xref: /llvm-project/clang/test/OpenMP/for_codegen.cpp (revision 0c6f2f629cc0017361310fa4c132090413a874db) 
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 -fopenmp-version=50 -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 -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=OMP51
4 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
5 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK --check-prefix=OMP51
6 // RUN: %clang_cc1 -fopenmp -fopenmp-version=50 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
7 // RUN: %clang_cc1 -fopenmp -fopenmp-version=50 -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
8 // 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
9 // 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
10 // 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
11 // 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
12 
13 // 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
14 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
15 // 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
16 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
17 // 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
18 // 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
19 // 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
20 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
21 //
22 // expected-no-diagnostics
23 #ifndef HEADER
24 #define HEADER
25 // PROF-INSTR-PATH: constant [25 x i8] c"for_codegen-test.profraw\00"
26 
27 // CHECK: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, ptr }
28 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 {{[0-9]+}}, ptr
29 // CHECK-DAG: [[LOOP_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 514, i32 0, i32 {{[0-9]+}}, ptr
30 // CHECK-DAG: [[I:@.+]] ={{.*}} global i8 1,
31 // CHECK-DAG: [[J:@.+]] ={{.*}} global i8 2,
32 // CHECK-DAG: [[K:@.+]] ={{.*}} global i8 3,
33 
34 // CHECK-LABEL: loop_with_counter_collapse
loop_with_counter_collapse()35 void loop_with_counter_collapse() {
36   // Captured initializations.
37   // CHECK: store i32 0, ptr [[I_TMP:%.+]],
38   // CHECK: [[VAL:%.+]] = load i32, ptr [[I_TMP]],
39   // CHECK: store i32 [[VAL]], ptr [[J_LB_MIN:%.+]],
40   // CHECK: store i32 3, ptr [[I_TMP]],
41   // CHECK: [[VAL:%.+]] = load i32, ptr [[I_TMP]],
42   // CHECK: store i32 [[VAL]], ptr [[J_LB_MAX:%.+]],
43   // CHECK: [[J_LB_MIN_VAL:%.+]] = load i32, ptr [[J_LB_MIN]],
44   // CHECK: [[J_LB_MAX_VAL:%.+]] = load i32, ptr [[J_LB_MAX]],
45   // CHECK: [[CMP:%.+]] = icmp slt i32 [[J_LB_MIN_VAL]], [[J_LB_MAX_VAL]]
46   // CHECK: [[BOOL:%.+]] = zext i1 [[CMP]] to i8
47   // CHECK: store i8 [[BOOL]], ptr [[J_LB_CMP:%.+]],
48   // CHECK: store i32 0, ptr [[I_TMP]],
49   // CHECK: [[VAL:%.+]] = load i32, ptr [[I_TMP]],
50   // CHECK: [[J_UB_MIN_VAL:%.+]] = add nsw i32 4, [[VAL]]
51   // CHECK: store i32 [[J_UB_MIN_VAL]], ptr [[J_UB_MIN:%.+]],
52   // CHECK: store i32 3, ptr [[I_TMP]],
53   // CHECK: [[VAL:%.+]] = load i32, ptr [[I_TMP]],
54   // CHECK: [[J_UB_MAX_VAL:%.+]] = add nsw i32 4, [[VAL]]
55   // CHECK: store i32 [[J_UB_MAX_VAL]], ptr [[J_UB_MAX:%.+]],
56   // CHECK: [[J_UB_MIN_VAL:%.+]] = load i32, ptr [[J_UB_MIN]],
57   // CHECK: [[J_UB_MAX_VAL:%.+]] = load i32, ptr [[J_UB_MAX]],
58   // CHECK: [[CMP:%.+]] = icmp sgt i32 [[J_UB_MIN_VAL]], [[J_UB_MAX_VAL]]
59   // CHECK: [[BOOL:%.+]] = zext i1 [[CMP]] to i8
60   // CHECK: store i8 [[BOOL]], ptr [[J_UB_CMP:%.+]],
61   // CHECK: [[J_UB_CMP_VAL:%.+]] = load i8, ptr [[J_UB_CMP]],
62   // CHECK: [[BOOL:%.+]] = trunc i8 [[J_UB_CMP_VAL]] to i1
63   // CHECK: br i1 [[BOOL]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]]
64   // CHECK: [[TRUE]]:
65   // CHECK: [[J_UB_MIN_VAL:%.+]] = load i32, ptr [[J_UB_MIN]],
66   // CHECK: br label %[[EXIT:[^,]+]]
67   // CHECK: [[FALSE]]:
68   // CHECK: [[J_UB_MAX_VAL:%.+]] = load i32, ptr [[J_UB_MAX]],
69   // CHECK: br label %[[EXIT]]
70   // CHECK: [[EXIT]]:
71   // CHECK: [[J_UB_VAL:%.+]] = phi i32 [ [[J_UB_MIN_VAL]], %[[TRUE]] ], [ [[J_UB_MAX_VAL]], %[[FALSE]] ]
72   // CHECK: store i32 [[J_UB_VAL]], ptr [[J_UB:%.+]],
73   // CHECK: [[J_LB_CMP_VAL:%.+]] = load i8, ptr [[J_LB_CMP]],
74   // CHECK: [[BOOL:%.+]] = trunc i8 [[J_LB_CMP_VAL]] to i1
75   // CHECK: br i1 [[BOOL]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]]
76   // CHECK: [[TRUE]]:
77   // CHECK: [[J_LB_MIN_VAL:%.+]] = load i32, ptr [[J_LB_MIN]],
78   // CHECK: br label %[[EXIT:[^,]+]]
79   // CHECK: [[FALSE]]:
80   // CHECK: [[J_LB_MAX_VAL:%.+]] = load i32, ptr [[J_LB_MAX]],
81   // CHECK: br label %[[EXIT]]
82   // CHECK: [[EXIT]]:
83   // CHECK: [[J_LB_VAL:%.+]] = phi i32 [ [[J_LB_MIN_VAL]], %[[TRUE]] ], [ [[J_LB_MAX_VAL]], %[[FALSE]] ]
84   // CHECK: store i32 [[J_LB_VAL]], ptr [[I_TMP]],
