1 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -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 -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=OMP5
3 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
4 // 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=OMP5
5 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
6 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -fopenmp-version=45 -emit-llvm -o - | FileCheck %s --check-prefix=CHECK --check-prefix=OMP45
7 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp -fexceptions -fcxx-exceptions -debug-info-kind=line-tables-only -gno-column-info -x c++ -emit-llvm %s -o - | FileCheck %s --check-prefix=TERM_DEBUG
8 // 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
9
10 // 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
11 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
12 // 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
13 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
14 // 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
15 // 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
16 // 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
17 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
18 //
19 // expected-no-diagnostics
20 #ifndef HEADER
21 #define HEADER
22 // PROF-INSTR-PATH: constant [25 x i8] c"for_codegen-test.profraw\00"
23
24 // CHECK: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* }
25 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8*
26 // CHECK-DAG: [[LOOP_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 514, i32 0, i32 0, i8*
27 // CHECK-DAG: [[I:@.+]] ={{.*}} global i8 1,
28 // CHECK-DAG: [[J:@.+]] ={{.*}} global i8 2,
29 // CHECK-DAG: [[K:@.+]] ={{.*}} global i8 3,
30
31 // CHECK-LABEL: loop_with_counter_collapse
loop_with_counter_collapse()32 void loop_with_counter_collapse() {
33 // Captured initializations.
34 // CHECK: store i32 0, i32* [[I_TMP:%.+]],
35 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]],
36 // CHECK: store i32 [[VAL]], i32* [[J_LB_MIN:%.+]],
37 // CHECK: store i32 3, i32* [[I_TMP]],
38 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]],
39 // CHECK: store i32 [[VAL]], i32* [[J_LB_MAX:%.+]],
40 // CHECK: [[J_LB_MIN_VAL:%.+]] = load i32, i32* [[J_LB_MIN]],
41 // CHECK: [[J_LB_MAX_VAL:%.+]] = load i32, i32* [[J_LB_MAX]],
42 // CHECK: [[CMP:%.+]] = icmp slt i32 [[J_LB_MIN_VAL]], [[J_LB_MAX_VAL]]
43 // CHECK: [[BOOL:%.+]] = zext i1 [[CMP]] to i8
44 // CHECK: store i8 [[BOOL]], i8* [[J_LB_CMP:%.+]],
45 // CHECK: store i32 0, i32* [[I_TMP]],
46 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]],
47 // CHECK: [[J_UB_MIN_VAL:%.+]] = add nsw i32 4, [[VAL]]
48 // CHECK: store i32 [[J_UB_MIN_VAL]], i32* [[J_UB_MIN:%.+]],
49 // CHECK: store i32 3, i32* [[I_TMP]],
50 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]],
51 // CHECK: [[J_UB_MAX_VAL:%.+]] = add nsw i32 4, [[VAL]]
52 // CHECK: store i32 [[J_UB_MAX_VAL]], i32* [[J_UB_MAX:%.+]],
53 // CHECK: [[J_UB_MIN_VAL:%.+]] = load i32, i32* [[J_UB_MIN]],
54 // CHECK: [[J_UB_MAX_VAL:%.+]] = load i32, i32* [[J_UB_MAX]],
55 // CHECK: [[CMP:%.+]] = icmp sgt i32 [[J_UB_MIN_VAL]], [[J_UB_MAX_VAL]]
