1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefixes=CHECK,CHECK-NORMAL
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 --check-prefixes=CHECK,CHECK-NORMAL
4 
5 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-enable-irbuilder -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefixes=CHECK,CHECK-IRBUILDER
6 // RUN: %clang_cc1 -fopenmp -fopenmp-enable-irbuilder -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
7 // RUN: %clang_cc1 -fopenmp -fopenmp-enable-irbuilder -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK,CHECK-IRBUILDER
8 
9 // 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
10 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
11 // 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
12 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
13 // expected-no-diagnostics
14 
15 #ifndef HEADER
16 #define HEADER
17 
18 // CHECK: [[KMP_DIM:%.+]] = type { i64, i64, i64 }
19 extern int n;
20 int a[10], b[10], c[10], d[10];
21 void foo();
22 
23 // CHECK-LABEL:bar
24 void bar() {
25   int i,j;
26 // CHECK: call void @__kmpc_doacross_init(
27 // CHECK: call void @__kmpc_doacross_fini(
28 #pragma omp parallel for ordered(2)
29   for (i = 0; i < n; ++i)
30   for (j = 0; j < n; ++j)
31     a[i] = b[i] + 1;
32 // CHECK: call void @__kmpc_doacross_init(
33 // CHECK: call void @__kmpc_doacross_fini(
34  #pragma omp for collapse(2) ordered(2)
35   for (int i = 0; i < n; i++)
36     for (int j = 0; j < n; j++)
37       ;
38 }
39 
40 // CHECK-LABEL: @main()
41 int main() {
42   int i;
43 // CHECK: [[DIMS:%.+]] = alloca [1 x [[KMP_DIM]]],
44 // CHECK-NORMAL: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
45 // CHECK: icmp
46 // CHECK-NEXT: br i1 %
47 // CHECK: [[CAST:%.+]] = bitcast [1 x [[KMP_DIM]]]* [[DIMS]] to i8*
48 // CHECK: call void @llvm.memset.p0i8.i64(i8* align 8 [[CAST]], i8 0, i64 24, i1 false)
49 // CHECK: [[DIM:%.+]] = getelementptr inbounds [1 x [[KMP_DIM]]], [1 x [[KMP_DIM]]]* [[DIMS]], i64 0, i64 0
50 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 1
51 // CHECK: store i64 %{{.+}}, i64* %
52 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 2
53 // CHECK: store i64 1, i64* %
54 // CHECK: [[DIM:%.+]] = getelementptr inbounds [1 x [[KMP_DIM]]], [1 x [[KMP_DIM]]]* [[DIMS]], i64 0, i64 0
55 // CHECK: [[CAST:%.+]] = bitcast [[KMP_DIM]]* [[DIM]] to i8*
56 // CHECK-NORMAL: call void @__kmpc_doacross_init([[IDENT]], i32 [[GTID]], i32 1, i8* [[CAST]])
57 // CHECK-NORMAL: call void @__kmpc_for_static_init_4(%struct.ident_t* @{{.+}}, i32 [[GTID]], i32 33, i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32 1, i32 1)
58 #pragma omp for ordered(1)
59   for (int i = 0; i < n; ++i) {
60     a[i] = b[i] + 1;
61     foo();
62 // CHECK: invoke void [[FOO:.+]](
63 // CHECK: load i32, i32* [[I:%.+]],
64 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
65 // CHECK-NEXT: sdiv i32 %{{.+}}, 1
66 // CHECK-NEXT: sext i32 %{{.+}} to i64
67 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[CNT:%.+]], i64 0, i64 0
68 // CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP]],
69 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[CNT]], i64 0, i64 0
70 // CHECK-NORMAL-NEXT: call void @__kmpc_doacross_post([[IDENT]], i32 [[GTID]], i64* [[TMP]])
71 // CHECK-IRBUILDER-NEXT: [[GTID18:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
72 // CHECK-IRBUILDER-NEXT: call void @__kmpc_doacross_post([[IDENT]], i32 [[GTID18]], i64* [[TMP]])
73 #pragma omp ordered depend(source)
74     c[i] = c[i] + 1;
75     foo();
76 // CHECK: invoke void [[FOO]]
77 // CHECK: load i32, i32* [[I]],