85   // CHECK: store i32 [[J_LB_VAL]], ptr [[J_LB:%.+]],
86   // CHECK: [[J_UB_VAL:%.+]] = load i32, ptr [[J_UB]],
87   // CHECK: [[J_LB_VAL:%.+]] = load i32, ptr [[J_LB]],
88   // CHECK: [[SUB:%.+]] = sub i32 [[J_UB_VAL]], [[J_LB_VAL]]
89   // CHECK: [[SUB_ST:%.+]] = sub i32 [[SUB]], 1
90   // CHECK: [[ADD_ST:%.+]] = add i32 [[SUB_ST]], 1
91   // CHECK: [[DIV_ST:%.+]] = udiv i32 [[ADD_ST]], 1
92   // CHECK: [[CAST:%.+]] = zext i32 [[DIV_ST]] to i64
93   // CHECK: [[MUL:%.+]] = mul nsw i64 4, [[CAST]]
94   // CHECK: [[NUM_ITERS_VAL:%.+]] = sub nsw i64 [[MUL]], 1
95   // CHECK: store i64 [[NUM_ITERS_VAL]], ptr [[NUM_ITERS:%.+]],
96 
97   // CHECK: store i64 0, ptr [[LB:%.+]],
98   // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, ptr [[NUM_ITERS]],
99   // CHECK: store i64 [[NUM_ITERS_VAL]], ptr [[UB:%.+]],
100   // CHECK: store i64 1, ptr [[STRIDE:%.+]],
101   // CHECK: store i32 0, ptr [[IS_LAST:%.+]],
102   // CHECK: call void @__kmpc_for_static_init_8(ptr @{{.+}}, i32 %{{.+}}, i32 34, ptr [[IS_LAST]], ptr [[LB]], ptr [[UB]], ptr [[STRIDE]], i64 1, i64 1)
103   // CHECK: [[UB_VAL:%.+]] = load i64, ptr [[UB]],
104   // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, ptr [[NUM_ITERS]],
105   // CHECK: [[CMP:%.+]] = icmp sgt i64 [[UB_VAL]], [[NUM_ITERS_VAL]]
106   // CHECK: br i1 [[CMP]], label %[[TRUE:[^,]+]], label %[[FALSE:[^,]+]]
107   // CHECK: [[TRUE]]:
108   // CHECK: [[NUM_ITERS_VAL:%.+]] = load i64, ptr [[NUM_ITERS]],
109   // CHECK: br label %[[DONE:[^,]+]]
110   // CHECK: [[FALSE]]:
111   // CHECK: [[UB_VAL:%.+]] = load i64, ptr [[UB]],
112   // CHECK: br label %[[DONE]]
113   // CHECK: [[DONE]]:
114   // CHECK: [[TOP:%.+]] = phi i64 [ [[NUM_ITERS_VAL]], %[[TRUE]] ], [ [[UB_VAL]], %[[FALSE]] ]
115   // CHECK: store i64 [[TOP]], ptr [[UB]],
116   // CHECK: [[LB_VAL:%.+]] = load i64, ptr [[LB]],
117   // CHECK: store i64 [[LB_VAL]], ptr [[IV:%.+]],
118   // CHECK: br label %[[COND:[^,]+]]
119   // CHECK: [[COND]]:
120   // CHECK: [[IV_VAL:%.+]] = load i64, ptr [[IV]],
121   // CHECK: [[UB_VAL:%.+]] = load i64, ptr [[UB]],
122   // CHECK: [[CMP:%.+]] = icmp sle i64 [[IV_VAL]], [[UB_VAL]]
123   // CHECK: br i1 [[CMP]], label %[[BODY:[^,]+]], label %[[CLEANUP:[^,]+]]
124   // LIFETIME: [[CLEANUP]]:
125   // LIFETIME: br label %[[CLEANUP:[^,]+]]
126   // CHECK: [[BODY]]:
127   // CHECK: [[IV_VAL:%.+]] = load i64, ptr [[IV]],
128   // CHECK: [[J_UB_VAL:%.+]] = load i32, ptr [[J_UB]],
129   // CHECK: [[J_LB_VAL:%.+]] = load i32, ptr [[J_LB]],
130   // CHECK: [[SUB:%.+]] = sub i32 [[J_UB_VAL]], [[J_LB_VAL]]
131   // CHECK: [[SUB_ST:%.+]] = sub i32 [[SUB]], 1
132   // CHECK: [[ADD_ST:%.+]] = add i32 [[SUB_ST]], 1
133   // CHECK: [[DIV_ST:%.+]] = udiv i32 [[ADD_ST]], 1
134   // CHECK: [[MUL:%.+]] = mul i32 1, [[DIV_ST]]
135   // CHECK: [[CAST:%.+]] = zext i32 [[MUL]] to i64
136   // CHECK: [[DIV:%.+]] = sdiv i64 [[IV_VAL]], [[CAST]]
137   // CHECK: [[MUL:%.+]] = mul nsw i64 [[DIV]], 1
138   // CHECK: [[ADD:%.+]] = add nsw i64 0, [[MUL]]
139   // CHECK: [[CAST:%.+]] = trunc i64 [[ADD]] to i32
140   // CHECK: store i32 [[CAST]], ptr [[I_PRIV:%.+]],
141   // CHECK: [[I_VAL:%.+]] = load i32, ptr [[I_PRIV]],
142   // CHECK: [[CONV:%.+]] = sext i32 [[I_VAL]] to i64
143   // CHECK: [[IV_VAL:%.+]] = load i64, ptr [[IV]],
144   // CHECK: [[IV_VAL1:%.+]] = load i64, ptr [[IV]],
145   // CHECK: [[J_UB_VAL:%.+]] = load i32, ptr [[J_UB]],
146   // CHECK: [[J_LB_VAL:%.+]] = load i32, ptr [[J_LB]],
147   // CHECK: [[SUB:%.+]] = sub i32 [[J_UB_VAL]], [[J_LB_VAL]]
148   // CHECK: [[SUB_ST:%.+]] = sub i32 [[SUB]], 1
149   // CHECK: [[ADD_ST:%.+]] = add i32 [[SUB_ST]], 1
150   // CHECK: [[DIV_ST:%.+]] = udiv i32 [[ADD_ST]], 1
151   // CHECK: [[MUL:%.+]] = mul i32 1, [[DIV_ST]]
152   // CHECK: [[CAST:%.+]] = zext i32 [[MUL]] to i64
153   // CHECK: [[DIV:%.+]] = sdiv i64 [[IV_VAL1]], [[CAST]]
154   // CHECK: [[J_UB_VAL:%.+]] = load i32, ptr [[J_UB]],
155   // CHECK: [[J_LB_VAL:%.+]] = load i32, ptr [[J_LB]],
156   // CHECK: [[SUB:%.+]] = sub i32 [[J_UB_VAL]], [[J_LB_VAL]]
157   // CHECK: [[SUB_ST:%.+]] = sub i32 [[SUB]], 1
158   // CHECK: [[ADD_ST:%.+]] = add i32 [[SUB_ST]], 1
159   // CHECK: [[DIV_ST:%.+]] = udiv i32 [[ADD_ST]], 1
160   // CHECK: [[MUL:%.+]] = mul i32 1, [[DIV_ST]]
161   // CHECK: [[CAST:%.+]] = zext i32 [[MUL]] to i64
162   // CHECK: [[MUL:%.+]] = mul nsw i64 [[DIV]], [[CAST]]
163   // CHECK: [[SUB:%.+]] = sub nsw i64 [[IV_VAL]], [[MUL]]
164   // CHECK: [[MUL:%.+]] = mul nsw i64 [[SUB:%.+]], 1
165   // CHECK: [[ADD:%.+]] = add nsw i64 [[CONV]], [[MUL]]
166   // CHECK: [[CAST:%.+]] = trunc i64 [[ADD]] to i32
167   // CHECK: store i32 [[CAST]], ptr [[J_PRIV:%.+]],
168 
169   // Check that the loop variable is not out of its boundaries.