56 // CHECK: [[BOOL:%.+]] = zext i1 [[CMP]] to i8
57 // CHECK: store i8 [[BOOL]], i8* [[J_UB_CMP:%.+]],
58 // CHECK: [[J_UB_CMP_VAL:%.+]] = load i8, i8* [[J_UB_CMP]],
59 // CHECK: [[BOOL:%.+]] = trunc i8 [[J_UB_CMP_VAL]] to i1
60 // CHECK: br i1 [[BOOL]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]]
61 // CHECK: [[TRUE]]:
62 // CHECK: [[J_UB_MIN_VAL:%.+]] = load i32, i32* [[J_UB_MIN]],
63 // CHECK: br label %[[EXIT:[^,]+]]
64 // CHECK: [[FALSE]]:
65 // CHECK: [[J_UB_MAX_VAL:%.+]] = load i32, i32* [[J_UB_MAX]],
66 // CHECK: br label %[[EXIT]]
67 // CHECK: [[EXIT]]:
68 // CHECK: [[J_UB_VAL:%.+]] = phi i32 [ [[J_UB_MIN_VAL]], %[[TRUE]] ], [ [[J_UB_MAX_VAL]], %[[FALSE]] ]
69 // CHECK: store i32 [[J_UB_VAL]], i32* [[J_UB:%.+]],
70 // CHECK: [[J_LB_CMP_VAL:%.+]] = load i8, i8* [[J_LB_CMP]],
71 // CHECK: [[BOOL:%.+]] = trunc i8 [[J_LB_CMP_VAL]] to i1
72 // CHECK: br i1 [[BOOL]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]]
73 // CHECK: [[TRUE]]:
74 // CHECK: [[J_LB_MIN_VAL:%.+]] = load i32, i32* [[J_LB_MIN]],
75 // CHECK: br label %[[EXIT:[^,]+]]
76 // CHECK: [[FALSE]]:
77 // CHECK: [[J_LB_MAX_VAL:%.+]] = load i32, i32* [[J_LB_MAX]],
78 // CHECK: br label %[[EXIT]]
79 // CHECK: [[EXIT]]:
80 // CHECK: [[J_LB_VAL:%.+]] = phi i32 [ [[J_LB_MIN_VAL]], %[[TRUE]] ], [ [[J_LB_MAX_VAL]], %[[FALSE]] ]
81 // CHECK: store i32 [[J_LB_VAL]], i32* [[J_LB:%.+]],
82 // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]],
83 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]],
84 // CHECK: [[SUB:%.+]] = sub i32 [[J_UB_VAL]], [[J_LB_VAL]]
85 // CHECK: [[SUB_ST:%.+]] = sub i32 [[SUB]], 1
86 // CHECK: [[ADD_ST:%.+]] = add i32 [[SUB_ST]], 1
87 // CHECK: [[DIV_ST:%.+]] = udiv i32 [[ADD_ST]], 1
88 // CHECK: [[CAST:%.+]] = zext i32 [[DIV_ST]] to i64
89 // CHECK: [[MUL:%.+]] = mul nsw i64 4, [[CAST]]
90 // CHECK: [[NUM_ITERS_VAL:%.+]] = sub nsw i64 [[MUL]], 1
91 // CHECK: store i64 [[NUM_ITERS_VAL]], i64* [[NUM_ITERS:%.+]],
92
93 // CHECK: store i64 0, i64* [[LB:%.+]],
94 // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, i64* [[NUM_ITERS]],
95 // CHECK: store i64 [[NUM_ITERS_VAL]], i64* [[UB:%.+]],
96 // CHECK: store i64 1, i64* [[STRIDE:%.+]],
97 // CHECK: store i32 0, i32* [[IS_LAST:%.+]],
98 // 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)
99 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]],
100 // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, i64* [[NUM_ITERS]],
101 // CHECK: [[CMP:%.+]] = icmp sgt i64 [[UB_VAL]], [[NUM_ITERS_VAL]]
102 // CHECK: br i1 [[CMP]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]]
103 // CHECK: [[TRUE]]:
104 // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, i64* [[NUM_ITERS]],
105 // CHECK: br label %[[DONE:[^,]+]]
106 // CHECK: [[FALSE]]:
107 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]],
108 // CHECK: br label %[[DONE]]
109 // CHECK: [[DONE]]:
110 // CHECK: [[TOP:%.+]] = phi i64 [ [[NUM_ITERS_VAL]], %[[TRUE]] ], [ [[UB_VAL]], %[[FALSE]] ]
111 // CHECK: store i64 [[TOP]], i64* [[UB]],
112 // CHECK: [[LB_VAL:%.+]] = load i64, i64* [[LB]],
113 // CHECK: store i64 [[LB_VAL]], i64* [[IV:%.+]],
