1 // RUN: %clang_cc1 -fopenmp -x c++ %s -verify -debug-info-kind=limited -emit-llvm -o - -triple powerpc64le-unknown-linux-gnu -std=c++98 -fnoopenmp-use-tls | FileCheck %s
2
3 // RUN: %clang_cc1 -fopenmp-simd -x c++ %s -verify -debug-info-kind=limited -emit-llvm -o - -triple powerpc64le-unknown-linux-gnu -std=c++98 -fnoopenmp-use-tls | FileCheck --check-prefix SIMD-ONLY0 %s
4 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
5 // expected-no-diagnostics
6
7 struct S {
8 float a;
SS9 S() : a(0.0f) {}
~SS10 ~S() {}
11 };
12
13 #pragma omp declare reduction(+:S:omp_out.a += omp_in.a) initializer(omp_priv = omp_orig)
14
15 float g;
16
17 int a;
18 #pragma omp threadprivate(a)
main(int argc,char * argv[])19 int main (int argc, char *argv[])
20 {
21 int i, n;
22 float a[100], b[100], sum, e[argc + 100];
23 S c[100];
24 float &d = g;
25
26 /* Some initializations */
27 n = 100;
28 for (i=0; i < n; i++)
29 a[i] = b[i] = i * 1.0;
30 sum = 0.0;
31
32 #pragma omp master taskloop simd reduction(+:sum, c[:n], d, e)
33 for (i=0; i < n; i++) {
34 sum = sum + (a[i] * b[i]);
35 c[i].a = i*i;
36 d += i*i;
37 e[i] = i;
38 }
39
40 }
41
42 // CHECK-LABEL: @main(
43 // CHECK: [[RETVAL:%.*]] = alloca i32,
44 // CHECK: [[ARGC_ADDR:%.*]] = alloca i32,
45 // CHECK: [[ARGV_ADDR:%.*]] = alloca i8**,
46 // CHECK: [[I:%.*]] = alloca i32,
47 // CHECK: [[N:%.*]] = alloca i32,
48 // CHECK: [[A:%.*]] = alloca [100 x float],
49 // CHECK: [[B:%.*]] = alloca [100 x float],
50 // CHECK: [[SUM:%.*]] = alloca float,
51 // CHECK: [[SAVED_STACK:%.*]] = alloca i8*,
52 // CHECK: [[C:%.*]] = alloca [100 x %struct.S],
53 // CHECK: [[D:%.*]] = alloca float*,
54 // CHECK: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]],
55 // CHECK: [[DOTRD_INPUT_:%.*]] = alloca [4 x %struct.kmp_task_red_input_t],
56 // CHECK: alloca i32,
57 // CHECK: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32,
58 // CHECK: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32,
59 // CHECK: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t*
60 // CHECK: store i32 0, i32* [[RETVAL]],
61 // CHECK: store i32 [[ARGC:%.*]], i32* [[ARGC_ADDR]],
62 // CHECK: store i8** [[ARGV:%.*]], i8*** [[ARGV_ADDR]],
63 // CHECK: [[TMP1:%.*]] = load i32, i32* [[ARGC_ADDR]],
64 // CHECK: [[ADD:%.*]] = add nsw i32 [[TMP1]], 100
65 // CHECK: [[TMP2:%.*]] = zext i32 [[ADD]] to i64
66 // CHECK: [[VLA:%.+]] = alloca float, i64 %
67
68 // CHECK: call void @__kmpc_taskgroup(%struct.ident_t*
69 // CHECK-DAG: [[TMP21:%.*]] = bitcast float* [[SUM]] to i8*
70 // CHECK-DAG: store i8* [[TMP21]], i8** [[TMP20:%[^,]+]],
71 // CHECK-DAG: [[TMP20]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T:%.+]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_:%.+]], i32 0, i32 0
72 // CHECK-DAG: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
73 // CHECK-DAG: store i64 4, i64* [[TMP22]],
74 // CHECK-DAG: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
75 // CHECK-DAG: store i8* bitcast (void (i8*)* @[[RED_INIT1:.+]] to i8*), i8** [[TMP23]],
76 // CHECK-DAG: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
77 // CHECK-DAG: store i8* null, i8** [[TMP24]],
78 // CHECK-DAG: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
79 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_COMB1:.+]] to i8*), i8** [[TMP25]],
80 // CHECK-DAG: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
81 // CHECK-DAG: [[TMP27:%.*]] = bitcast i32* [[TMP26]] to i8*
82 // CHECK-DAG: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP27]], i8 0, i64 4, i1 false)
83 // CHECK-DAG: [[ARRAYIDX5:%.*]] = getelementptr inbounds [100 x %struct.S], [100 x %struct.S]* [[C]], i64 0, i64 0
84 // CHECK-DAG: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, %
85 // CHECK-DAG: [[ARRAYIDX6:%.*]] = getelementptr inbounds [100 x %struct.S], [100 x %struct.S]* [[C]], i64 0, i64 [[LB_ADD_LEN]]
86 // CHECK-DAG: [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8*