170   // CHECK: [[J_VAL:%.+]] = load i32, ptr [[J_PRIV]],
171   // CHECK: [[I_VAL:%.+]] = load i32, ptr [[I_PRIV]],
172   // CHECK: [[J_COND:%.+]] = add nsw i32 4, [[I_VAL]]
173   // CHECK: [[CMP:%.+]] = icmp slt i32 [[J_VAL]], [[J_COND]]
174   // CHECK: br i1 [[CMP]], label %[[NEXT:[^,]+]], label %[[BODY_CONT:[^,]+]]
175   // CHECK: [[NEXT]]:
176 
177   // Main body is empty.
178   // CHECK: br label %[[BODY_CONT]]
179   // CHECK: [[BODY_CONT]]:
180   // CHECK: br label %[[INC:[^,]+]]
181   // CHECK: [[INC]]:
182   // CHECK: [[IV_VAL:%.+]] = load i64, ptr [[IV]],
183   // CHECK: [[ADD:%.+]] = add nsw i64 [[IV_VAL]], 1
184   // CHECK: store i64 [[ADD]], ptr [[IV]],
185   // CHECK: br label %[[COND]]
186   // CHECK: [[CLEANUP]]:
187   // CHECK: br label %[[EXIT:[^,]+]]
188   // CHECK: [[EXIT]]:
189   // CHECK: call void @__kmpc_for_static_fini(ptr @{{.+}}, i32 %{{.+}})
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   // LIFETIME: call void @llvm.lifetime.end
196   // LIFETIME: call void @llvm.lifetime.end
197   // LIFETIME: call void @llvm.lifetime.end
198   // LIFETIME: call void @llvm.lifetime.end
199   #pragma omp for collapse(2)
200   for (int i = 0; i < 4; i++) {
201     for (int j = i; j < 4 + i; j++) {
202     }
203   }
204 }
205 
206 // CHECK-LABEL: loop_with_counter_collapse4
loop_with_counter_collapse4()207 void loop_with_counter_collapse4() {
208 
209   // Check bounds calculation when collapse > 2
210   // CHECK: store i32 0, ptr [[I_TMP:%.+]],
211   // CHECK: [[VAL:%.+]] = load i32, ptr [[I_TMP]],
212   // CHECK: store i32 [[VAL]], ptr [[K_LB_MIN:%.+]],
213   // CHECK: store i32 6, ptr [[I_TMP]],
214   // CHECK: [[VAL:%.+]] = load i32, ptr [[I_TMP]],
215   // CHECK: store i32 [[VAL]], ptr [[K_LB_MAX:%.+]],
216   #pragma omp for collapse(4)
217   for (int i = 0; i < 7; i++) {
218     for (int j = 0; j < 11; j++) {
219       for (int k = i; k < 7; k++) {
220         for (int l = 0; l < 11; l++) {
221         }
222       }
223     }
224   }
225 }
226 
227 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}})
without_schedule_clause(float * a,float * b,float * c,float * d)228 void without_schedule_clause(float *a, float *b, float *c, float *d) {
229 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(ptr [[DEFAULT_LOC:[@%].+]])
230   #pragma omp for nowait
231 // CHECK: call void @__kmpc_for_static_init_4(ptr [[LOOP_LOC]], i32 [[GTID]], i32 34, ptr [[IS_LAST:%[^,]+]], ptr [[OMP_LB:%[^,]+]], ptr [[OMP_UB:%[^,]+]], ptr [[OMP_ST:%[^,]+]], i32 1, i32 1)
232 // UB = min(UB, GlobalUB)
233 // CHECK-NEXT: [[UB:%.+]] = load i32, ptr [[OMP_UB]]
234 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
235 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
236 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
237 // CHECK-NEXT: store i32 [[UBRESULT]], ptr [[OMP_UB]]
238 // CHECK-NEXT: [[LB:%.+]] = load i32, ptr [[OMP_LB]]
239 // CHECK-NEXT: store i32 [[LB]], ptr [[OMP_IV:[^,]+]]
240 // Loop header
241 // CHECK: [[IV:%.+]] = load i32, ptr [[OMP_IV]]
242 // CHECK-NEXT: [[UB:%.+]] = load i32, ptr [[OMP_UB]]
243 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
244 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
245   for (int i = 33; i < 32000000; i += 7) {
246 // CHECK: [[LOOP1_BODY]]
247 // Start of body: calculate i from IV:
248 // CHECK: [[IV1_1:%.+]] = load i32, ptr [[OMP_IV]]
249 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7
250 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]]
251 // CHECK-NEXT: store i32 [[CALC_I_2]], ptr [[LC_I:.+]]
252 // ... loop body ...