114 // CHECK: br label %[[COND:[^,]+]]
115 // CHECK: [[COND]]:
116 // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]],
117 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]],
118 // CHECK: [[CMP:%.+]] = icmp sle i64 [[IV_VAL]], [[UB_VAL]]
119 // CHECK: br i1 [[CMP]], label %[[BODY:[^,]+]], label %[[CLEANUP:[^,]+]]
120 // LIFETIME: [[CLEANUP]]:
121 // LIFETIME: br label %[[CLEANUP:[^,]+]]
122 // CHECK: [[BODY]]:
123 // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]],
124 // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]],
125 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]],
126 // CHECK: [[SUB:%.+]] = sub i32 [[J_UB_VAL]], [[J_LB_VAL]]
127 // CHECK: [[SUB_ST:%.+]] = sub i32 [[SUB]], 1
128 // CHECK: [[ADD_ST:%.+]] = add i32 [[SUB_ST]], 1
129 // CHECK: [[DIV_ST:%.+]] = udiv i32 [[ADD_ST]], 1
130 // CHECK: [[MUL:%.+]] = mul i32 1, [[DIV_ST]]
131 // CHECK: [[CAST:%.+]] = zext i32 [[MUL]] to i64
132 // CHECK: [[DIV:%.+]] = sdiv i64 [[IV_VAL]], [[CAST]]
133 // CHECK: [[MUL:%.+]] = mul nsw i64 [[DIV]], 1
134 // CHECK: [[ADD:%.+]] = add nsw i64 0, [[MUL]]
135 // CHECK: [[CAST:%.+]] = trunc i64 [[ADD]] to i32
136 // CHECK: store i32 [[CAST]], i32* [[I_PRIV:%.+]],
137 // CHECK: [[I_VAL:%.+]] = load i32, i32* [[I_PRIV]],
138 // CHECK: [[CONV:%.+]] = sext i32 [[I_VAL]] to i64
139 // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]],
140 // CHECK: [[IV_VAL1:%.+]] = load i64, i64* [[IV]],
141 // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]],
142 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]],
143 // CHECK: [[SUB:%.+]] = sub i32 [[J_UB_VAL]], [[J_LB_VAL]]
144 // CHECK: [[SUB_ST:%.+]] = sub i32 [[SUB]], 1
145 // CHECK: [[ADD_ST:%.+]] = add i32 [[SUB_ST]], 1
146 // CHECK: [[DIV_ST:%.+]] = udiv i32 [[ADD_ST]], 1
147 // CHECK: [[MUL:%.+]] = mul i32 1, [[DIV_ST]]
148 // CHECK: [[CAST:%.+]] = zext i32 [[MUL]] to i64
149 // CHECK: [[DIV:%.+]] = sdiv i64 [[IV_VAL1]], [[CAST]]
150 // CHECK: [[J_UB_VAL:%.+]] = load i32, i32* [[J_UB]],
151 // CHECK: [[J_LB_VAL:%.+]] = load i32, i32* [[J_LB]],
152 // CHECK: [[SUB:%.+]] = sub i32 [[J_UB_VAL]], [[J_LB_VAL]]
153 // CHECK: [[SUB_ST:%.+]] = sub i32 [[SUB]], 1
154 // CHECK: [[ADD_ST:%.+]] = add i32 [[SUB_ST]], 1
155 // CHECK: [[DIV_ST:%.+]] = udiv i32 [[ADD_ST]], 1
156 // CHECK: [[MUL:%.+]] = mul i32 1, [[DIV_ST]]
157 // CHECK: [[CAST:%.+]] = zext i32 [[MUL]] to i64
158 // CHECK: [[MUL:%.+]] = mul nsw i64 [[DIV]], [[CAST]]
159 // CHECK: [[SUB:%.+]] = sub nsw i64 [[IV_VAL]], [[MUL]]
160 // CHECK: [[MUL:%.+]] = mul nsw i64 [[SUB:%.+]], 1
161 // CHECK: [[ADD:%.+]] = add nsw i64 [[CONV]], [[MUL]]
162 // CHECK: [[CAST:%.+]] = trunc i64 [[ADD]] to i32
163 // CHECK: store i32 [[CAST]], i32* [[J_PRIV:%.+]],
164
165 // Check that the loop variable is not out of its boundaries.
166 // CHECK: [[J_VAL:%.+]] = load i32, i32* [[J_PRIV]],
167 // CHECK: [[I_VAL:%.+]] = load i32, i32* [[I_PRIV]],
168 // CHECK: [[J_COND:%.+]] = add nsw i32 4, [[I_VAL]]
169 // CHECK: [[CMP:%.+]] = icmp slt i32 [[J_VAL]], [[J_COND]]
170 // CHECK: br i1 [[CMP]], label %[[NEXT:[^,]+]], label %[[BODY_CONT:[^,]+]]
171 // CHECK: [[NEXT]]:
172
173 // Main body is empty.
