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; 9 S() : a(0.0f) {} 10 ~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) 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 parallel 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: store i32 0, i32* [[RETVAL]], 55 // CHECK: store i32 [[ARGC:%.*]], i32* [[ARGC_ADDR]], 56 // CHECK: store i8** [[ARGV:%.*]], i8*** [[ARGV_ADDR]], 57 // CHECK: [[TMP1:%.*]] = load i32, i32* [[ARGC_ADDR]], 58 // CHECK: [[ADD:%.*]] = add nsw i32 [[TMP1]], 100 59 // CHECK: [[TMP2:%.*]] = zext i32 [[ADD]] to i64 60 // CHECK: [[VLA:%.+]] = alloca float, i64 % 61 62 // CHECK: [[SUM_ADDR:%.*]] = alloca float*, 63 // CHECK: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], 64 // CHECK: [[DOTRD_INPUT_:%.*]] = alloca [4 x %struct.kmp_taskred_input_t], 65 // CHECK: alloca i32, 66 // CHECK: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, 67 // CHECK: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, 68 // CHECK: [[RES:%.+]] = call {{.*}}i32 @__kmpc_master( 69 // CHECK-NEXT: [[IS_MASTER:%.+]] = icmp ne i32 [[RES]], 0 70 // CHECK-NEXT: br i1 [[IS_MASTER]], label {{%?}}[[THEN:.+]], label {{%?}}[[EXIT:[^,]+]] 71 // CHECK: [[THEN]] 72 // CHECK: call void @__kmpc_taskgroup(%struct.ident_t* 73 // CHECK-DAG: [[TMP21:%.*]] = bitcast float* %{{.+}} to i8* 74 // CHECK-DAG: store i8* [[TMP21]], i8** [[TMP20:%[^,]+]], 75 // CHECK-DAG: [[TMP20]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_:%.+]], i32 0, i32 0 76 // CHECK-DAG: [[TMP21:%.*]] = bitcast float* %{{.+}} to i8* 77 // CHECK-DAG: store i8* [[TMP21]], i8** [[TMP20:%[^,]+]], 78 // CHECK-DAG: [[TMP20]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1 79 // CHECK-DAG: [[TMP22:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2 80 // CHECK-DAG: store i64 4, i64* [[TMP22]], 81 // CHECK-DAG: [[TMP23:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3 82 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_INIT1:.+]] to i8*), i8** [[TMP23]], 83 // CHECK-DAG: [[TMP24:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4 84 // CHECK-DAG: store i8* null, i8** [[TMP24]], 85 // CHECK-DAG: [[TMP25:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5 86 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_COMB1:.+]] to i8*), i8** [[TMP25]], 87 // CHECK-DAG: [[TMP26:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6 88 // CHECK-DAG: [[TMP27:%.*]] = bitcast i32* [[TMP26]] to i8* 89 // CHECK-DAG: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP27]], i8 0, i64 4, i1 false) 90 // CHECK-DAG: [[ARRAYIDX5:%.*]] = getelementptr inbounds [100 x %struct.S], [100 x %struct.S]* [[C:%.+]], i64 0, i64 0 91 // CHECK-DAG: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, % 92 // CHECK-DAG: [[ARRAYIDX6:%.*]] = getelementptr inbounds [100 x %struct.S], [100 x %struct.S]* [[C]], i64 0, i64 [[LB_ADD_LEN]] 93 // CHECK-DAG: [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8* 94 // CHECK-DAG: store i8* [[TMP31]], i8** [[TMP28:%[^,]+]], 95 // CHECK-DAG: [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 0 96 // CHECK-DAG: [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8* 97 // CHECK-DAG: store i8* [[TMP31]], i8** [[TMP28:%[^,]+]], 98 // CHECK-DAG: [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 1 99 // CHECK-DAG: [[TMP32:%.*]] = ptrtoint %struct.S* [[ARRAYIDX6]] to i64 100 // CHECK-DAG: [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX5]] to i64 101 // CHECK-DAG: [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]] 102 // CHECK-DAG: [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 103 // CHECK-DAG: [[TMP36:%.*]] = add nuw i64 [[TMP35]], 1 104 // CHECK-DAG: [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 105 // CHECK-DAG: store i64 [[TMP37]], i64* [[TMP38:%[^,]+]], 106 // CHECK-DAG: [[TMP38]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 2 107 // CHECK-DAG: [[TMP39:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 3 108 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_INIT2:.