253 // End of body: store into a[i]:
254 // CHECK: store float [[RESULT:%.+]], ptr {{%.+}}
255 // CHECK-NOT: !llvm.access.group
256     a[i] = b[i] * c[i] * d[i];
257 // CHECK: [[IV1_2:%.+]] = load i32, ptr [[OMP_IV]]{{.*}}
258 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
259 // CHECK-NEXT: store i32 [[ADD1_2]], ptr [[OMP_IV]]
260 // CHECK-NEXT: br label %{{.+}}
261   }
262 // CHECK: [[LOOP1_END]]
263 // CHECK: call void @__kmpc_for_static_fini(ptr [[LOOP_LOC]], i32 [[GTID]])
264 // CHECK-NOT: __kmpc_barrier
265 // CHECK: ret void
266 }
267 
268 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}})
static_not_chunked(float * a,float * b,float * c,float * d)269 void static_not_chunked(float *a, float *b, float *c, float *d) {
270 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(ptr [[DEFAULT_LOC:[@%].+]])
271   #pragma omp for schedule(static)
272 // CHECK: call void @__kmpc_for_static_init_4(ptr [[LOOP_LOC]], i32 [[GTID]], i32 34, ptr [[IS_LAST:%[^,]+]], ptr [[OMP_LB:%[^,]+]], ptr [[OMP_UB:%[^,]+]], ptr [[OMP_ST:%[^,]+]], i32 1, i32 1)
273 // UB = min(UB, GlobalUB)
274 // CHECK-NEXT: [[UB:%.+]] = load i32, ptr [[OMP_UB]]
275 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
276 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
277 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
278 // CHECK-NEXT: store i32 [[UBRESULT]], ptr [[OMP_UB]]
279 // CHECK-NEXT: [[LB:%.+]] = load i32, ptr [[OMP_LB]]
280 // CHECK-NEXT: store i32 [[LB]], ptr [[OMP_IV:[^,]+]]
281 // Loop header
282 // CHECK: [[IV:%.+]] = load i32, ptr [[OMP_IV]]
283 // CHECK-NEXT: [[UB:%.+]] = load i32, ptr [[OMP_UB]]
284 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
285 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
286   for (int i = 32000000; i > 33; i += -7) {
287 // CHECK: [[LOOP1_BODY]]
288 // Start of body: calculate i from IV:
289 // CHECK: [[IV1_1:%.+]] = load i32, ptr [[OMP_IV]]
290 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7
291 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]]
292 // CHECK-NEXT: store i32 [[CALC_I_2]], ptr [[LC_I:.+]]
293 // ... loop body ...
294 // End of body: store into a[i]:
295 // CHECK: store float [[RESULT:%.+]], ptr {{%.+}}
296 // CHECK-NOT: !llvm.access.group
297     a[i] = b[i] * c[i] * d[i];
298 // CHECK: [[IV1_2:%.+]] = load i32, ptr [[OMP_IV]]{{.*}}
299 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
300 // CHECK-NEXT: store i32 [[ADD1_2]], ptr [[OMP_IV]]
301 // CHECK-NEXT: br label %{{.+}}
302   }
303 // CHECK: [[LOOP1_END]]
304 // CHECK: call void @__kmpc_for_static_fini(ptr [[LOOP_LOC]], i32 [[GTID]])
305 // CHECK: call {{.+}} @__kmpc_barrier(ptr [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
306 // CHECK: ret void
307 }
308 
309 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}})
static_chunked(float * a,float * b,float * c,float * d)310 void static_chunked(float *a, float *b, float *c, float *d) {
311 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(ptr [[DEFAULT_LOC:[@%].+]])
312   #pragma omp for schedule(monotonic: static, 5)
313 // CHECK: call void @__kmpc_for_static_init_4u(ptr [[LOOP_LOC]], i32 [[GTID]], i32 536870945, ptr [[IS_LAST:%[^,]+]], ptr [[OMP_LB:%[^,]+]], ptr [[OMP_UB:%[^,]+]], ptr [[OMP_ST:%[^,]+]], i32 1, i32 5)
314 // UB = min(UB, GlobalUB)
315 // CHECK: [[UB:%.+]] = load i32, ptr [[OMP_UB]]
316 // CHECK-NEXT: [[UBCMP:%.+]] = icmp ugt i32 [[UB]], 16908288
317 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
318 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 16908288, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
319 // CHECK-NEXT: store i32 [[UBRESULT]], ptr [[OMP_UB]]
320 // CHECK-NEXT: [[LB:%.+]] = load i32, ptr [[OMP_LB]]
321 // CHECK-NEXT: store i32 [[LB]], ptr [[OMP_IV:[^,]+]]
322 
323 // Outer loop header
324 // CHECK: [[O_IV:%.+]] = load i32, ptr [[OMP_IV]]
325 // CHECK-NEXT: [[O_UB:%.+]] = load i32, ptr [[OMP_UB]]
326 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ule i32 [[O_IV]], [[O_UB]]
327 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
328 
329 // Loop header
330 // CHECK: [[O_LOOP1_BODY]]
331 // CHECK: [[IV:%.+]] = load i32, ptr [[OMP_IV]]
332 // CHECK-NEXT: [[UB:%.+]] = load i32, ptr [[OMP_UB]]
333 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i32 [[IV]], [[UB]]
334 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
335   for (unsigned i = 131071; i <= 2147483647; i += 127) {
336 // CHECK: [[LOOP1_BODY]]
337 // Start of body: calculate i from IV:
338 // CHECK: [[IV1_1:%.+]] = load i32, ptr [[OMP_IV]]
339 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i32 [[IV1_1]], 127
340 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 131071, [[CALC_I_1]]
341 // CHECK-NEXT: store i32 [[CALC_I_2]], ptr [[LC_I:.+]]
342 // ... loop body ...
