OMP5-DAG: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* }
388 // OMP5-DAG: [[LOOP_LOC:@.+]] = private unnamed_addr global [[IDENT_T_TY]] { i32 0, i32 514, i32 0, i32 0, i8*
389 
390 // OMP5-LABEL: increment
391 int increment () {
392 // OMP5: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
393   #pragma omp for
394 // Determine UB = min(UB, GlobalUB)
395 // OMP5: 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)
396 // OMP5-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
397 // OMP5-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4
398 // OMP5-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
399 // OMP5: [[UBRESULT:%.+]] = phi i32 [ 4, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
400 // OMP5-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
401 // OMP5-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
402 // OMP5-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
403 // OMP5-NEXT: br label %[[LOOP1_HEAD:.+]]
404 
405 // Loop header
406 // OMP5: [[LOOP1_HEAD]]
407 // OMP5: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
408 // OMP5-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
409 // OMP5-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
410 // OMP5-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
411 
412   for (int i = 0 ; i != 5; ++i)
413 // Start of body: calculate i from IV:
414 // OMP5: [[LOOP1_BODY]]
415 // OMP5: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
416 // OMP5-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 1
417 // OMP5-NEXT: [[CALC_I_2:%.+]] = add nsw i32 0, [[CALC_I_1]]
418 // OMP5-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
419 // OMP5: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
420 // OMP5-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
421 // OMP5-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
422 // OMP5-NEXT: br label %[[LOOP1_HEAD]]
423     ;
424 // OMP5: [[LOOP1_END]]
425 // OMP5: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
426 // OMP5: __kmpc_barrier
427   return 0;
428 // OMP5: ret i32 0
429 }
430 
431 // OMP5-LABEL: decrement_nowait
432 int decrement_nowait () {
433 // OMP5: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
434   #pragma omp for nowait
435 // Determine UB = min(UB, GlobalUB)
436 // OMP5: 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)
437 // OMP5-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
438 // OMP5-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4
439 // OMP5-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
440 // OMP5: [[UBRESULT:%.+]] = phi i32 [ 4, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
441 // OMP5-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
442 // OMP5-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
443 // OMP5-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
444 // OMP5-NEXT: br label %[[LOOP1_HEAD:.+]]
445 
446 // Loop header
447 // OMP5: [[LOOP1_HEAD]]
448 // OMP5: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
449 // OMP5-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
450 // OMP5-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
451 // OMP5-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
452   for (int j = 5 ; j != 0; --j)
453 // Start of body: calculate i from IV:
454 // OMP5: [[LOOP1_BODY]]
455 // OMP5: [[IV2_1:%.+]] = load i32, i32* [[OMP_IV]]
456 // OMP5-NEXT: [[CALC_II_1:%.+]] = mul nsw i32 [[IV2_1]], 1
457 // OMP5-NEXT: [[CALC_II_2:%.+]] = sub nsw i32 5, [[CALC_II_1]]
458 // OMP5-NEXT: store i32 [[CALC_II_2]], i32* [[LC_I:.+]]
459 // OMP5: [[IV2_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
460 // OMP5-NEXT: [[ADD2_2:%.+]] = add nsw i32 [[IV2_2]], 1
461 // OMP5-NEXT: store i32 [[ADD2_2]], i32* [[OMP_IV]]
462 // OMP5-NEXT: br label %[[LOOP1_HEAD]]
463     ;
464 // OMP5: [[LOOP1_END]]
465 // OMP5: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
466 // OMP5-NOT: __kmpc_barrier
467   return 0;
468 // OMP5: ret i32 0
469 }
470 #endif
471 
472 #endif // HEADER
473 
474