78 // CHECK-NEXT: sub nsw i32 %{{.+}}, 2
79 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
80 // CHECK-NEXT: sdiv i32 %{{.+}}, 1
81 // CHECK-NEXT: sext i32 %{{.+}} to i64
82 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[CNT:%.+]], i64 0, i64 0
83 // CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP]],
84 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[CNT]], i64 0, i64 0
85 // CHECK-NORMAL-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID]], i64* [[TMP]])
86 // CHECK-IRBUILDER-NEXT: [[GTID30:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
87 // CHECK-IRBUILDER-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID30]], i64* [[TMP]])
88 #pragma omp ordered depend(sink : i - 2)
89     d[i] = a[i - 2];
90   }
91   // CHECK: landingpad
92   // CHECK-NORMAL: call void @__kmpc_doacross_fini([[IDENT]], i32 [[GTID]])
93   // CHECK: br label %
94 
95   // CHECK: call void @__kmpc_for_static_fini(
96   // CHECK-NORMAL: call void @__kmpc_doacross_fini([[IDENT]], i32 [[GTID]])
97   // CHECK: ret i32 0
98   return 0;
99 }
100 
101 // CHECK-LABEL: main1
102 int main1() {
103 // CHECK: [[DIMS:%.+]] = alloca [1 x [[KMP_DIM]]],
104 // CHECK-NORMAL: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
105 // CHECK: icmp
106 // CHECK-NEXT: br i1 %
107 // CHECK: [[CAST:%.+]] = bitcast [1 x [[KMP_DIM]]]* [[DIMS]] to i8*
108 // CHECK: call void @llvm.memset.p0i8.i64(i8* align 8 [[CAST]], i8 0, i64 24, i1 false)
109 // CHECK: [[DIM:%.+]] = getelementptr inbounds [1 x [[KMP_DIM]]], [1 x [[KMP_DIM]]]* [[DIMS]], i64 0, i64 0
110 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 1
111 // CHECK: store i64 %{{.+}}, i64* %
112 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 2
113 // CHECK: store i64 1, i64* %
114 // CHECK: [[DIM:%.+]] = getelementptr inbounds [1 x [[KMP_DIM]]], [1 x [[KMP_DIM]]]* [[DIMS]], i64 0, i64 0
115 // CHECK: [[CAST:%.+]] = bitcast [[KMP_DIM]]* [[DIM]] to i8*
116 // CHECK-NORMAL: call void @__kmpc_doacross_init([[IDENT]], i32 [[GTID]], i32 1, i8* [[CAST]])
117 // CHECK-NORMAL: call void @__kmpc_for_static_init_4(%struct.ident_t* @{{.+}}, i32 [[GTID]], i32 33, i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32 1, i32 1)
118 #pragma omp for ordered(1)
119   for (int i = n; i > 0; --i) {
120     a[i] = b[i] + 1;
121     foo();
122 // CHECK: invoke void [[FOO:.+]](
123 // CHECK: [[UB_VAL:%.+]] = load i32, i32* [[UB:%.+]],
124 // CHECK-NEXT: [[I_VAL:%.+]] = load i32, i32* [[I:%.+]],
125 // CHECK-NEXT: sub i32 [[UB_VAL]], [[I_VAL]]
126 // CHECK-NEXT: udiv i32 %{{.+}}, 1
127 // CHECK-NEXT: zext i32 %{{.+}} to i64
128 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[CNT:%.+]], i64 0, i64 0
129 // CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP]],
130 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[CNT]], i64 0, i64 0
131 // CHECK-NORMAL-NEXT: call void @__kmpc_doacross_post([[IDENT]], i32 [[GTID]], i64* [[TMP]])
132 // CHECK-IRBUILDER-NEXT: [[GTID17:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
133 // CHECK-IRBUILDER-NEXT: call void @__kmpc_doacross_post([[IDENT]], i32 [[GTID17]], i64* [[TMP]])
134 #pragma omp ordered depend(source)
135     c[i] = c[i] + 1;
136     foo();
137 // CHECK: invoke void [[FOO]]
138 // CHECK: [[UB_VAL:%.+]] = load i32, i32* [[UB]],
139 // CHECK-NEXT: [[I_VAL:%.+]] = load i32, i32* [[I]],
140 // CHECK-NEXT: [[SUB:%.+]] = sub nsw i32 [[I_VAL]], 2
141 // CHECK-NEXT: sub i32 [[UB_VAL]], [[SUB]]
142 // CHECK-NEXT: udiv i32 %{{.+}}, 1
143 // CHECK-NEXT: zext i32 %{{.+}} to i64
144 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[CNT:%.+]], i64 0, i64 0
