1 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ %s -verify -debug-info-kind=limited -emit-llvm -o - -triple powerpc64le-unknown-linux-gnu -std=c++98 | FileCheck %s 2 3 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ %s -verify -debug-info-kind=limited -emit-llvm -o - -triple powerpc64le-unknown-linux-gnu -std=c++98 | FileCheck --check-prefix SIMD-ONLY0 %s 4 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}} 5 // expected-no-diagnostics 6 7 // CHECK: [[RED_SIZE1:@reduction_size[.].+]] = common thread_local global i64 0 8 // CHECK: [[RED_SIZE2:@reduction_size[.].+]] = common thread_local global i64 0 9 10 struct S { 11 float a; 12 S() : a(0.0f) {} 13 ~S() {} 14 }; 15 16 #pragma omp declare reduction(+:S:omp_out.a += omp_in.a) initializer(omp_priv = omp_orig) 17 18 float g; 19 20 int a; 21 #pragma omp threadprivate(a) 22 int main (int argc, char *argv[]) 23 { 24 int i, n; 25 float a[100], b[100], sum, e[argc + 100]; 26 S c[100]; 27 float &d = g; 28 29 /* Some initializations */ 30 n = 100; 31 for (i=0; i < n; i++) 32 a[i] = b[i] = i * 1.0; 33 sum = 0.0; 34 35 #pragma omp taskloop reduction(+:sum, c[:n], d, e) 36 for (i=0; i < n; i++) { 37 sum = sum + (a[i] * b[i]); 38 c[i].a = i*i; 39 d += i*i; 40 e[i] = i; 41 } 42 43 } 44 45 // CHECK-LABEL: @main( 46 // CHECK: [[RETVAL:%.*]] = alloca i32, 47 // CHECK: [[ARGC_ADDR:%.*]] = alloca i32, 48 // CHECK: [[ARGV_ADDR:%.*]] = alloca i8**, 49 // CHECK: [[I:%.*]] = alloca i32, 50 // CHECK: [[N:%.*]] = alloca i32, 51 // CHECK: [[A:%.*]] = alloca [100 x float], 52 // CHECK: [[B:%.*]] = alloca [100 x float], 53 // CHECK: [[SUM:%.*]] = alloca float, 54 // CHECK: [[SAVED_STACK:%.*]] = alloca i8*, 55 // CHECK: [[C:%.*]] = alloca [100 x %struct.S], 56 // CHECK: [[D:%.*]] = alloca float*, 57 // CHECK: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], 58 // CHECK: [[DOTRD_INPUT_:%.*]] = alloca [4 x %struct.kmp_taskred_input_t], 59 // CHECK: alloca i32, 60 // CHECK: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, 61 // CHECK: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, 62 // CHECK: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* 63 // CHECK: store i32 0, i32* [[RETVAL]], 64 // CHECK: store i32 [[ARGC:%.*]], i32* [[ARGC_ADDR]], 65 // CHECK: store i8** [[ARGV:%.*]], i8*** [[ARGV_ADDR]], 66 // CHECK: [[TMP1:%.*]] = load i32, i32* [[ARGC_ADDR]], 67 // CHECK: [[ADD:%.*]] = add nsw i32 [[TMP1]], 100 68 // CHECK: [[TMP2:%.*]] = zext i32 [[ADD]] to i64 69 // CHECK: [[VLA:%.+]] = alloca float, i64 % 70 71 // CHECK: call void @__kmpc_taskgroup(%struct.ident_t* 72 // CHECK-DAG: [[TMP21:%.*]] = bitcast float* [[SUM]] to i8* 73 // CHECK-DAG: store i8* [[TMP21]], i8** [[TMP20:%[^,]+]], 74 // CHECK-DAG: [[TMP20]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_:%.+]], i32 0, i32 0 75 // CHECK-DAG: [[TMP21:%.*]] = bitcast float* [[SUM]] to i8* 76 // CHECK-DAG: store i8* [[TMP21]], i8** [[TMP20:%[^,]+]], 77 // CHECK-DAG: [[TMP20]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1 78 // CHECK-DAG: [[TMP22:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2 79 // CHECK-DAG: store i64 4, i64* [[TMP22]], 80 // CHECK-DAG: [[TMP23:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3 81 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_INIT1:.+]] to i8*), i8** [[TMP23]], 82 // CHECK-DAG: [[TMP24:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4 83 // CHECK-DAG: store i8* null, i8** [[TMP24]], 84 // CHECK-DAG: [[TMP25:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5 85 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_COMB1:.+]] to i8*), i8** [[TMP25]], 86 // CHECK-DAG: [[TMP26:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6 87 // CHECK-DAG: [[TMP27:%.*]] = bitcast i32* [[TMP26]] to i8* 88 // CHECK-DAG: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP27]], i8 0, i64 4, i1 false) 89 // CHECK-DAG: [[ARRAYIDX5:%.*]] = getelementptr inbounds [100 x %struct.S], [100 x %struct.S]* [[C]], i64 0, i64 0 90 // CHECK-DAG: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, % 91 // CHECK-DAG: [[ARRAYIDX6:%.