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