145 // CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP]],
146 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[CNT]], i64 0, i64 0
147 // CHECK-NORMAL-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID]], i64* [[TMP]])
148 // CHECK-IRBUILDER-NEXT: [[GTID29:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
149 // CHECK-IRBUILDER-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID29]], i64* [[TMP]])
150 #pragma omp ordered depend(sink : i - 2)
151     d[i] = a[i - 2];
152   }
153   // CHECK: landingpad
154   // CHECK-NORMAL: call void @__kmpc_doacross_fini([[IDENT]], i32 [[GTID]])
155   // CHECK: br label %
156 
157   // CHECK: call void @__kmpc_for_static_fini(
158   // CHECK-NORMAL: call void @__kmpc_doacross_fini([[IDENT]], i32 [[GTID]])
159   // CHECK: ret i32 0
160   return 0;
161 }
162 
163 // CHECK: define {{.+}}TestStruct
164 template <typename T>
165 struct TestStruct {
166   static const int M = 10;
167   static const int N = 20;
168   T i;
169   T a[N][M];
170   T b[N][M];
171   T foo(T, T);
172   T bar(T, T, T);
173   void baz(T, T);
174   TestStruct() {
175 // CHECK: [[DIMS:%.+]] = alloca [2 x [[KMP_DIM]]],
176 // CHECK-NORMAL: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
177 // CHECK: [[CAST:%.+]] = bitcast [2 x [[KMP_DIM]]]* [[DIMS]] to i8*
178 // CHECK: call void @llvm.memset.p0i8.i64(i8* align 8 [[CAST]], i8 0, i64 48, i1 false)
179 // CHECK: [[DIM:%.+]] = getelementptr inbounds [2 x [[KMP_DIM]]], [2 x [[KMP_DIM]]]* [[DIMS]], i64 0, i64 0
180 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 1
181 // CHECK: store i64 10, i64* %
182 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 2
183 // CHECK: store i64 1, i64* %
184 // CHECK: [[DIM:%.+]] = getelementptr inbounds [2 x [[KMP_DIM]]], [2 x [[KMP_DIM]]]* [[DIMS]], i64 0, i64 1
185 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 1
186 // CHECK: store i64 %{{.+}}, i64* %
187 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 2
188 // CHECK: store i64 1, i64* %
189 // CHECK: [[DIM:%.+]] = getelementptr inbounds [2 x [[KMP_DIM]]], [2 x [[KMP_DIM]]]* [[DIMS]], i64 0, i64 0
190 // CHECK: [[CAST:%.+]] = bitcast [[KMP_DIM]]* [[DIM]] to i8*
191 // CHECK-NORMAL: call void @__kmpc_doacross_init([[IDENT]], i32 [[GTID]], i32 2, i8* [[CAST]])
192 // CHECK-NORMAL: call void @__kmpc_for_static_init_4(%struct.ident_t* @{{.+}}, i32 [[GTID]], i32 33, i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32 1, i32 1)
193 #pragma omp for ordered(2)
194     for (T j = 0; j < M; j++)
195       for (i = 0; i < n; i += 2) {
196         a[i][j] = foo(i, j);
197 // CHECK: invoke {{.+TestStruct.+foo}}
198 // CHECK: load i32*, i32** %
199 // CHECK: load i32, i32* %
200 // CHECK: load i32, i32* %
201 // CHECK: load i32, i32* [[J:%.+]],
202 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
203 // CHECK-NEXT: sdiv i32 %{{.+}}, 1
204 // CHECK-NEXT: sext i32 %{{.+}} to i64
205 // CHECK-NORMAL-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT:%.+]], i64 0, i64 0
206 // CHECK-NORMAL-NEXT: store i64 %{{.+}}, i64* [[TMP:%.+]],
207 // CHECK-NEXT: [[I:%.+]] = load i32*, i32** [[I_REF:%.+]],
208 // CHECK-NEXT: load i32, i32* [[I]],
209 // CHECK-NEXT: sub nsw i32 %{{.+}}, 2
210 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
211 // CHECK-NEXT: sdiv i32 %{{.+}}, 2
212 // CHECK-NEXT: sext i32 %{{.+}} to i64
213 // CHECK-IRBUILDER-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT:%.+]], i64 0, i64 0
214 // CHECK-IRBUILDER-NEXT: store i64 %{{.+}}, i64* [[TMP:%.+]],
215 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT]], i64 0, i64 1
216 // CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP]],