+]] to i8*), i8** [[TMP39]], 109 // CHECK-DAG: [[TMP40:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 4 110 // CHECK-DAG: store i8* bitcast (void (i8*)* @[[RED_FINI2:.+]] to i8*), i8** [[TMP40]], 111 // CHECK-DAG: [[TMP41:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 5 112 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_COMB2:.+]] to i8*), i8** [[TMP41]], 113 // CHECK-DAG: [[TMP42:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 6 114 // CHECK-DAG: store i32 1, i32* [[TMP42]], 115 // CHECK-DAG: [[TMP44:%.*]] = load float*, float** [[D:%.+]], 116 // CHECK-DAG: [[TMP45:%.*]] = bitcast float* [[TMP44]] to i8* 117 // CHECK-DAG: store i8* [[TMP45]], i8** [[TMP43:%[^,]+]], 118 // CHECK-DAG: [[TMP43]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7:%.+]], i32 0, i32 0 119 // CHECK-DAG: [[TMP45:%.*]] = bitcast float* [[TMP44]] to i8* 120 // CHECK-DAG: store i8* [[TMP45]], i8** [[TMP43:%[^,]+]], 121 // CHECK-DAG: [[TMP43]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 1 122 // CHECK-DAG: [[TMP46:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 2 123 // CHECK-DAG: store i64 4, i64* [[TMP46]], 124 // CHECK-DAG: [[TMP47:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 3 125 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_INIT3:.+]] to i8*), i8** [[TMP47]], 126 // CHECK-DAG: [[TMP48:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 4 127 // CHECK-DAG: store i8* null, i8** [[TMP48]], 128 // CHECK-DAG: [[TMP49:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 5 129 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_COMB3:.+]] to i8*), i8** [[TMP49]], 130 // CHECK-DAG: [[TMP50:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 6 131 // CHECK-DAG: [[TMP51:%.*]] = bitcast i32* [[TMP50]] to i8* 132 // CHECK-DAG: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP51]], i8 0, i64 4, i1 false) 133 // CHECK-DAG: [[TMP53:%.*]] = bitcast float* [[VLA:%.+]] to i8* 134 // CHECK-DAG: store i8* [[TMP53]], i8** [[TMP52:%[^,]+]], 135 // CHECK-DAG: [[TMP52]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8:%.+]], i32 0, i32 0 136 // CHECK-DAG: [[TMP53:%.*]] = bitcast float* [[VLA:%.+]] to i8* 137 // CHECK-DAG: store i8* [[TMP53]], i8** [[TMP52:%[^,]+]], 138 // CHECK-DAG: [[TMP52]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 1 139 // CHECK-DAG: [[TMP54:%.*]] = mul nuw i64 [[TMP2:%.+]], 4 140 // CHECK-DAG: [[TMP55:%.*]] = udiv exact i64 [[TMP54]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 141 // CHECK-DAG: store i64 [[TMP54]], i64* [[TMP56:%[^,]+]], 142 // CHECK-DAG: [[TMP56]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 2 143 // CHECK-DAG: [[TMP57:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 3 144 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_INIT4:.+]] to i8*), i8** [[TMP57]], 145 // CHECK-DAG: [[TMP58:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 4 146 // CHECK-DAG: store i8* null, i8** [[TMP58]], 147 // CHECK-DAG: [[TMP59:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 5 148 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_COMB4:.+]] to i8*), i8** [[TMP59]], 149 // CHECK-DAG: [[TMP60:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 6 150 // CHECK-DAG: store i32 1, i32* [[TMP60]], 151 // CHECK-DAG: [[DOTRD_INPUT_GEP_]] = getelementptr inbounds [4 x %struct.kmp_taskred_input_t], [4 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 152 // CHECK-DAG: [[DOTRD_INPUT_GEP_4]] = getelementptr inbounds [4 x %struct.kmp_taskred_input_t], [4 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 153 // CHECK-DAG: [[DOTRD_INPUT_GEP_7]] = getelementptr inbounds [4 x %struct.kmp_taskred_input_t], [4 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 154 // CHECK-DAG: [[DOTRD_INPUT_GEP_8]] = getelementptr inbounds [4 x %struct.kmp_taskred_input_t], [4 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 155 // CHECK: [[TMP61:%.*]] = bitcast [4 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8* 156 // CHECK: [[TMP62:%.*]] = call i8* @__kmpc_taskred_init(i32 [[TMP0:%.+]], i32 4, i8* [[TMP61]]) 157 // CHECK: [[TMP63:%.