174 // CHECK: br label %[[BODY_CONT]]
175 // CHECK: [[BODY_CONT]]:
176 // CHECK: br label %[[INC:[^,]+]]
177 // CHECK: [[INC]]:
178 // CHECK: [[IV_VAL:%.+]] = load i64, i64* [[IV]],
179 // CHECK: [[ADD:%.+]] = add nsw i64 [[IV_VAL]], 1
180 // CHECK: store i64 [[ADD]], i64* [[IV]],
181 // CHECK: br label %[[COND]]
182 // CHECK: [[CLEANUP]]:
183 // CHECK: br label %[[EXIT:[^,]+]]
184 // CHECK: [[EXIT]]:
185 // CHECK: call void @__kmpc_for_static_fini(%struct.ident_t* @{{.+}}, i32 %{{.+}})
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 // LIFETIME: call void @llvm.lifetime.end
195 #pragma omp for collapse(2)
196 for (int i = 0; i < 4; i++) {
197 for (int j = i; j < 4 + i; j++) {
198 }
199 }
200 }
201
202 // CHECK-LABEL: loop_with_counter_collapse4
loop_with_counter_collapse4()203 void loop_with_counter_collapse4() {
204
205 // Check bounds calculation when collapse > 2
206 // CHECK: store i32 0, i32* [[I_TMP:%.+]],
207 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]],
208 // CHECK: store i32 [[VAL]], i32* [[K_LB_MIN:%.+]],
209 // CHECK: store i32 6, i32* [[I_TMP]],
210 // CHECK: [[VAL:%.+]] = load i32, i32* [[I_TMP]],
211 // CHECK: store i32 [[VAL]], i32* [[K_LB_MAX:%.+]],
212 #pragma omp for collapse(4)
213 for (int i = 0; i < 7; i++) {
214 for (int j = 0; j < 11; j++) {
215 for (int k = i; k < 7; k++) {
216 for (int l = 0; l < 11; l++) {
217 }
218 }
219 }
220 }
221 }
222
223 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
without_schedule_clause(float * a,float * b,float * c,float * d)224 void without_schedule_clause(float *a, float *b, float *c, float *d) {
225 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
226 #pragma omp for nowait
227 // 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)
228 // UB = min(UB, GlobalUB)
229 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
230 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
231 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
232 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
233 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
234 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
235 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
236 // Loop header
237 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
238 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
239 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
240 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
241 for (int i = 33; i < 32000000; i += 7) {
242 // CHECK: [[LOOP1_BODY]]
243 // Start of body: calculate i from IV:
244 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
245 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7
246 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]]
247 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
248 // ... loop body ...
249 // End of body: store into a[i]:
250 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
251 // CHECK-NOT: !llvm.access.group
252 a[i] = b[i] * c[i] * d[i];
253 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
254 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
255 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
256 // CHECK-NEXT: br label %{{.+}}
257 }
258 // CHECK: [[LOOP1_END]]
259 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
260 // CHECK-NOT: __kmpc_barrier
261 // CHECK: ret void
262 }
263
264 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
static_not_chunked(float * a,float * b,float * c,float * d)265 void static_not_chunked(float *a, float *b, float *c, float *d) {
266 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
267 #pragma omp for schedule(static)
268 // 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)
269 // UB = min(UB, GlobalUB)
270 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
271 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
272 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
273 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
274 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
275 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
276 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
277 // Loop header
278 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
279 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
280 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
281 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
282 for (int i = 32000000; i > 33; i += -7) {
283 // CHECK: [[LOOP1_BODY]]
284 // Start of body: calculate i from IV:
285 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
286 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7
287 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]]
288 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
289 // ... loop body ...
290 // End of body: store into a[i]:
291 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
292 // CHECK-NOT: !llvm.access.group
293 a[i] = b[i] * c[i] * d[i];
294 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
295 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
296 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
297 // CHECK-NEXT: br label %{{.+}}
298 }
299 // CHECK: [[LOOP1_END]]
300 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
301 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
302 // CHECK: ret void
303 }
304
305 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
static_chunked(float * a,float * b,float * c,float * d)306 void static_chunked(float *a, float *b, float *c, float *d) {
307 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
308 #pragma omp for schedule(monotonic: static, 5)
309 // 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)
310 // UB = min(UB, GlobalUB)
311 // CHECK: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
312 // CHECK-NEXT: [[UBCMP:%.+]] = icmp ugt i32 [[UB]], 16908288
313 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
314 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 16908288, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
315 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
316 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
317 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
318
319 // Outer loop header
320 // CHECK: [[O_IV:%.+]] = load i32, i32* [[OMP_IV]]
321 // CHECK-NEXT: [[O_UB:%.+]] = load i32, i32* [[OMP_UB]]
322 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ule i32 [[O_IV]], [[O_UB]]
323 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
324
325 // Loop header
326 // CHECK: [[O_LOOP1_BODY]]
327 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
328 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
329 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i32 [[IV]], [[UB]]
330 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
331 for (unsigned i = 131071; i <= 2147483647; i += 127) {
332 // CHECK: [[LOOP1_BODY]]
333 // Start of body: calculate i from IV:
334 // CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
335 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i32 [[IV1_1]], 127
336 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 131071, [[CALC_I_1]]
337 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]]
338 // ... loop body ...