343 // End of body: store into a[i]:
344 // CHECK: store float [[RESULT:%.+]], ptr {{%.+}}
345 // CHECK-NOT: !llvm.access.group
346     a[i] = b[i] * c[i] * d[i];
347 // CHECK: [[IV1_2:%.+]] = load i32, ptr [[OMP_IV]]{{.*}}
348 // CHECK-NEXT: [[ADD1_2:%.+]] = add i32 [[IV1_2]], 1
349 // CHECK-NEXT: store i32 [[ADD1_2]], ptr [[OMP_IV]]
350 // CHECK-NEXT: br label %{{.+}}
351   }
352 // CHECK: [[LOOP1_END]]
353 // Update the counters, adding stride
354 // CHECK:  [[LB:%.+]] = load i32, ptr [[OMP_LB]]
355 // CHECK-NEXT: [[ST:%.+]] = load i32, ptr [[OMP_ST]]
356 // CHECK-NEXT: [[ADD_LB:%.+]] = add i32 [[LB]], [[ST]]
357 // CHECK-NEXT: store i32 [[ADD_LB]], ptr [[OMP_LB]]
358 // CHECK-NEXT: [[UB:%.+]] = load i32, ptr [[OMP_UB]]
359 // CHECK-NEXT: [[ST:%.+]] = load i32, ptr [[OMP_ST]]
360 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]]
361 // CHECK-NEXT: store i32 [[ADD_UB]], ptr [[OMP_UB]]
362 
363 // CHECK: [[O_LOOP1_END]]
364 // CHECK: call void @__kmpc_for_static_fini(ptr [[LOOP_LOC]], i32 [[GTID]])
365 // CHECK: call {{.+}} @__kmpc_barrier(ptr [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
366 // CHECK: ret void
367 }
368 
369 // CHECK-LABEL: define {{.*void}} @{{.*}}dynamic1{{.*}}(ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}})
dynamic1(float * a,float * b,float * c,float * d)370 void dynamic1(float *a, float *b, float *c, float *d) {
371 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(ptr [[DEFAULT_LOC:[@%].+]])
372   #pragma omp for schedule(nonmonotonic: dynamic)
373 // CHECK: call void @__kmpc_dispatch_init_8u(ptr [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741859, i64 0, i64 16908287, i64 1, i64 1)
374 //
375 // CHECK: [[HASWORK:%.+]] = call i32 @__kmpc_dispatch_next_8u(ptr [[DEFAULT_LOC]], i32 [[GTID]], ptr [[OMP_ISLAST:%[^,]+]], ptr [[OMP_LB:%[^,]+]], ptr [[OMP_UB:%[^,]+]], ptr [[OMP_ST:%[^,]+]])
376 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
377 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
378 
379 // Loop header
380 // CHECK: [[O_LOOP1_BODY]]
381 // CHECK: [[LB:%.+]] = load i64, ptr [[OMP_LB]]
382 // CHECK-NEXT: store i64 [[LB]], ptr [[OMP_IV:[^,]+]]
383 // CHECK: [[IV:%.+]] = load i64, ptr [[OMP_IV]]
384 
385 // CHECK-NEXT: [[UB:%.+]] = load i64, ptr [[OMP_UB]]
386 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1
387 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]]
388 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
389   for (unsigned long long i = 131071; i < 2147483647; i += 127) {
390 // CHECK: [[LOOP1_BODY]]
391 // Start of body: calculate i from IV:
392 // CHECK: [[IV1_1:%.+]] = load i64, ptr [[OMP_IV]]
393 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127
394 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]]
395 // CHECK-NEXT: store i64 [[CALC_I_2]], ptr [[LC_I:.+]]
396 // ... loop body ...
397 // End of body: store into a[i]:
398 // CHECK: store float [[RESULT:%.+]], ptr {{%.+}}!llvm.access.group
399     a[i] = b[i] * c[i] * d[i];
400 // CHECK: [[IV1_2:%.+]] = load i64, ptr [[OMP_IV]]{{.*}}
401 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
402 // CHECK-NEXT: store i64 [[ADD1_2]], ptr [[OMP_IV]]
403 // CHECK-NEXT: br label %{{.+}}
404   }
405 // CHECK: [[LOOP1_END]]
406 // CHECK: [[O_LOOP1_END]]
407 // CHECK: call {{.+}} @__kmpc_barrier(ptr [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
408 // CHECK: ret void
409 }
410 
411 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}})
guided7(float * a,float * b,float * c,float * d)412 void guided7(float *a, float *b, float *c, float *d) {
413 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(ptr [[DEFAULT_LOC:[@%].+]])
414   #pragma omp for schedule(guided, 7)
415 // OMP45: call void @__kmpc_dispatch_init_8u(ptr [[DEFAULT_LOC]], i32 [[GTID]], i32 36, i64 0, i64 16908287, i64 1, i64 7)
416 // OMP5: call void @__kmpc_dispatch_init_8u(ptr [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741860, i64 0, i64 16908287, i64 1, i64 7)
417 // OMP51: call void @__kmpc_dispatch_init_8u(ptr [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741860, i64 0, i64 16908287, i64 1, i64 7)
418 //
419 // CHECK: [[HASWORK:%.+]] = call i32 @__kmpc_dispatch_next_8u(ptr [[DEFAULT_LOC]], i32 [[GTID]], ptr [[OMP_ISLAST:%[^,]+]], ptr [[OMP_LB:%[^,]+]], ptr [[OMP_UB:%[^,]+]], ptr [[OMP_ST:%[^,]+]])
420 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
421 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
422 
423 // Loop header
424 // CHECK: [[O_LOOP1_BODY]]
425 // CHECK: [[LB:%.+]] = load i64, ptr [[OMP_LB]]
426 // CHECK-NEXT: store i64 [[LB]], ptr [[OMP_IV:[^,]+]]
427 // CHECK: [[IV:%.+]] = load i64, ptr [[OMP_IV]]
428 
429 // CHECK-NEXT: [[UB:%.+]] = load i64, ptr [[OMP_UB]]
430 // CHECK-NEXT: [[BOUND:%.+]] = add i64 [[UB]], 1
431 // CHECK-NEXT: [[CMP:%.+]] = icmp ult i64 [[IV]], [[BOUND]]
432 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
433   for (unsigned long long i = 131071; i < 2147483647; i += 127) {
434 // CHECK: [[LOOP1_BODY]]
435 // Start of body: calculate i from IV:
436 // CHECK: [[IV1_1:%.+]] = load i64, ptr [[OMP_IV]]
437 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i64 [[IV1_1]], 127
438 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i64 131071, [[CALC_I_1]]
439 // CHECK-NEXT: store i64 [[CALC_I_2]], ptr [[LC_I:.+]]
440 // ... loop body ...
