1 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
3 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
11 // RUN: %clang_cc1 -no-opaque-pointers -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 
92   // CHECK: call void @__kmpc_for_static_fini(
93   // CHECK-NORMAL: call void @__kmpc_doacross_fini([[IDENT]], i32 [[GTID]])
94   // CHECK: ret i32 0
95   return 0;
96 }
97 
98 // CHECK-LABEL: main1
99 int main1() {
100 // CHECK: [[DIMS:%.+]] = alloca [1 x [[KMP_DIM]]],
101 // CHECK-NORMAL: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
102 // CHECK: icmp
103 // CHECK-NEXT: br i1 %
104 // CHECK: [[CAST:%.+]] = bitcast [1 x [[KMP_DIM]]]* [[DIMS]] to i8*
105 // CHECK: call void @llvm.memset.p0i8.i64(i8* align 8 [[CAST]], i8 0, i64 24, i1 false)
106 // CHECK: [[DIM:%.+]] = getelementptr inbounds [1 x [[KMP_DIM]]], [1 x [[KMP_DIM]]]* [[DIMS]], i64 0, i64 0
107 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 1
108 // CHECK: store i64 %{{.+}}, i64* %
109 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 2
110 // CHECK: store i64 1, i64* %
111 // CHECK: [[DIM:%.+]] = getelementptr inbounds [1 x [[KMP_DIM]]], [1 x [[KMP_DIM]]]* [[DIMS]], i64 0, i64 0
112 // CHECK: [[CAST:%.+]] = bitcast [[KMP_DIM]]* [[DIM]] to i8*
113 // CHECK-NORMAL: call void @__kmpc_doacross_init([[IDENT]], i32 [[GTID]], i32 1, i8* [[CAST]])
114 // CHECK-NORMAL: call void @__kmpc_for_static_init_4(%struct.ident_t* @{{.+}}, i32 [[GTID]], i32 33, i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32 1, i32 1)
115 #pragma omp for ordered(1)
116   for (int i = n; i > 0; --i) {
117     a[i] = b[i] + 1;
118     foo();
119 // CHECK: invoke void [[FOO:.+]](
120 // CHECK: [[UB_VAL:%.+]] = load i32, i32* [[UB:%.+]],
121 // CHECK-NEXT: [[I_VAL:%.+]] = load i32, i32* [[I:%.+]],
122 // CHECK-NEXT: sub i32 [[UB_VAL]], [[I_VAL]]
123 // CHECK-NEXT: udiv i32 %{{.+}}, 1
124 // CHECK-NEXT: zext i32 %{{.+}} to i64
125 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[CNT:%.+]], i64 0, i64 0
126 // CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP]],
127 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[CNT]], i64 0, i64 0
128 // CHECK-NORMAL-NEXT: call void @__kmpc_doacross_post([[IDENT]], i32 [[GTID]], i64* [[TMP]])
129 // CHECK-IRBUILDER-NEXT: [[GTID17:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
130 // CHECK-IRBUILDER-NEXT: call void @__kmpc_doacross_post([[IDENT]], i32 [[GTID17]], i64* [[TMP]])
131 #pragma omp ordered depend(source)
132     c[i] = c[i] + 1;
133     foo();
134 // CHECK: invoke void [[FOO]]
135 // CHECK: [[UB_VAL:%.+]] = load i32, i32* [[UB]],
136 // CHECK-NEXT: [[I_VAL:%.+]] = load i32, i32* [[I]],
137 // CHECK-NEXT: [[SUB:%.+]] = sub nsw i32 [[I_VAL]], 2
138 // CHECK-NEXT: sub i32 [[UB_VAL]], [[SUB]]
139 // CHECK-NEXT: udiv i32 %{{.+}}, 1
140 // CHECK-NEXT: zext i32 %{{.+}} to i64
141 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[CNT:%.+]], i64 0, i64 0
142 // CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP]],
143 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [1 x i64], [1 x i64]* [[CNT]], i64 0, i64 0
144 // CHECK-NORMAL-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID]], i64* [[TMP]])
145 // CHECK-IRBUILDER-NEXT: [[GTID29:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
