1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s
2 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s
3 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
4 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -check-prefix=LAMBDA %s
5 // RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -check-prefix=BLOCKS %s
6 // expected-no-diagnostics
7 // REQUIRES: x86-registered-target
8 #ifndef HEADER
9 #define HEADER
10 
11 struct St {
12   int a, b;
13   St() : a(0), b(0) {}
14   St(const St &st) : a(st.a + st.b), b(0) {}
15   ~St() {}
16 };
17 
18 volatile int g = 1212;
19 
20 template <class T>
21 struct S {
22   T f;
23   S(T a) : f(a + g) {}
24   S() : f(g) {}
25   S(const S &s, St t = St()) : f(s.f + t.a) {}
26   operator T() { return T(); }
27   ~S() {}
28 };
29 
30 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float }
31 // CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} }
32 // CHECK-DAG: [[ST_TY:%.+]] = type { i{{[0-9]+}}, i{{[0-9]+}} }
33 
34 template <typename T>
35 T tmain() {
36   S<T> test;
37   T t_var = T();
38   T vec[] = {1, 2};
39   S<T> s_arr[] = {1, 2};
40   S<T> var(3);
41 #pragma omp parallel
42 #pragma omp sections firstprivate(t_var, vec, s_arr, var)
43   {
44     vec[0] = t_var;
45 #pragma omp section
46     s_arr[0] = var;
47   }
48   return T();
49 }
50 
51 // CHECK: [[TEST:@.+]] = global [[S_FLOAT_TY]] zeroinitializer,
52 S<float> test;
53 // CHECK-DAG: [[T_VAR:@.+]] = global i{{[0-9]+}} 333,
54 int t_var = 333;
55 // CHECK-DAG: [[VEC:@.+]] = global [2 x i{{[0-9]+}}] [i{{[0-9]+}} 1, i{{[0-9]+}} 2],
56 int vec[] = {1, 2};
57 // CHECK-DAG: [[S_ARR:@.+]] = global [2 x [[S_FLOAT_TY]]] zeroinitializer,
58 S<float> s_arr[] = {1, 2};
59 // CHECK-DAG: [[VAR:@.+]] = global [[S_FLOAT_TY]] zeroinitializer,
60 S<float> var(3);
61 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8*
62 // CHECK-DAG: [[SECTIONS_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 194, i32 0, i32 0, i8*
63 
64 // CHECK: call {{.*}} [[S_FLOAT_TY_DEF_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]])
65 // CHECK: ([[S_FLOAT_TY]]*)* [[S_FLOAT_TY_DESTR:@[^ ]+]] {{[^,]+}}, {{.+}}([[S_FLOAT_TY]]* [[TEST]]
66 int main() {
67 #ifdef LAMBDA
68   // LAMBDA: [[G:@.+]] = global i{{[0-9]+}} 1212,
69   // LAMBDA-LABEL: @main
70   // LAMBDA: call void [[OUTER_LAMBDA:@.+]](
71   [&]() {
72 // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]](
73 // LAMBDA: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}})
74 #pragma omp parallel
75 #pragma omp sections firstprivate(g)
76   {
77     // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}})
78     // Skip temp vars for loop
79     // LAMBDA: alloca i{{[0-9]+}},
80     // LAMBDA: alloca i{{[0-9]+}},
81     // LAMBDA: alloca i{{[0-9]+}},
82     // LAMBDA: alloca i{{[0-9]+}},
83     // LAMBDA: alloca i{{[0-9]+}},
84     // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}},
85     // LAMBDA: [[G_VAL:%.+]] = load volatile i{{[0-9]+}}, i{{[0-9]+}}* [[G]]
86     // LAMBDA: store i{{[0-9]+}} [[G_VAL]], i{{[0-9]+}}* [[G_PRIVATE_ADDR]]
87     // LAMBDA: call void @__kmpc_barrier(
88     g = 1;
89     // LAMBDA: call void @__kmpc_for_static_init_4(
90     // LAMBDA: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]],
91     // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
92     // LAMBDA: store i{{[0-9]+}}* [[G_PRIVATE_ADDR]], i{{[0-9]+}}** [[G_PRIVATE_ADDR_REF]]
93     // LAMBDA: call void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]])
94     // LAMBDA: call void @__kmpc_for_static_fini(
95     // LAMBDA: call i32 @__kmpc_cancel_barrier(
96 #pragma omp section
97     [&]() {
98       // LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* [[ARG_PTR:%.+]])
99       // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]],
100       g = 2;
101       // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]]
102       // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
103       // LAMBDA: [[G_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G_PTR_REF]]
104       // LAMBDA: store i{{[0-9]+}} 2, i{{[0-9]+}}* [[G_REF]]
105     }();
106   }
107   }();
108   return 0;
109 #elif defined(BLOCKS)