441 // End of body: store into a[i]:
442 // CHECK: store float [[RESULT:%.+]], ptr {{%.+}}!llvm.access.group
443     a[i] = b[i] * c[i] * d[i];
444 // CHECK: [[IV1_2:%.+]] = load i64, ptr [[OMP_IV]]{{.*}}
445 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
446 // CHECK-NEXT: store i64 [[ADD1_2]], ptr [[OMP_IV]]
447 // CHECK-NEXT: br label %{{.+}}
448   }
449 // CHECK: [[LOOP1_END]]
450 // CHECK: [[O_LOOP1_END]]
451 // CHECK: call {{.+}} @__kmpc_barrier(ptr [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
452 // CHECK: ret void
453 }
454 
455 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}})
test_auto(float * a,float * b,float * c,float * d)456 void test_auto(float *a, float *b, float *c, float *d) {
457   unsigned int x = 0;
458   unsigned int y = 0;
459 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(ptr [[DEFAULT_LOC:[@%].+]])
460   #pragma omp for schedule(auto) collapse(2)
461 // OMP45: call void @__kmpc_dispatch_init_8(ptr [[DEFAULT_LOC]], i32 [[GTID]], i32 38, i64 0, i64 [[LAST_ITER:%[^,]+]], i64 1, i64 1)
462 // OMP5: call void @__kmpc_dispatch_init_8(ptr [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741862, i64 0, i64 [[LAST_ITER:%[^,]+]], i64 1, i64 1)
463 // OMP51: call void @__kmpc_dispatch_init_8(ptr [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741862, i64 0, i64 [[LAST_ITER:%[^,]+]], i64 1, i64 1)
464 //
465 // CHECK: [[HASWORK:%.+]] = call i32 @__kmpc_dispatch_next_8(ptr [[DEFAULT_LOC]], i32 [[GTID]], ptr [[OMP_ISLAST:%[^,]+]], ptr [[OMP_LB:%[^,]+]], ptr [[OMP_UB:%[^,]+]], ptr [[OMP_ST:%[^,]+]])
466 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
467 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
468 
469 // Loop header
470 // CHECK: [[O_LOOP1_BODY]]
471 // CHECK: [[LB:%.+]] = load i64, ptr [[OMP_LB]]
472 // CHECK-NEXT: store i64 [[LB]], ptr [[OMP_IV:[^,]+]]
473 // CHECK: [[IV:%.+]] = load i64, ptr [[OMP_IV]]
474 
475 // CHECK-NEXT: [[UB:%.+]] = load i64, ptr [[OMP_UB]]
476 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i64 [[IV]], [[UB]]
477 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
478 // FIXME: When the iteration count of some nested loop is not a known constant,
479 // we should pre-calculate it, like we do for the total number of iterations!
480   for (char i = static_cast<char>(y); i <= '9'; ++i)
481     for (x = 11; x > 0; --x) {
482 // CHECK: [[LOOP1_BODY]]
483 // Start of body: indices are calculated from IV:
484 // CHECK: store i8 {{%[^,]+}}, ptr {{%[^,]+}}
485 // CHECK: store i32 {{%[^,]+}}, ptr {{%[^,]+}}
486 // ... loop body ...
487 // End of body: store into a[i]:
488 // CHECK: store float [[RESULT:%.+]], ptr {{%.+}}
489 // CHECK-NOT: !llvm.access.group
490     a[i] = b[i] * c[i] * d[i];
491 // CHECK: [[IV1_2:%.+]] = load i64, ptr [[OMP_IV]]{{.*}}
492 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1
493 // CHECK-NEXT: store i64 [[ADD1_2]], ptr [[OMP_IV]]
494 // CHECK-NEXT: br label %{{.+}}
495   }
496 // CHECK: [[LOOP1_END]]
497 // CHECK: [[O_LOOP1_END]]
498 // CHECK: call {{.+}} @__kmpc_barrier(ptr [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
499 // CHECK: ret void
500 }
501 
502 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}}, ptr noundef {{.+}})
runtime(float * a,float * b,float * c,float * d)503 void runtime(float *a, float *b, float *c, float *d) {
504   int x = 0;
505 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(ptr [[DEFAULT_LOC:[@%].+]])
506   #pragma omp for collapse(2) schedule(runtime)
507 
508 // OMP45: call void @__kmpc_dispatch_init_4(ptr [[DEFAULT_LOC]], i32 [[GTID]], i32 37, i32 0, i32 199, i32 1, i32 1)
509 // OMP5: call void @__kmpc_dispatch_init_4(ptr [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741861, i32 0, i32 199, i32 1, i32 1)
510 // OMP51: call void @__kmpc_dispatch_init_4(ptr [[DEFAULT_LOC]], i32 [[GTID]], i32 1073741861, i32 0, i32 199, i32 1, i32 1)
511 //
512 // CHECK: [[HASWORK:%.+]] = call i32 @__kmpc_dispatch_next_4(ptr [[DEFAULT_LOC]], i32 [[GTID]], ptr [[OMP_ISLAST:%[^,]+]], ptr [[OMP_LB:%[^,]+]], ptr [[OMP_UB:%[^,]+]], ptr [[OMP_ST:%[^,]+]])
513 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ne i32 [[HASWORK]], 0
514 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]]
515 
516 // Loop header
517 // CHECK: [[O_LOOP1_BODY]]
518 // CHECK: [[LB:%.+]] = load i32, ptr [[OMP_LB]]
519 // CHECK-NEXT: store i32 [[LB]], ptr [[OMP_IV:[^,]+]]
520 // CHECK: [[IV:%.+]] = load i32, ptr [[OMP_IV]]
521 
522 // CHECK-NEXT: [[UB:%.+]] = load i32, ptr [[OMP_UB]]
523 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]]
524 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]]
525   for (unsigned char i = '0' ; i <= '9'; ++i)
526     for (x = -10; x < 10; ++x) {
527 // CHECK: [[LOOP1_BODY]]
528 // Start of body: indices are calculated from IV:
529 // CHECK: store i8 {{%[^,]+}}, ptr {{%[^,]+}}
530 // CHECK: store i32 {{%[^,]+}}, ptr {{%[^,]+}}
531 // ... loop body ...