217 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT]], i64 0, i64 0
218 // CHECK-NORMAL-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID]], i64* [[TMP]])
219 // CHECK-IRBUILDER-NEXT: [[GTID18:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
220 // CHECK-IRBUILDER-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID18]], i64* [[TMP]])
221 // CHECK-NEXT: load i32, i32* [[J:%.+]],
222 // CHECK-NEXT: sub nsw i32 %{{.+}}, 1
223 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
224 // CHECK-NEXT: sdiv i32 %{{.+}}, 1
225 // CHECK-NEXT: sext i32 %{{.+}} to i64
226 // CHECK-NORMAL-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT:%.+]], i64 0, i64 0
227 // CHECK-NORMAL-NEXT: store i64 %{{.+}}, i64* [[TMP]],
228 // CHECK-NEXT: [[I:%.+]] = load i32*, i32** [[I_REF]],
229 // CHECK-NEXT: load i32, i32* [[I]],
230 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
231 // CHECK-NEXT: sdiv i32 %{{.+}}, 2
232 // CHECK-NEXT: sext i32 %{{.+}} to i64
233 // CHECK-IRBUILDER-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT:%.+]], i64 0, i64 0
234 // CHECK-IRBUILDER-NEXT: store i64 %{{.+}}, i64* [[TMP]],
235 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT]], i64 0, i64 1
236 // CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP]],
237 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT]], i64 0, i64 0
238 // CHECK-NORMAL-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID]], i64* [[TMP]])
239 // CHECK-IRBUILDER-NEXT: [[GTID27:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
240 // CHECK-IRBUILDER-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID27]], i64* [[TMP]])
241 #pragma omp ordered depend(sink : j, i - 2) depend(sink : j - 1, i)
242         b[i][j] = bar(a[i][j], b[i - 1][j], b[i][j - 1]);
243 // CHECK: invoke {{.+TestStruct.+bar}}
244 // CHECK: load i32*, i32** %
245 // CHECK: load i32, i32* %
246 // CHECK: load i32, i32* %
247 // CHECK: load i32, i32* [[J]],
248 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
249 // CHECK-NEXT: sdiv i32 %{{.+}}, 1
250 // CHECK-NEXT: sext i32 %{{.+}} to i64
251 // CHECK-NORMAL-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT:%.+]], i64 0, i64 0
252 // CHECK-NORMAL-NEXT: store i64 %{{.+}}, i64* [[TMP]],
253 // CHECK-NEXT: [[I:%.+]] = load i32*, i32** [[I_REF]],
254 // CHECK-NEXT: load i32, i32* [[I]],
255 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
256 // CHECK-NEXT: sdiv i32 %{{.+}}, 2
257 // CHECK-NEXT: sext i32 %{{.+}} to i64
258 // CHECK-IRBUILDER-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT:%.+]], i64 0, i64 0
259 // CHECK-IRBUILDER-NEXT: store i64 %{{.+}}, i64* [[TMP]],
260 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT]], i64 0, i64 1
261 // CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP]],
262 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT]], i64 0, i64 0
263 // CHECK-NORMAL-NEXT: call void @__kmpc_doacross_post([[IDENT]], i32 [[GTID]], i64* [[TMP]])
264 // CHECK-IRBUILDER-NEXT: [[GTID58:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
265 // CHECK-IRBUILDER-NEXT: call void @__kmpc_doacross_post([[IDENT]], i32 [[GTID58]], i64* [[TMP]])
266 #pragma omp ordered depend(source)
267         baz(a[i][j], b[i][j]);
268       }
269   }
270   // CHECK: landingpad
271   // CHECK-NORMAL: call void @__kmpc_doacross_fini([[IDENT]], i32 [[GTID]])
272   // CHECK: br label %
273 
274   // CHECK: call void @__kmpc_for_static_fini(
275   // CHECK-NORMAL: call void @__kmpc_doacross_fini([[IDENT]], i32 [[GTID]])
276   // CHECK: ret
277 };
278 
279 TestStruct<int> s;
280 #endif // HEADER
281