*]] = load i32, i32* [[N:%.+]], 158 // CHECK: store i32 [[TMP63]], i32* [[DOTCAPTURE_EXPR_]], 159 // CHECK: [[TMP64:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], 160 // CHECK: [[SUB:%.*]] = sub nsw i32 [[TMP64]], 0 161 // CHECK: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 162 // CHECK: [[SUB12:%.*]] = sub nsw i32 [[DIV]], 1 163 // CHECK: store i32 [[SUB12]], i32* [[DOTCAPTURE_EXPR_9]], 164 // CHECK: [[TMP65:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* {{.+}}, i32 [[TMP0]], i32 1, i64 888, i64 40, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @[[TASK:.+]] to i32 (i32, i8*)*)) 165 // 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) 166 // CHECK: call void @__kmpc_end_taskgroup(%struct.ident_t* 167 // CHECK: call {{.*}}void @__kmpc_end_master( 168 // CHECK-NEXT: br label {{%?}}[[EXIT]] 169 // CHECK: [[EXIT]] 170 171 // CHECK: define internal void @[[RED_INIT1]](i8* noalias %{{.+}}, i8* noalias %{{.+}}) 172 // CHECK: store float 0.000000e+00, float* % 173 // CHECK: ret void 174 175 // CHECK: define internal void @[[RED_COMB1]](i8* %0, i8* %1) 176 // CHECK: fadd float % 177 // CHECK: store float %{{.+}}, float* % 178 // CHECK: ret void 179 180 // CHECK: define internal void @[[RED_INIT2]](i8* noalias %{{.+}}, i8* noalias %{{.+}}) 181 // CHECK: call i8* @__kmpc_threadprivate_cached( 182 // CHECK: call void [[OMP_INIT1:@.+]]( 183 // CHECK: ret void 184 185 // CHECK: define internal void [[OMP_COMB1:@.+]](%struct.S* noalias %0, %struct.S* noalias %1) 186 // CHECK: fadd float % 187 188 // CHECK: define internal void [[OMP_INIT1]](%struct.S* noalias %0, %struct.S* noalias %1) 189 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64( 190 191 // CHECK: define internal void @[[RED_FINI2]](i8* %0) 192 // CHECK: call i8* @__kmpc_threadprivate_cached( 193 // CHECK: call void @ 194 // CHECK: ret void 195 196 // CHECK: define internal void @[[RED_COMB2]](i8* %0, i8* %1) 197 // CHECK: call i8* @__kmpc_threadprivate_cached( 198 // CHECK: call void [[OMP_COMB1]]( 199 // CHECK: ret void 200 201 // CHECK: define internal void @[[RED_INIT3]](i8* noalias %{{.+}}, i8* noalias %{{.+}}) 202 // CHECK: store float 0.000000e+00, float* % 203 // CHECK: ret void 204 205 // CHECK: define internal void @[[RED_COMB3]](i8* %0, i8* %1) 206 // CHECK: fadd float % 207 // CHECK: store float %{{.+}}, float* % 208 // CHECK: ret void 209 210 // CHECK: define internal void @[[RED_INIT4]](i8* noalias %{{.+}}, i8* noalias %{{.+}}) 211 // CHECK: call i8* @__kmpc_threadprivate_cached( 212 // CHECK: store float 0.000000e+00, float* % 213 // CHECK: ret void 214 215 // CHECK: define internal void @[[RED_COMB4]](i8* %0, i8* %1) 216 // CHECK: call i8* @__kmpc_threadprivate_cached( 217 // CHECK: fadd float % 218 // CHECK: store float %{{.+}}, float* % 219 // CHECK: ret void 220 221 // CHECK-NOT: call i8* @__kmpc_threadprivate_cached( 222 // CHECK: call i8* @__kmpc_task_reduction_get_th_data( 223 // CHECK: call i8* @__kmpc_threadprivate_cached( 224 // CHECK: call i8* @__kmpc_task_reduction_get_th_data( 225 // CHECK-NOT: call i8* @__kmpc_threadprivate_cached( 226 // CHECK: call i8* @__kmpc_task_reduction_get_th_data( 227 // CHECK: call i8* @__kmpc_threadprivate_cached( 228 // CHECK: call i8* @__kmpc_task_reduction_get_th_data( 229 // CHECK-NOT: call i8* @__kmpc_threadprivate_cached( 230 231 // CHECK-DAG: distinct !DISubprogram(linkageName: "[[TASK]]", scope: ! 232 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_INIT1]]" 233 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_COMB1]]" 234 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_INIT2]]" 235 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_FINI2]]" 236 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_COMB2]]" 237 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_INIT3]]" 238 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_COMB3]]" 239 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_INIT4]]" 240 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_COMB4]]" 241