532 // End of body: store into a[i]:
533 // CHECK: store float [[RESULT:%.+]], ptr {{%.+}}
534 // CHECK-NOT: !llvm.access.group
535     a[i] = b[i] * c[i] * d[i];
536 // CHECK: [[IV1_2:%.+]] = load i32, ptr [[OMP_IV]]{{.*}}
537 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
538 // CHECK-NEXT: store i32 [[ADD1_2]], ptr [[OMP_IV]]
539 // CHECK-NEXT: br label %{{.+}}
540   }
541 // CHECK: [[LOOP1_END]]
542 // CHECK: [[O_LOOP1_END]]
543 // CHECK: call {{.+}} @__kmpc_barrier(ptr [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
544 // CHECK: ret void
545 }
546 
547 // CHECK-LABEL: test_precond
test_precond()548 void test_precond() {
549   // CHECK: [[A_ADDR:%.+]] = alloca i8,
550   // CHECK: [[I_ADDR:%.+]] = alloca i8,
551   // CHECK: [[CAP:%.+]] = alloca i8,
552   char a = 0;
553   // CHECK: store i8 0,
554   // CHECK: store i32
555   // CHECK: store i8
556   // CHECK: [[A:%.+]] = load i8, ptr [[CAP]],
557   // CHECK: [[CONV:%.+]] = sext i8 [[A]] to i32
558   // CHECK: [[CMP:%.+]] = icmp slt i32 [[CONV]], 10
559   // CHECK: br i1 [[CMP]], label %[[PRECOND_THEN:[^,]+]], label %[[PRECOND_END:[^,]+]]
560   // CHECK: [[PRECOND_THEN]]
561   // CHECK: call void @__kmpc_for_static_init_4
562 #pragma omp for
563   for(char i = a; i < 10; ++i);
564   // CHECK: call void @__kmpc_for_static_fini
565   // CHECK: [[PRECOND_END]]
566 }
567 
568 // TERM_DEBUG-LABEL: foo
foo()569 int foo() { extern void mayThrow(); mayThrow(); return 0;};
570 
571 // TERM_DEBUG-LABEL: parallel_for
parallel_for(float * a)572 void parallel_for(float *a) {
573 #pragma omp parallel
574 #pragma omp for schedule(static, 5)
575   // TERM_DEBUG-NOT: __kmpc_global_thread_num
576   // TERM_DEBUG:     call void @__kmpc_for_static_init_4u({{.+}}), !dbg [[DBG_LOC:![0-9]+]]
577   // TERM_DEBUG:     invoke noundef i32 {{.*}}foo{{.*}}()
578   // TERM_DEBUG:     unwind label %[[TERM_LPAD:.+]],
579   // TERM_DEBUG-NOT: __kmpc_global_thread_num
580   // TERM_DEBUG:     call void @__kmpc_for_static_fini({{.+}}), !dbg [[DBG_LOC]]
581   // TERM_DEBUG:     call {{.+}} @__kmpc_barrier({{.+}}), !dbg [[DBG_LOC]]
582   // TERM_DEBUG:     [[TERM_LPAD]]
583   // TERM_DEBUG:     call void @__clang_call_terminate
584   // TERM_DEBUG:     unreachable
585   for (unsigned i = 131071; i <= 2147483647; i += 127)
586     a[i] += foo();
587 }
588 // Check source line corresponds to "#pragma omp for schedule(static, 5)" above:
589 // TERM_DEBUG: [[DBG_LOC]] = !DILocation(line: [[@LINE-15]],
590 
591 char i = 1, j = 2, k = 3;
592 // CHECK-LABEL: for_with_global_lcv
for_with_global_lcv()593 void for_with_global_lcv() {
594 // CHECK: alloca i8,
595 // CHECK: [[I_ADDR:%.+]] = alloca i8,
596 // CHECK: alloca i8,
597 // CHECK: [[J_ADDR:%.+]] = alloca i8,
598 
599 // CHECK: call void @__kmpc_for_static_init_4(
600 // CHECK-NOT: [[I]]
601 // CHECK: store i8 %{{.+}}, ptr [[I_ADDR]]
602 // CHECK-NOT: [[I]]
603 // CHECK: [[I_VAL:%.+]] = load i8, ptr [[I_ADDR]],
604 // CHECK-NOT: [[I]]
605 // CHECK: store i8 [[I_VAL]], ptr [[K]]
606 // CHECK-NOT: [[I]]
607 // CHECK: call void @__kmpc_for_static_fini(
608 // CHECK: call void @__kmpc_barrier(
609 #pragma omp for
610   for (i = 0; i < 2; ++i) {
611     k = i;
612   }
613 // CHECK: call void @__kmpc_for_static_init_4(
614 // CHECK-NOT: [[J]]
615 // CHECK: store i8 %{{.+}}, ptr [[J_ADDR]]
616 // CHECK-NOT: [[J]]
617 // CHECK: [[J_VAL:%.+]] = load i8, ptr [[J_ADDR]],
618 // CHECK-NOT: [[J]]
619 // CHECK: store i8 [[J_VAL]], ptr [[K]]
620 // CHECK-NOT: [[J]]
621 // CHECK: call void @__kmpc_for_static_fini(
622 #pragma omp for collapse(2)
623   for (int i = 0; i < 2; ++i)
624   for (j = 0; j < 2; ++j) {
625     k = i;
626     k = j;
627   }
628   char &cnt = i;
629 #pragma omp for
630   for (cnt = 0; cnt < 2; ++cnt)
631     k = cnt;
632 }
633 
634 // CHECK-LABEL: for_with_references
for_with_references()635 void for_with_references() {
636 // CHECK: [[I:%.+]] = alloca i8,
637 // CHECK: [[CNT:%.+]] = alloca ptr,
638 // CHECK: [[CNT_PRIV:%.+]] = alloca i8,
639 // CHECK: call void @__kmpc_for_static_init_8(
640 // CHECK-NOT: load i8, ptr [[CNT]],
641 // CHECK: call void @__kmpc_for_static_fini(
642   char i = 0;
643   char &cnt = i;
644 #pragma omp for collapse(2)
645   for (cnt = 0; cnt < 2; ++cnt)
646     for (int j = cnt; j < 4 + cnt; j++)
647     k = cnt;
648 }
649 
650 // CHECK-LABEL: for_with_references_dep_cond
for_with_references_dep_cond()651 void for_with_references_dep_cond() {
652 // CHECK: [[I:%.+]] = alloca i8,
653 // CHECK: [[CNT:%.+]] = alloca ptr,
654 // CHECK: [[CNT_PRIV:%.+]] = alloca i8,
655 // CHECK: call void @__kmpc_for_static_init_8(
656 // CHECK-NOT: load i8, ptr [[CNT]],
657 // CHECK: call void @__kmpc_for_static_fini(
658   char i = 0;
659   char &cnt = i;
660 #pragma omp for collapse(2)
661   for (cnt = 0; cnt < 2; ++cnt)
662     for (int j = 0; j < 4 + cnt; j++)
663     k = cnt;
664 }
665 
666 struct Bool {
BoolBool667   Bool(bool b) : b(b) {}