339 // End of body: store into a[i]:
340 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
341 // CHECK-NOT: !llvm.access.group
342 a[i] = b[i] * c[i] * d[i];
343 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
344 // CHECK-NEXT: [[ADD1_2:%.+]] = add i32 [[IV1_2]], 1
345 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
346 // CHECK-NEXT: br label %{{.+}}
347 }
348 // CHECK: [[LOOP1_END]]
349 // Update the counters, adding stride
350 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
351 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]]
352 // CHECK-NEXT: [[ADD_LB:%.+]] = add i32 [[LB]], [[ST]]
353 // CHECK-NEXT: store i32 [[ADD_LB]], i32* [[OMP_LB]]
354 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
355 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]]
356 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]]
357 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]]
358
359 // CHECK: [[O_LOOP1_END]]
360 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[LOOP_LOC]], i32 [[GTID]])
361 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
362 // CHECK: ret void
363 }
364
365 // CHECK-LABEL: define {{.*void}} @{{.*}}dynamic1{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
dynamic1(float * a,float * b,float * c,float * d)366 void dynamic1(float *a, float *b, float *c, float *d) {
367 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
368 #pragma omp for schedule(nonmonotonic: dynamic)
369 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741859, i64 0, i64 16908287, i64 1, i64 1)
370 //
371 // 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:%[^,]+]])
372 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
373 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
374
375 // Loop header
376 // CHECK: [[O_LOOP1_BODY]]
377 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
378 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
379 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
380
381 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
382 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1
383 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]]
384 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
385 for (unsigned long long i = 131071; i < 2147483647; i += 127) {
386 // CHECK: [[LOOP1_BODY]]
387 // Start of body: calculate i from IV:
388 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]]
389 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127
390 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]]
391 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]]
392 // ... loop body ...
393 // End of body: store into a[i]:
394 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.access.group
395 a[i] = b[i] * c[i] * d[i];
396 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
397 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
398 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
399 // CHECK-NEXT: br label %{{.+}}
400 }
401 // CHECK: [[LOOP1_END]]
402 // CHECK: [[O_LOOP1_END]]
403 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
404 // CHECK: ret void
405 }
406
407 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
guided7(float * a,float * b,float * c,float * d)408 void guided7(float *a, float *b, float *c, float *d) {
409 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
410 #pragma omp for schedule(guided, 7)
411 // OMP45: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 36, i64 0, i64 16908287, i64 1, i64 7)
412 // OMP5: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741860, i64 0, i64 16908287, i64 1, i64 7)
413 //
414 // 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:%[^,]+]])
415 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
416 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
417
418 // Loop header
419 // CHECK: [[O_LOOP1_BODY]]
420 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
421 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
422 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
423
424 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
425 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1
426 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]]
427 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
428 for (unsigned long long i = 131071; i < 2147483647; i += 127) {
429 // CHECK: [[LOOP1_BODY]]
430 // Start of body: calculate i from IV:
431 // CHECK: [[IV1_1:%.+]] = load i64, i64* [[OMP_IV]]
432 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127
433 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]]
434 // CHECK-NEXT: store i64 [[CALC_I_2]], i64* [[LC_I:.+]]
435 // ... loop body ...
436 // End of body: store into a[i]:
437 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}!llvm.access.group
438 a[i] = b[i] * c[i] * d[i];
439 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
440 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
441 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
442 // CHECK-NEXT: br label %{{.+}}
443 }
444 // CHECK: [[LOOP1_END]]
445 // CHECK: [[O_LOOP1_END]]
446 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
447 // CHECK: ret void
448 }
449
450 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
test_auto(float * a,float * b,float * c,float * d)451 void test_auto(float *a, float *b, float *c, float *d) {
452 unsigned int x = 0;
453 unsigned int y = 0;
454 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
455 #pragma omp for schedule(auto) collapse(2)
456 // 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)
457 // 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)
458 //
459 // 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:%[^,]+]])
460 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
461 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
462
463 // Loop header
464 // CHECK: [[O_LOOP1_BODY]]
465 // CHECK: [[LB:%.+]] = load i64, i64* [[OMP_LB]]
466 // CHECK-NEXT: store i64 [[LB]], i64* [[OMP_IV:[^,]+]]
467 // CHECK: [[IV:%.+]] = load i64, i64* [[OMP_IV]]
468
469 // CHECK-NEXT: [[UB:%.+]] = load i64, i64* [[OMP_UB]]
470 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]]
471 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
472 // FIXME: When the iteration count of some nested loop is not a known constant,
473 // we should pre-calculate it, like we do for the total number of iterations!
474 for (char i = static_cast<char>(y); i <= '9'; ++i)
475 for (x = 11; x > 0; --x) {
476 // CHECK: [[LOOP1_BODY]]
477 // Start of body: indices are calculated from IV:
478 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}}
479 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}}
480 // ... loop body ...