87 // CHECK-DAG: store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
88 // CHECK-DAG: [[TMP28]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 0
89 // CHECK-DAG: [[TMP32:%.*]] = ptrtoint %struct.S* [[ARRAYIDX6]] to i64
90 // CHECK-DAG: [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX5]] to i64
91 // CHECK-DAG: [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
92 // CHECK-DAG: [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
93 // CHECK-DAG: [[TMP36:%.*]] = add nuw i64 [[TMP35]], 1
94 // CHECK-DAG: [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
95 // CHECK-DAG: store i64 [[TMP37]], i64* [[TMP38:%[^,]+]],
96 // CHECK-DAG: [[TMP38]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 1
97 // CHECK-DAG: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 2
98 // CHECK-DAG: store i8* bitcast (void (i8*)* @[[RED_INIT2:.+]] to i8*), i8** [[TMP39]],
99 // CHECK-DAG: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 3
100 // CHECK-DAG: store i8* bitcast (void (i8*)* @[[RED_FINI2:.+]] to i8*), i8** [[TMP40]],
101 // CHECK-DAG: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 4
102 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_COMB2:.+]] to i8*), i8** [[TMP41]],
103 // CHECK-DAG: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 5
104 // CHECK-DAG: store i32 1, i32* [[TMP42]],
105 // CHECK-DAG: [[TMP44:%.*]] = load float*, float** [[D]],
106 // CHECK-DAG: [[TMP45:%.*]] = bitcast float* [[TMP44]] to i8*
107 // CHECK-DAG: store i8* [[TMP45]], i8** [[TMP43:%[^,]+]],
108 // CHECK-DAG: [[TMP43]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7:%.+]], i32 0, i32 0
109 // CHECK-DAG: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 1
110 // CHECK-DAG: store i64 4, i64* [[TMP46]],
111 // CHECK-DAG: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 2
112 // CHECK-DAG: store i8* bitcast (void (i8*)* @[[RED_INIT3:.+]] to i8*), i8** [[TMP47]],
113 // CHECK-DAG: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 3
114 // CHECK-DAG: store i8* null, i8** [[TMP48]],
115 // CHECK-DAG: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 4
116 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_COMB3:.+]] to i8*), i8** [[TMP49]],
117 // CHECK-DAG: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 5
118 // CHECK-DAG: [[TMP51:%.*]] = bitcast i32* [[TMP50]] to i8*
119 // CHECK-DAG: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP51]], i8 0, i64 4, i1 false)
120 // CHECK-DAG: [[TMP53:%.*]] = bitcast float* [[VLA]] to i8*
121 // CHECK-DAG: store i8* [[TMP53]], i8** [[TMP52:%[^,]+]],
122 // CHECK-DAG: [[TMP52]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8:%.+]], i32 0, i32 0
123 // CHECK-DAG: [[TMP54:%.*]] = mul nuw i64 [[TMP2]], 4
124 // CHECK-DAG: [[TMP55:%.*]] = udiv exact i64 [[TMP54]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
125 // CHECK-DAG: store i64 [[TMP54]], i64* [[TMP56:%[^,]+]],
126 // CHECK-DAG: [[TMP56]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 1
127 // CHECK-DAG: [[TMP57:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 2
128 // CHECK-DAG: store i8* bitcast (void (i8*)* @[[RED_INIT4:.+]] to i8*), i8** [[TMP57]],
129 // CHECK-DAG: [[TMP58:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 3
130 // CHECK-DAG: store i8* null, i8** [[TMP58]],
131 // CHECK-DAG: [[TMP59:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 4
132 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_COMB4:.+]] to i8*), i8** [[TMP59]],
133 // CHECK-DAG: [[TMP60:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 5
134 // CHECK-DAG: store i32 1, i32* [[TMP60]],
135 // CHECK-DAG: [[DOTRD_INPUT_GEP_]] = getelementptr inbounds [4 x %struct.kmp_task_red_input_t], [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]], i64 0, i64
136 // CHECK-DAG: [[DOTRD_INPUT_GEP_4]] = getelementptr inbounds [4 x %struct.kmp_task_red_input_t], [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]], i64 0, i64