110   // BLOCKS: [[G:@.+]] = global i{{[0-9]+}} 1212,
111   // BLOCKS-LABEL: @main
112   // BLOCKS: call void {{%.+}}(i8
113   ^{
114 // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8*
115 // BLOCKS: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}})
116 #pragma omp parallel
117 #pragma omp sections firstprivate(g)
118    {
119     // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}})
120     // Skip temp vars for loop
121     // BLOCKS: alloca i{{[0-9]+}},
122     // BLOCKS: alloca i{{[0-9]+}},
123     // BLOCKS: alloca i{{[0-9]+}},
124     // BLOCKS: alloca i{{[0-9]+}},
125     // BLOCKS: alloca i{{[0-9]+}},
126     // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}},
127     // BLOCKS: [[G_VAL:%.+]] = load volatile i{{[0-9]+}}, i{{[0-9]+}}* [[G]]
128     // BLOCKS: store i{{[0-9]+}} [[G_VAL]], i{{[0-9]+}}* [[G_PRIVATE_ADDR]]
129     // BLOCKS: call void @__kmpc_barrier(
130     g = 1;
131     // BLOCKS: call void @__kmpc_for_static_init_4(
132     // BLOCKS: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]],
133     // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
134     // BLOCKS: i{{[0-9]+}}* [[G_PRIVATE_ADDR]]
135     // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
136     // BLOCKS: call void {{%.+}}(i8
137     // BLOCKS: call void @__kmpc_for_static_fini(
138     // BLOCKS: call i32 @__kmpc_cancel_barrier(
139 #pragma omp section
140     ^{
141       // BLOCKS: define {{.+}} void {{@.+}}(i8*
142       g = 2;
143       // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
144       // BLOCKS: store i{{[0-9]+}} 2, i{{[0-9]+}}*
145       // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
146       // BLOCKS: ret
147     }();
148   }
149   }();
150   return 0;
151 #else
152 #pragma omp sections firstprivate(t_var, vec, s_arr, var) nowait
153   {
154     {
155     vec[0] = t_var;
156     s_arr[0] = var;
157     }
158   }
159   return tmain<int>();
160 #endif
161 }
162 
163 // CHECK: define {{.*}}i{{[0-9]+}} @main()
164 // CHECK: alloca i{{[0-9]+}},
165 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(
166 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}},
167 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}],
168 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]],
169 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]],
170 
171 // CHECK: call i32 @__kmpc_single(
172 // firstprivate t_var(t_var)
173 // CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR]],
174 // CHECK: store i{{[0-9]+}} [[T_VAR_VAL]], i{{[0-9]+}}* [[T_VAR_PRIV]],
175 
176 // firstprivate vec(vec)
177 // CHECK: [[VEC_DEST:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_PRIV]] to i8*
178 // CHECK: call void @llvm.memcpy.{{.+}}(i8* [[VEC_DEST]], i8* bitcast ([2 x i{{[0-9]+}}]* [[VEC]] to i8*),
179 
180 // firstprivate s_arr(s_arr)
181 // CHECK: [[S_ARR_PRIV_BEGIN:%.+]] = getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[S_ARR_PRIV]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
182 // CHECK: [[S_ARR_PRIV_END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[S_ARR_PRIV_BEGIN]], i{{[0-9]+}} 2
183 // CHECK: [[IS_EMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[S_ARR_PRIV_BEGIN]], [[S_ARR_PRIV_END]]
184 // CHECK: br i1 [[IS_EMPTY]], label %[[S_ARR_BODY_DONE:.+]], label %[[S_ARR_BODY:.+]]
185 // CHECK: [[S_ARR_BODY]]
186 // CHECK: getelementptr inbounds ([2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[S_ARR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0)
187 // CHECK: call {{.*}} [[ST_TY_DEFAULT_CONSTR:@.+]]([[ST_TY]]* [[ST_TY_TEMP:%.+]])
188 // CHECK: call {{.*}} [[S_FLOAT_TY_COPY_CONSTR:@.+]]([[S_FLOAT_TY]]* {{.+}}, [[S_FLOAT_TY]]* {{.+}}, [[ST_TY]]* [[ST_TY_TEMP]])
189 // CHECK: call {{.*}} [[ST_TY_DESTR:@.+]]([[ST_TY]]* [[ST_TY_TEMP]])
190 // CHECK: br i1 {{.+}}, label %{{.+}}, label %[[S_ARR_BODY]]
191 
192 // firstprivate var(var)
193 // CHECK: call {{.*}} [[ST_TY_DEFAULT_CONSTR]]([[ST_TY]]* [[ST_TY_TEMP:%.+]])
194 // CHECK: call {{.*}} [[S_FLOAT_TY_COPY_CONSTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]], [[S_FLOAT_TY]]* {{.*}} [[VAR]], [[ST_TY]]* [[ST_TY_TEMP]])
195 // CHECK: call {{.*}} [[ST_TY_DESTR]]([[ST_TY]]* [[ST_TY_TEMP]])