481 // End of body: store into a[i]:
482 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
483 // CHECK-NOT: !llvm.access.group
484 a[i] = b[i] * c[i] * d[i];
485 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
486 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1
487 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
488 // CHECK-NEXT: br label %{{.+}}
489 }
490 // CHECK: [[LOOP1_END]]
491 // CHECK: [[O_LOOP1_END]]
492 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
493 // CHECK: ret void
494 }
495
496 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
runtime(float * a,float * b,float * c,float * d)497 void runtime(float *a, float *b, float *c, float *d) {
498 int x = 0;
499 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
500 #pragma omp for collapse(2) schedule(runtime)
501 // OMP45: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 37, i32 0, i32 199, i32 1, i32 1)
502 // OMP5: call void @__kmpc_dispatch_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741861, i32 0, i32 199, i32 1, i32 1)
503 //
504 // 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:%[^,]+]])
505 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
506 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
507
508 // Loop header
509 // CHECK: [[O_LOOP1_BODY]]
510 // CHECK: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
511 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]]
512 // CHECK: [[IV:%.+]] = load i32, i32* [[OMP_IV]]
513
514 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
515 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
516 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
517 for (unsigned char i = '0' ; i <= '9'; ++i)
518 for (x = -10; x < 10; ++x) {
519 // CHECK: [[LOOP1_BODY]]
520 // Start of body: indices are calculated from IV:
521 // CHECK: store i8 {{%[^,]+}}, i8* {{%[^,]+}}
522 // CHECK: store i32 {{%[^,]+}}, i32* {{%[^,]+}}
523 // ... loop body ...
524 // End of body: store into a[i]:
525 // CHECK: store float [[RESULT:%.+]], float* {{%.+}}
526 // CHECK-NOT: !llvm.access.group
527 a[i] = b[i] * c[i] * d[i];
528 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
529 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
530 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
531 // CHECK-NEXT: br label %{{.+}}
532 }
533 // CHECK: [[LOOP1_END]]
534 // CHECK: [[O_LOOP1_END]]
535 // CHECK: call {{.+}} @__kmpc_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
536 // CHECK: ret void
537 }
538
539 // CHECK-LABEL: test_precond
test_precond()540 void test_precond() {
541 // CHECK: [[A_ADDR:%.+]] = alloca i8,
542 // CHECK: [[I_ADDR:%.+]] = alloca i8,
543 // CHECK: [[CAP:%.+]] = alloca i8,
544 char a = 0;
545 // CHECK: store i8 0,
546 // CHECK: store i32
547 // CHECK: store i8
548 // CHECK: [[A:%.+]] = load i8, i8* [[CAP]],
549 // CHECK: [[CONV:%.+]] = sext i8 [[A]] to i32
550 // CHECK: [[CMP:%.+]] = icmp slt i32 [[CONV]], 10
551 // CHECK: br i1 [[CMP]], label %[[PRECOND_THEN:[^,]+]], label %[[PRECOND_END:[^,]+]]
552 // CHECK: [[PRECOND_THEN]]
553 // CHECK: call void @__kmpc_for_static_init_4
554 #pragma omp for
555 for(char i = a; i < 10; ++i);
556 // CHECK: call void @__kmpc_for_static_fini
557 // CHECK: [[PRECOND_END]]
558 }
559
560 // TERM_DEBUG-LABEL: foo
foo()561 int foo() { extern void mayThrow(); mayThrow(); return 0;};
562
563 // TERM_DEBUG-LABEL: parallel_for
parallel_for(float * a)564 void parallel_for(float *a) {
565 #pragma omp parallel
566 #pragma omp for schedule(static, 5)
567 // TERM_DEBUG-NOT: __kmpc_global_thread_num
568 // TERM_DEBUG: call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC:![0-9]+]]
569 // TERM_DEBUG: invoke i32 {{.*}}foo{{.*}}()
570 // TERM_DEBUG: unwind label %[[TERM_LPAD:.+]],