137 // CHECK-DAG: [[DOTRD_INPUT_GEP_7]] = getelementptr inbounds [4 x %struct.kmp_task_red_input_t], [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]], i64 0, i64
138 // CHECK-DAG: [[DOTRD_INPUT_GEP_8]] = getelementptr inbounds [4 x %struct.kmp_task_red_input_t], [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]], i64 0, i64
139 // CHECK: [[TMP61:%.*]] = bitcast [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]] to i8*
140 // CHECK: [[TMP62:%.*]] = call i8* @__kmpc_task_reduction_init(i32 [[TMP0]], i32 4, i8* [[TMP61]])
141 // CHECK: [[TMP63:%.*]] = load i32, i32* [[N]],
142 // CHECK: store i32 [[TMP63]], i32* [[DOTCAPTURE_EXPR_]],
143 // CHECK: [[TMP64:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]],
144 // CHECK: [[SUB:%.*]] = sub nsw i32 [[TMP64]], 0
145 // CHECK: [[SUB10:%.*]] = sub nsw i32 [[SUB]], 1
146 // CHECK: [[ADD11:%.*]] = add nsw i32 [[SUB10]], 1
147 // CHECK: [[DIV:%.*]] = sdiv i32 [[ADD11]], 1
148 // CHECK: [[SUB12:%.*]] = sub nsw i32 [[DIV]], 1
149 // CHECK: store i32 [[SUB12]], i32* [[DOTCAPTURE_EXPR_9]],
150 // CHECK: [[TMP65:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* %{{.+}}, i32 [[TMP0]], i32 1, i64 888, i64 72, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @{{.+}} to i32 (i32, i8*)*))
151 // CHECK: call void @__kmpc_taskloop(%struct.ident_t* %{{.+}}, i32 [[TMP0]], i8* [[TMP65]], i32 1, i64* %{{.+}}, i64* %{{.+}}, i64 %{{.+}}, i32 1, i32 0, i64 0, i8* null)
152 // CHECK: call void @__kmpc_end_taskgroup(%struct.ident_t*
153
154 // CHECK: ret i32
155
156 // CHECK: define internal void @[[RED_INIT1]](i8* %0)
157 // CHECK: store float 0.000000e+00, float* %
158 // CHECK: ret void
159
160 // CHECK: define internal void @[[RED_COMB1]](i8* %0, i8* %1)
161 // CHECK: fadd float %
162 // CHECK: store float %{{.+}}, float* %
163 // CHECK: ret void
164
165 // CHECK: define internal void @[[RED_INIT2]](i8* %0)
166 // CHECK: call i8* @__kmpc_threadprivate_cached(
167 // CHECK: call i8* @__kmpc_threadprivate_cached(
168 // CHECK: call void [[OMP_INIT1:@.+]](
169 // CHECK: ret void
170
171 // CHECK: define internal void [[OMP_COMB1:@.+]](%struct.S* noalias %0, %struct.S* noalias %1)
172 // CHECK: fadd float %
173
174 // CHECK: define internal void [[OMP_INIT1]](%struct.S* noalias %0, %struct.S* noalias %1)
175 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(
176
177 // CHECK: define internal void @[[RED_FINI2]](i8* %0)
178 // CHECK: call i8* @__kmpc_threadprivate_cached(
179 // CHECK: call void @
180 // CHECK: ret void
181
182 // CHECK: define internal void @[[RED_COMB2]](i8* %0, i8* %1)
183 // CHECK: call i8* @__kmpc_threadprivate_cached(
184 // CHECK: call void [[OMP_COMB1]](
185 // CHECK: ret void
186
187 // CHECK: define internal void @[[RED_INIT3]](i8* %0)
188 // CHECK: store float 0.000000e+00, float* %
189 // CHECK: ret void
190
191 // CHECK: define internal void @[[RED_COMB3]](i8* %0, i8* %1)
192 // CHECK: fadd float %
193 // CHECK: store float %{{.+}}, float* %
194 // CHECK: ret void
195
196 // CHECK: define internal void @[[RED_INIT4]](i8* %0)
197 // CHECK: call i8* @__kmpc_threadprivate_cached(
198 // CHECK: store float 0.000000e+00, float* %
199 // CHECK: ret void
200
201 // CHECK: define internal void @[[RED_COMB4]](i8* %0, i8* %1)
202 // CHECK: call i8* @__kmpc_threadprivate_cached(
203 // CHECK: fadd float %
204 // CHECK: store float %{{.+}}, float* %
205 // CHECK: ret void
206
207 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_INIT1]]"
208 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_COMB1]]"
209 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_INIT2]]"
210 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_FINI2]]"
211 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_COMB2]]"
212 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_INIT3]]"
213 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_COMB3]]"
214 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_INIT4]]"
215 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_COMB4]]"
216