196 
197 // ~(firstprivate var), ~(firstprivate s_arr)
198 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]])
199 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]*
200 // CHECK: call void @__kmpc_end_single(
201 
202 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]])
203 
204 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]()
205 
206 // CHECK: ret void
207 
208 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]()
209 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]],
210 // CHECK: call {{.*}} [[S_INT_TY_DEF_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]])
211 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i32*, [2 x i32]*, [2 x [[S_INT_TY]]]*, [[S_INT_TY]]*)* [[TMAIN_MICROTASK:@.+]] to void
212 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]*
213 // CHECK: ret
214 //
215 // CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}},
216 // Skip temp vars for loop
217 // CHECK: alloca i{{[0-9]+}},
218 // CHECK: alloca i{{[0-9]+}},
219 // CHECK: alloca i{{[0-9]+}},
220 // CHECK: alloca i{{[0-9]+}},
221 // CHECK: alloca i{{[0-9]+}},
222 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}},
223 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}],
224 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]],
225 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]],
226 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
227 
228 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %
229 // CHECK: [[VEC_REF:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** %
230 // CHECK: [[S_ARR:%.+]] = load [2 x [[S_INT_TY]]]*, [2 x [[S_INT_TY]]]** %
231 // CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** %
232 
233 // firstprivate t_var(t_var)
234 // CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_REF]],
235 // CHECK: store i{{[0-9]+}} [[T_VAR_VAL]], i{{[0-9]+}}* [[T_VAR_PRIV]],
236 
237 // firstprivate vec(vec)
238 // CHECK: [[VEC_DEST:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_PRIV]] to i8*
239 // CHECK: [[VEC_SRC:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_REF]] to i8*
240 // CHECK: call void @llvm.memcpy.{{.+}}(i8* [[VEC_DEST]], i8* [[VEC_SRC]],
241 
242 // firstprivate s_arr(s_arr)
243 // CHECK: [[S_ARR_PRIV_BEGIN:%.+]] = getelementptr inbounds [2 x [[S_INT_TY]]], [2 x [[S_INT_TY]]]* [[S_ARR_PRIV]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
244 // CHECK: [[S_ARR_PRIV_END:%.+]] = getelementptr [[S_INT_TY]], [[S_INT_TY]]* [[S_ARR_PRIV_BEGIN]], i{{[0-9]+}} 2
245 // CHECK: [[IS_EMPTY:%.+]] = icmp eq [[S_INT_TY]]* [[S_ARR_PRIV_BEGIN]], [[S_ARR_PRIV_END]]
246 // CHECK: br i1 [[IS_EMPTY]], label %[[S_ARR_BODY_DONE:.+]], label %[[S_ARR_BODY:.+]]
247 // CHECK: [[S_ARR_BODY]]
248 // CHECK: call {{.*}} [[ST_TY_DEFAULT_CONSTR:@.+]]([[ST_TY]]* [[ST_TY_TEMP:%.+]])
249 // CHECK: call {{.*}} [[S_INT_TY_COPY_CONSTR:@.+]]([[S_INT_TY]]* {{.+}}, [[S_INT_TY]]* {{.+}}, [[ST_TY]]* [[ST_TY_TEMP]])
250 // CHECK: call {{.*}} [[ST_TY_DESTR:@.+]]([[ST_TY]]* [[ST_TY_TEMP]])
251 // CHECK: br i1 {{.+}}, label %{{.+}}, label %[[S_ARR_BODY]]
252 
253 // firstprivate var(var)
254 // CHECK: call {{.*}} [[ST_TY_DEFAULT_CONSTR]]([[ST_TY]]* [[ST_TY_TEMP:%.+]])
255 // CHECK: call {{.*}} [[S_INT_TY_COPY_CONSTR]]([[S_INT_TY]]* [[VAR_PRIV]], [[S_INT_TY]]* {{.*}} [[VAR_REF]], [[ST_TY]]* [[ST_TY_TEMP]])
256 // CHECK: call {{.*}} [[ST_TY_DESTR]]([[ST_TY]]* [[ST_TY_TEMP]])
257 
258 // Synchronization for initialization.
259 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
260 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
261 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]])
262 
263 // CHECK: call void @__kmpc_for_static_init_4(
264 // CHECK: call void @__kmpc_for_static_fini(
265 
266 // ~(firstprivate var), ~(firstprivate s_arr)
267 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]])
268 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]*
269 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
270 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
271 // CHECK: call i32 @__kmpc_cancel_barrier(%{{.+}}* [[SECTIONS_BARRIER_LOC]], i{{[0-9]+}} [[GTID]])
272 // CHECK: ret void
273 #endif
274 
275