571 // TERM_DEBUG-NOT: __kmpc_global_thread_num
572 // TERM_DEBUG: call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC]]
573 // TERM_DEBUG: call {{.+}} @__kmpc_barrier({{.+}}), !dbg [[DBG_LOC]]
574 // TERM_DEBUG: [[TERM_LPAD]]
575 // TERM_DEBUG: call void @__clang_call_terminate
576 // TERM_DEBUG: unreachable
577 for (unsigned i = 131071; i <= 2147483647; i += 127)
578 a[i] += foo();
579 }
580 // Check source line corresponds to "#pragma omp for schedule(static, 5)" above:
581 // TERM_DEBUG: [[DBG_LOC]] = !DILocation(line: [[@LINE-15]],
582
583 char i = 1, j = 2, k = 3;
584 // CHECK-LABEL: for_with_global_lcv
for_with_global_lcv()585 void for_with_global_lcv() {
586 // CHECK: alloca i8,
587 // CHECK: [[I_ADDR:%.+]] = alloca i8,
588 // CHECK: alloca i8,
589 // CHECK: [[J_ADDR:%.+]] = alloca i8,
590
591 // CHECK: call void @__kmpc_for_static_init_4(
592 // CHECK-NOT: [[I]]
593 // CHECK: store i8 %{{.+}}, i8* [[I_ADDR]]
594 // CHECK-NOT: [[I]]
595 // CHECK: [[I_VAL:%.+]] = load i8, i8* [[I_ADDR]],
596 // CHECK-NOT: [[I]]
597 // CHECK: store i8 [[I_VAL]], i8* [[K]]
598 // CHECK-NOT: [[I]]
599 // CHECK: call void @__kmpc_for_static_fini(
600 // CHECK: call void @__kmpc_barrier(
601 #pragma omp for
602 for (i = 0; i < 2; ++i) {
603 k = i;
604 }
605 // CHECK: call void @__kmpc_for_static_init_4(
606 // CHECK-NOT: [[J]]
607 // CHECK: store i8 %{{.+}}, i8* [[J_ADDR]]
608 // CHECK-NOT: [[J]]
609 // CHECK: [[J_VAL:%.+]] = load i8, i8* [[J_ADDR]],
610 // CHECK-NOT: [[J]]
611 // CHECK: store i8 [[J_VAL]], i8* [[K]]
612 // CHECK-NOT: [[J]]
613 // CHECK: call void @__kmpc_for_static_fini(
614 #pragma omp for collapse(2)
615 for (int i = 0; i < 2; ++i)
616 for (j = 0; j < 2; ++j) {
617 k = i;
618 k = j;
619 }
620 char &cnt = i;
621 #pragma omp for
622 for (cnt = 0; cnt < 2; ++cnt)
623 k = cnt;
624 }
625
626 // CHECK-LABEL: for_with_references
for_with_references()627 void for_with_references() {
628 // CHECK: [[I:%.+]] = alloca i8,
629 // CHECK: [[CNT:%.+]] = alloca i8*,
630 // CHECK: [[CNT_PRIV:%.+]] = alloca i8,
631 // CHECK: call void @__kmpc_for_static_init_8(
632 // CHECK-NOT: load i8, i8* [[CNT]],
633 // CHECK: call void @__kmpc_for_static_fini(
634 char i = 0;
635 char &cnt = i;
636 #pragma omp for collapse(2)
637 for (cnt = 0; cnt < 2; ++cnt)
638 for (int j = cnt; j < 4 + cnt; j++)
639 k = cnt;
640 }
641
642 // CHECK-LABEL: for_with_references_dep_cond
for_with_references_dep_cond()643 void for_with_references_dep_cond() {
644 // CHECK: [[I:%.+]] = alloca i8,
645 // CHECK: [[CNT:%.+]] = alloca i8*,
646 // CHECK: [[CNT_PRIV:%.+]] = alloca i8,
647 // CHECK: call void @__kmpc_for_static_init_8(
648 // CHECK-NOT: load i8, i8* [[CNT]],
649 // CHECK: call void @__kmpc_for_static_fini(
650 char i = 0;
651 char &cnt = i;
652 #pragma omp for collapse(2)
653 for (cnt = 0; cnt < 2; ++cnt)
654 for (int j = 0; j < 4 + cnt; j++)
655 k = cnt;
656 }
657
658 struct Bool {
BoolBool659 Bool(bool b) : b(b) {}
operator boolBool660 operator bool() const { return b; }
661 const bool b;
662 };
663
664 template <typename T>
665 struct It {
ItIt666 It() : p(0) {}
667 It(const It &, int = 0) ;
668 template <typename U>
669 It(U &, int = 0) ;
670 It &operator=(const It &);
671 It &operator=(It &);
~ItIt672 ~It() {}
673
ItIt674 It(T *p) : p(p) {}
675
operator T*&It676 operator T *&() { return p; }
operator T*It677 operator T *() const { return p; }
operator ->It678 T *operator->() const { return p; }
679
operator ++It680 It &operator++() { ++p; return *this; }
operator --It681 It &operator--() { --p; return *this; }
operator +=It682 It &operator+=(unsigned n) { p += n; return *this; }
operator -=It683 It &operator-=(unsigned n) { p -= n; return *this; }
684
685 T *p;