*]] = getelementptr inbounds [100 x %struct.S], [100 x %struct.S]* [[C]], i64 0, i64 [[LB_ADD_LEN]] 92 // CHECK-DAG: [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8* 93 // CHECK-DAG: store i8* [[TMP31]], i8** [[TMP28:%[^,]+]], 94 // CHECK-DAG: [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 0 95 // CHECK-DAG: [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8* 96 // CHECK-DAG: store i8* [[TMP31]], i8** [[TMP28:%[^,]+]], 97 // CHECK-DAG: [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 1 98 // CHECK-DAG: [[TMP32:%.*]] = ptrtoint %struct.S* [[ARRAYIDX6]] to i64 99 // CHECK-DAG: [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX5]] to i64 100 // CHECK-DAG: [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]] 101 // CHECK-DAG: [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64) 102 // CHECK-DAG: [[TMP36:%.*]] = add nuw i64 [[TMP35]], 1 103 // CHECK-DAG: [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64) 104 // CHECK-DAG: store i64 [[TMP37]], i64* [[TMP38:%[^,]+]], 105 // CHECK-DAG: [[TMP38]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 2 106 // CHECK-DAG: [[TMP39:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 3 107 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_INIT2:.+]] to i8*), i8** [[TMP39]], 108 // CHECK-DAG: [[TMP40:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 4 109 // CHECK-DAG: store i8* bitcast (void (i8*)* @[[RED_FINI2:.+]] to i8*), i8** [[TMP40]], 110 // CHECK-DAG: [[TMP41:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 5 111 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_COMB2:.+]] to i8*), i8** [[TMP41]], 112 // CHECK-DAG: [[TMP42:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 6 113 // CHECK-DAG: store i32 1, i32* [[TMP42]], 114 // CHECK-DAG: [[TMP44:%.*]] = load float*, float** [[D]], 115 // CHECK-DAG: [[TMP45:%.*]] = bitcast float* [[TMP44]] to i8* 116 // CHECK-DAG: store i8* [[TMP45]], i8** [[TMP43:%[^,]+]], 117 // CHECK-DAG: [[TMP43]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7:%.+]], i32 0, i32 0 118 // CHECK-DAG: [[TMP45:%.*]] = bitcast float* [[TMP44]] to i8* 119 // CHECK-DAG: store i8* [[TMP45]], i8** [[TMP43:%[^,]+]], 120 // CHECK-DAG: [[TMP43]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 1 121 // CHECK-DAG: [[TMP46:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 2 122 // CHECK-DAG: store i64 4, i64* [[TMP46]], 123 // CHECK-DAG: [[TMP47:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 3 124 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_INIT3:.+]] to i8*), i8** [[TMP47]], 125 // CHECK-DAG: [[TMP48:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 4 126 // CHECK-DAG: store i8* null, i8** [[TMP48]], 127 // CHECK-DAG: [[TMP49:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 5 128 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_COMB3:.+]] to i8*), i8** [[TMP49]], 129 // CHECK-DAG: [[TMP50:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 6 130 // CHECK-DAG: [[TMP51:%.*]] = bitcast i32* [[TMP50]] to i8* 131 // CHECK-DAG: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP51]], i8 0, i64 4, i1 false) 132 // CHECK-DAG: [[TMP53:%.*]] = bitcast float* [[VLA]] to i8* 133 // CHECK-DAG: store i8* [[TMP53]], i8** [[TMP52:%[^,]+]], 134 // CHECK-DAG: [[TMP52]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8:%.+]], i32 0, i32 0 135 // CHECK-DAG: [[TMP53:%.*]] = bitcast float* [[VLA]] to i8* 136 // CHECK-DAG: store i8* [[TMP53]], i8** [[TMP52:%[^,]+]], 137 // CHECK-DAG: [[TMP52]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 1 138 // CHECK-DAG: [[TMP54:%.*]] = mul nuw i64 [[TMP2]], 4 139 // CHECK-DAG: [[TMP55:%.*]] = udiv exact i64 [[TMP54]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64) 140 // CHECK-DAG: store i64 [[TMP54]], i64* [[TMP56:%[^,]+]], 141 // CHECK-DAG: [[TMP56]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 2 142 // CHECK-DAG: [[TMP57:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 3 143 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_INIT4:.+]] to i8*), i8** [[TMP57]], 144 // CHECK-DAG: [[TMP58:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 4 145 // CHECK-DAG: store i8* null, i8** [[TMP58]], 146 // CHECK-DAG: [[TMP59:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 5 147 // CHECK-DAG: store i8* bitcast (void (i8*, i8*)* @[[RED_COMB4:.