146 // CHECK-IRBUILDER-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID29]], i64* [[TMP]])
147 #pragma omp ordered depend(sink : i - 2)
148     d[i] = a[i - 2];
149   }
150   // CHECK: call void @__kmpc_for_static_fini(
151   // CHECK-NORMAL: call void @__kmpc_doacross_fini([[IDENT]], i32 [[GTID]])
152   // CHECK: ret i32 0
153   return 0;
154 }
155 
156 // CHECK: define {{.+}}TestStruct
157 template <typename T>
158 struct TestStruct {
159   static const int M = 10;
160   static const int N = 20;
161   T i;
162   T a[N][M];
163   T b[N][M];
164   T foo(T, T);
165   T bar(T, T, T);
166   void baz(T, T);
167   TestStruct() {
168 // CHECK: [[DIMS:%.+]] = alloca [2 x [[KMP_DIM]]],
169 // CHECK-NORMAL: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
170 // CHECK: [[CAST:%.+]] = bitcast [2 x [[KMP_DIM]]]* [[DIMS]] to i8*
171 // CHECK: call void @llvm.memset.p0i8.i64(i8* align 8 [[CAST]], i8 0, i64 48, i1 false)
172 // CHECK: [[DIM:%.+]] = getelementptr inbounds [2 x [[KMP_DIM]]], [2 x [[KMP_DIM]]]* [[DIMS]], i64 0, i64 0
173 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 1
174 // CHECK: store i64 10, i64* %
175 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 2
176 // CHECK: store i64 1, i64* %
177 // CHECK: [[DIM:%.+]] = getelementptr inbounds [2 x [[KMP_DIM]]], [2 x [[KMP_DIM]]]* [[DIMS]], i64 0, i64 1
178 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 1
179 // CHECK: store i64 %{{.+}}, i64* %
180 // CHECK: getelementptr inbounds [[KMP_DIM]], [[KMP_DIM]]* [[DIM]], i32 0, i32 2
181 // CHECK: store i64 1, i64* %
182 // CHECK: [[DIM:%.+]] = getelementptr inbounds [2 x [[KMP_DIM]]], [2 x [[KMP_DIM]]]* [[DIMS]], i64 0, i64 0
183 // CHECK: [[CAST:%.+]] = bitcast [[KMP_DIM]]* [[DIM]] to i8*
184 // CHECK-NORMAL: call void @__kmpc_doacross_init([[IDENT]], i32 [[GTID]], i32 2, i8* [[CAST]])
185 // CHECK-NORMAL: call void @__kmpc_for_static_init_4(%struct.ident_t* @{{.+}}, i32 [[GTID]], i32 33, i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32 1, i32 1)
186 #pragma omp for ordered(2)
187     for (T j = 0; j < M; j++)
188       for (i = 0; i < n; i += 2) {
189         a[i][j] = foo(i, j);
190 // CHECK: invoke {{.+TestStruct.+foo}}
191 // CHECK: load i32*, i32** %
192 // CHECK: load i32, i32* %
193 // CHECK: load i32, i32* %
194 // CHECK: load i32, i32* [[J:%.+]],
195 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
196 // CHECK-NEXT: sdiv i32 %{{.+}}, 1
197 // CHECK-NEXT: sext i32 %{{.+}} to i64
198 // CHECK-NORMAL-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT:%.+]], i64 0, i64 0
199 // CHECK-NORMAL-NEXT: store i64 %{{.+}}, i64* [[TMP:%.+]],
200 // CHECK-NEXT: [[I:%.+]] = load i32*, i32** [[I_REF:%.+]],
201 // CHECK-NEXT: load i32, i32* [[I]],
202 // CHECK-NEXT: sub nsw i32 %{{.+}}, 2
203 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
204 // CHECK-NEXT: sdiv i32 %{{.+}}, 2
205 // CHECK-NEXT: sext i32 %{{.+}} to i64
206 // CHECK-IRBUILDER-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT:%.+]], i64 0, i64 0
207 // CHECK-IRBUILDER-NEXT: store i64 %{{.+}}, i64* [[TMP:%.+]],
208 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT]], i64 0, i64 1
209 // CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP]],
210 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT]], i64 0, i64 0
211 // CHECK-NORMAL-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID]], i64* [[TMP]])