686 };
687
688 template <typename T>
operator +(It<T> a,typename It<T>::difference_type n)689 It<T> operator+(It<T> a, typename It<T>::difference_type n) { return a.p + n; }
690
691 template <typename T>
operator +(typename It<T>::difference_type n,It<T> a)692 It<T> operator+(typename It<T>::difference_type n, It<T> a) { return a.p + n; }
693
694 template <typename T>
operator -(It<T> a,typename It<T>::difference_type n)695 It<T> operator-(It<T> a, typename It<T>::difference_type n) { return a.p - n; }
696
697 typedef Bool BoolType;
698
699 template <typename T>
operator <(It<T> a,It<T> b)700 BoolType operator<(It<T> a, It<T> b) { return a.p < b.p; }
701
loop_with_It(It<char> begin,It<char> end)702 void loop_with_It(It<char> begin, It<char> end) {
703 #pragma omp for
704 for (It<char> it = begin; it < end; ++it) {
705 *it = 0;
706 }
707 }
708
709 // CHECK-LABEL: loop_with_It
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
loop_with_It_plus(It<char> begin,It<char> end)714 void loop_with_It_plus(It<char> begin, It<char> end) {
715 #pragma omp for
716 for (It<char> it = begin; it < end; it+=1u) {
717 *it = 0;
718 }
719 }
720
721 // CHECK-LABEL: loop_with_It_plus
722 // CHECK: call i32 @__kmpc_global_thread_num(
723 // CHECK: call void @__kmpc_for_static_init_8(
724 // CHECK: call void @__kmpc_for_static_fini(
725
loop_with_stmt_expr()726 void loop_with_stmt_expr() {
727 #pragma omp for collapse(2)
728 for (int i = __extension__({float b = 0;b; }); i < __extension__({double c = 1;c; }); i += __extension__({char d = 1; d; }))
729 for (int j = i; j < 4 + i; j++)
730 ;
731 }
732 // CHECK-LABEL: loop_with_stmt_expr
733 // CHECK: call i32 @__kmpc_global_thread_num(
734 // CHECK: call void @__kmpc_for_static_init_8(
735 // CHECK: call void @__kmpc_for_static_fini(
736
737
738 // CHECK-LABEL: fint
739 // CHECK: call {{.*}}i32 {{.*}}ftemplate
740 // CHECK: ret i32
741
742 // CHECK: load i16, i16*
743 // CHECK: store i16 %
744 // CHECK: call void {{.+}}@__kmpc_fork_call(
745 // CHECK: call void @__kmpc_for_static_init_4(
746 template <typename T>
ftemplate()747 T ftemplate() {
748 short aa = 0;
749
750 #pragma omp parallel for schedule(static, aa)
751 for (int i = 0; i < 100; i++) {
752 }
753 return T();
754 }
755
fint(void)756 int fint(void) { return ftemplate<int>(); }
757
758 // Check for imperfectly loop nests codegen.
759 #if _OPENMP == 201811
760 void first();
761 void last();
762 void inner_f();
763 void inner_l();
764 void body_f();
765
766 // OMP5-LABEL: imperfectly_nested_loop
imperfectly_nested_loop()767 void imperfectly_nested_loop() {
768 // OMP5: call void @__kmpc_for_static_init_4(
769 #pragma omp for collapse(3) order(concurrent)
770 for (int i = 0; i < 10; ++i) {
771 {
772 int a, d;
773 // OMP5: invoke void @{{.+}}first{{.+}}()
774 first();
775 // OMP5: load i32{{.*}}!llvm.access.group ![[AG:[0-9]+]]
776 // OMP5: store i32{{.*}}!llvm.access.group ![[AG]]
777 a = d;
778 for (int j = 0; j < 10; ++j) {
779 int a, d;
780 // OMP5: invoke void @{{.+}}inner_f{{.+}}()
781 inner_f();
782 // OMP5: load i32{{.*}}!llvm.access.group ![[AG]]
783 // OMP5: store i32{{.*}}!llvm.access.group ![[AG]]
784 a = d;
785 for (int k = 0; k < 10; ++k) {
786 int a, d;
787 // OMP5: invoke void @{{.+}}body_f{{.+}}()
788 body_f();
789 // OMP5: load i32{{.*}}!llvm.access.group ![[AG]]
790 // OMP5: store i32{{.*}}!llvm.access.group ![[AG]]
791 a = d;
792 }
793 // OMP5: invoke void @{{.+}}inner_l{{.+}}()
794 inner_l();
795 }
796 // OMP5: invoke void @{{.+}}last{{.+}}()
797 last();
798 }
799 }
800 // OMP5: call void @__kmpc_for_static_fini(
801 }
802
803 // OMP5: ![[AG]] = distinct !{}
804 // OMP5: !{!"llvm.loop.parallel_accesses", ![[AG]]}
805
806 #endif
807
808 #endif // HEADER
809