operator boolBool668   operator bool() const { return b; }
669   const bool b;
670 };
671 
672 template <typename T>
673 struct It {
ItIt674   It() : p(0) {}
675   It(const It &, int = 0) ;
676   template <typename U>
677   It(U &, int = 0) ;
678   It &operator=(const It &);
679   It &operator=(It &);
~ItIt680   ~It() {}
681 
ItIt682   It(T *p) : p(p) {}
683 
operator T*&It684   operator T *&() { return p; }
operator T*It685   operator T *() const { return p; }
operator ->It686   T *operator->() const { return p; }
687 
operator ++It688   It &operator++() { ++p; return *this; }
operator --It689   It &operator--() { --p; return *this; }
operator +=It690   It &operator+=(unsigned n) { p += n; return *this; }
operator -=It691   It &operator-=(unsigned n) { p -= n; return *this; }
692 
693   T *p;
694 };
695 
696 template <typename T>
operator +(It<T> a,typename It<T>::difference_type n)697 It<T> operator+(It<T> a, typename It<T>::difference_type n) { return a.p + n; }
698 
699 template <typename T>
operator +(typename It<T>::difference_type n,It<T> a)700 It<T> operator+(typename It<T>::difference_type n, It<T> a) { return a.p + n; }
701 
702 template <typename T>
operator -(It<T> a,typename It<T>::difference_type n)703 It<T> operator-(It<T> a, typename It<T>::difference_type n) { return a.p - n; }
704 
705 typedef Bool BoolType;
706 
707 template <typename T>
operator <(It<T> a,It<T> b)708 BoolType operator<(It<T> a, It<T> b) { return a.p < b.p; }
709 
loop_with_It(It<char> begin,It<char> end)710 void loop_with_It(It<char> begin, It<char> end) {
711 #pragma omp for
712   for (It<char> it = begin; it < end; ++it) {
713     *it = 0;
714   }
715 }
716 
717 // CHECK-LABEL: loop_with_It
718 // CHECK: call i32 @__kmpc_global_thread_num(
719 // CHECK: call void @__kmpc_for_static_init_8(
720 // CHECK: call void @__kmpc_for_static_fini(
721 
loop_with_It_plus(It<char> begin,It<char> end)722 void loop_with_It_plus(It<char> begin, It<char> end) {
723 #pragma omp for
724   for (It<char> it = begin; it < end; it+=1u) {
725     *it = 0;
726   }
727 }
728 
729 // CHECK-LABEL: loop_with_It_plus
730 // CHECK: call i32 @__kmpc_global_thread_num(
731 // CHECK: call void @__kmpc_for_static_init_8(
732 // CHECK: call void @__kmpc_for_static_fini(
733 
loop_with_stmt_expr()734 void loop_with_stmt_expr() {
735 #pragma omp for collapse(2)
736   for (int i = __extension__({float b = 0;b; }); i < __extension__({double c = 1;c; }); i += __extension__({char d = 1; d; }))
737     for (int j = i; j < 4 + i; j++)
738     ;
739 }
740 // CHECK-LABEL: loop_with_stmt_expr
741 // CHECK: call i32 @__kmpc_global_thread_num(
742 // CHECK: call void @__kmpc_for_static_init_8(
743 // CHECK: call void @__kmpc_for_static_fini(
744 
745 
746 // CHECK-LABEL: fint
747 // CHECK: call {{.*}}i32 {{.*}}ftemplate
748 // CHECK: ret i32
749 
750 // CHECK: load i16, ptr
751 // CHECK: store i16 %
752 // CHECK: call void {{.+}}@__kmpc_fork_call(
753 // CHECK: call void @__kmpc_for_static_init_4(
754 template <typename T>
ftemplate()755 T ftemplate() {
756   short aa = 0;
757 
758 #pragma omp parallel for schedule(static, aa)
759   for (int i = 0; i < 100; i++) {
760   }
761   return T();
762 }
763 
fint(void)764 int fint(void) { return ftemplate<int>(); }
765 
766 // Check for imperfectly loop nests codegen.
767 #if _OPENMP == 201811
768 void first();
769 void last();
770 void inner_f();
771 void inner_l();
772 void body_f();
773 
774 // OMP51-NOT: imperfectly_nested_loop
775 // OMP5-LABEL: imperfectly_nested_loop
imperfectly_nested_loop()776 void imperfectly_nested_loop() {
777   // OMP5: call void @__kmpc_for_static_init_4(
778 #pragma omp for collapse(3) order(concurrent)
779   for (int i = 0; i < 10; ++i) {
780     {
781       int a, d;
782       // OMP5: invoke void @{{.+}}first{{.+}}()
783       first();
784       // OMP5: load i32{{.*}}!llvm.access.group ![[AG:[0-9]+]]
785       // OMP5: store i32{{.*}}!llvm.access.group ![[AG]]
786       a = d;
787       for (int j = 0; j < 10; ++j) {
788         int a, d;
789         // OMP5: invoke void @{{.+}}inner_f{{.+}}()
790         inner_f();
791         // OMP5: load i32{{.*}}!llvm.access.group ![[AG]]
792         // OMP5: store i32{{.*}}!llvm.access.group ![[AG]]
793         a = d;
794         for (int k = 0; k < 10; ++k) {
795           int a, d;
796           // OMP5: invoke void @{{.+}}body_f{{.+}}()
797           body_f();
798           // OMP5: load i32{{.*}}!llvm.access.group ![[AG]]
799           // OMP5: store i32{{.*}}!llvm.access.group ![[AG]]
800           a = d;
801         }
802         // OMP5: invoke void @{{.+}}inner_l{{.+}}()
803         inner_l();
804       }
805       // OMP5: invoke void @{{.+}}last{{.+}}()
806       last();
807     }
808   }
809   // OMP5: call void @__kmpc_for_static_fini(
810 }
811 
812 // OMP5: ![[AG]] = distinct !{}
813 // OMP5: !{!"llvm.loop.parallel_accesses", ![[AG]]}
814 
815 #endif
816 
817 #endif // HEADER
818