212 // CHECK-IRBUILDER-NEXT: [[GTID18:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
213 // CHECK-IRBUILDER-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID18]], i64* [[TMP]])
214 // CHECK-NEXT: load i32, i32* [[J:%.+]],
215 // CHECK-NEXT: sub nsw i32 %{{.+}}, 1
216 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
217 // CHECK-NEXT: sdiv i32 %{{.+}}, 1
218 // CHECK-NEXT: sext i32 %{{.+}} to i64
219 // CHECK-NORMAL-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT:%.+]], i64 0, i64 0
220 // CHECK-NORMAL-NEXT: store i64 %{{.+}}, i64* [[TMP]],
221 // CHECK-NEXT: [[I:%.+]] = load i32*, i32** [[I_REF]],
222 // CHECK-NEXT: load i32, i32* [[I]],
223 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
224 // CHECK-NEXT: sdiv i32 %{{.+}}, 2
225 // CHECK-NEXT: sext i32 %{{.+}} to i64
226 // CHECK-IRBUILDER-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT:%.+]], i64 0, i64 0
227 // CHECK-IRBUILDER-NEXT: store i64 %{{.+}}, i64* [[TMP]],
228 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT]], i64 0, i64 1
229 // CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP]],
230 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT]], i64 0, i64 0
231 // CHECK-NORMAL-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID]], i64* [[TMP]])
232 // CHECK-IRBUILDER-NEXT: [[GTID27:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
233 // CHECK-IRBUILDER-NEXT: call void @__kmpc_doacross_wait([[IDENT]], i32 [[GTID27]], i64* [[TMP]])
234 #pragma omp ordered depend(sink : j, i - 2) depend(sink : j - 1, i)
235         b[i][j] = bar(a[i][j], b[i - 1][j], b[i][j - 1]);
236 // CHECK: invoke {{.+TestStruct.+bar}}
237 // CHECK: load i32*, i32** %
238 // CHECK: load i32, i32* %
239 // CHECK: load i32, i32* %
240 // CHECK: load i32, i32* [[J]],
241 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
242 // CHECK-NEXT: sdiv i32 %{{.+}}, 1
243 // CHECK-NEXT: sext i32 %{{.+}} to i64
244 // CHECK-NORMAL-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT:%.+]], i64 0, i64 0
245 // CHECK-NORMAL-NEXT: store i64 %{{.+}}, i64* [[TMP]],
246 // CHECK-NEXT: [[I:%.+]] = load i32*, i32** [[I_REF]],
247 // CHECK-NEXT: load i32, i32* [[I]],
248 // CHECK-NEXT: sub nsw i32 %{{.+}}, 0
249 // CHECK-NEXT: sdiv i32 %{{.+}}, 2
250 // CHECK-NEXT: sext i32 %{{.+}} to i64
251 // CHECK-IRBUILDER-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT:%.+]], i64 0, i64 0
252 // CHECK-IRBUILDER-NEXT: store i64 %{{.+}}, i64* [[TMP]],
253 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT]], i64 0, i64 1
254 // CHECK-NEXT: store i64 %{{.+}}, i64* [[TMP]],
255 // CHECK-NEXT: [[TMP:%.+]] = getelementptr inbounds [2 x i64], [2 x i64]* [[CNT]], i64 0, i64 0
256 // CHECK-NORMAL-NEXT: call void @__kmpc_doacross_post([[IDENT]], i32 [[GTID]], i64* [[TMP]])
257 // CHECK-IRBUILDER-NEXT: [[GTID58:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT:%.+]])
258 // CHECK-IRBUILDER-NEXT: call void @__kmpc_doacross_post([[IDENT]], i32 [[GTID58]], i64* [[TMP]])
259 #pragma omp ordered depend(source)
260         baz(a[i][j], b[i][j]);
261       }
262   }
263   // CHECK: call void @__kmpc_for_static_fini(
264   // CHECK-NORMAL: call void @__kmpc_doacross_fini([[IDENT]], i32 [[GTID]])
265   // CHECK: ret
266 };
267 
268 TestStruct<int> s;
269 #endif // HEADER
270