+]] to i8*), i8** [[TMP59]], 148 // CHECK-DAG: [[TMP60:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 6 149 // CHECK-DAG: store i32 1, i32* [[TMP60]], 150 // 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 151 // 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 152 // 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 153 // 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 154 // CHECK: [[TMP61:%.*]] = bitcast [4 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8* 155 // CHECK: [[TMP62:%.*]] = call i8* @__kmpc_taskred_init(i32 [[TMP0]], i32 4, i8* [[TMP61]]) 156 // CHECK: [[TMP63:%.*]] = load i32, i32* [[N]], 157 // CHECK: store i32 [[TMP63]], i32* [[DOTCAPTURE_EXPR_]], 158 // CHECK: [[TMP64:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], 159 // CHECK: [[SUB:%.*]] = sub nsw i32 [[TMP64]], 0 160 // CHECK: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 161 // CHECK: [[SUB12:%.*]] = sub nsw i32 [[DIV]], 1 162 // CHECK: store i32 [[SUB12]], i32* [[DOTCAPTURE_EXPR_9]], 163 // 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*)*)) 164 // 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) 165 // CHECK: call void @__kmpc_end_taskgroup(%struct.ident_t* 166 167 // CHECK: ret i32 168 169 // CHECK: define internal void @[[RED_INIT1]](i8* noalias noundef %{{.+}}, i8* noalias noundef %{{.+}}) 170 // CHECK: store float 0.000000e+00, float* % 171 // CHECK: ret void 172 173 // CHECK: define internal void @[[RED_COMB1]](i8* noundef %0, i8* noundef %1) 174 // CHECK: fadd float % 175 // CHECK: store float %{{.+}}, float* % 176 // CHECK: ret void 177 178 // CHECK: define internal void @[[RED_INIT2]](i8* noalias noundef %{{.+}}, i8* noalias noundef %{{.+}}) 179 // CHECK: call void [[OMP_INIT1:@.+]](%struct.S* 180 // CHECK: ret void 181 182 // CHECK: define internal void [[OMP_COMB1:@.+]](%struct.S* noalias noundef %0, %struct.S* noalias noundef %1) 183 // CHECK: fadd float % 184 185 // CHECK: define internal void [[OMP_INIT1]](%struct.S* noalias noundef %0, %struct.S* noalias noundef %1) 186 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64( 187 188 // CHECK: define internal void @[[RED_FINI2]](i8* noundef %0) 189 // CHECK: load i64, i64* [[RED_SIZE1]] 190 // CHECK: call void @ 191 // CHECK: ret void 192 193 // CHECK: define internal void @[[RED_COMB2]](i8* noundef %0, i8* noundef %1) 194 // CHECK: load i64, i64* [[RED_SIZE1]] 195 // CHECK: call void [[OMP_COMB1]]( 196 // CHECK: ret void 197 198 // CHECK: define internal void @[[RED_INIT3]](i8* noalias noundef %{{.+}}, i8* noalias noundef %{{.+}}) 199 // CHECK: store float 0.000000e+00, float* % 200 // CHECK: ret void 201 202 // CHECK: define internal void @[[RED_COMB3]](i8* noundef %0, i8* noundef %1) 203 // CHECK: fadd float % 204 // CHECK: store float %{{.+}}, float* % 205 // CHECK: ret void 206 207 // CHECK: define internal void @[[RED_INIT4]](i8* noalias noundef %{{.+}}, i8* noalias noundef %{{.+}}) 208 // CHECK: load i64, i64* [[RED_SIZE2]] 209 // CHECK: store float 0.000000e+00, float* % 210 // CHECK: ret void 211 212 // CHECK: define internal void @[[RED_COMB4]](i8* noundef %0, i8* noundef %1) 213 // CHECK: load i64, i64* [[RED_SIZE2]] 214 // CHECK: fadd float % 215 // CHECK: store float %{{.+}}, float* % 216 // CHECK: ret void 217 218 // CHECK: call i8* @__kmpc_task_reduction_get_th_data( 219 // CHECK: call i8* @__kmpc_task_reduction_get_th_data( 220 // CHECK: call i8* @__kmpc_task_reduction_get_th_data( 221 // CHECK: call i8* @__kmpc_task_reduction_get_th_data( 222 223 // CHECK-DAG: distinct !DISubprogram(linkageName: "[[TASK]]", scope: ! 224 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_INIT1]]" 225 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_COMB1]]" 226 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_INIT2]]" 227 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_FINI2]]" 228 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_COMB2]]" 229 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_INIT3]]" 230 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_COMB3]]" 231 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_INIT4]]" 232 // CHECK-DAG: !DISubprogram(linkageName: "[[RED_COMB4]]" 233