1 // Test device global memory data sharing codegen. 2 ///==========================================================================/// 3 4 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm-bc %s -o %t-ppc-host.bc 5 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple nvptx64-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - -disable-llvm-optzns | FileCheck %s --check-prefix CK1 --check-prefix SEQ 6 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple nvptx64-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - -disable-llvm-optzns -fopenmp-cuda-parallel-target-regions | FileCheck %s --check-prefix CK1 --check-prefix PAR 7 8 // expected-no-diagnostics 9 10 #ifndef HEADER 11 #define HEADER 12 13 void test_ds(){ 14 #pragma omp target 15 { 16 int a = 10; 17 #pragma omp parallel 18 { 19 a = 1000; 20 } 21 int b = 100; 22 int c = 1000; 23 #pragma omp parallel private(c) 24 { 25 int *c1 = &c; 26 b = a + 10000; 27 } 28 } 29 } 30 // SEQ: [[MEM_TY:%.+]] = type { [128 x i8] } 31 // SEQ-DAG: [[SHARED_GLOBAL_RD:@.+]] = weak addrspace(3) global [[MEM_TY]] undef 32 // SEQ-DAG: [[KERNEL_PTR:@.+]] = internal addrspace(3) global i8* undef 33 // SEQ-DAG: [[KERNEL_SIZE:@.+]] = internal unnamed_addr constant i64 8 34 // SEQ-DAG: [[KERNEL_SHARED:@.+]] = internal unnamed_addr constant i16 1 35 36 /// ========= In the worker function ========= /// 37 // CK1: {{.*}}define internal void @__omp_offloading{{.*}}test_ds{{.*}}_worker() 38 // CK1: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* null, i32 0) 39 // CK1-NOT: call void @__kmpc_data_sharing_init_stack 40 41 /// ========= In the kernel function ========= /// 42 43 // CK1: {{.*}}define weak void @__omp_offloading{{.*}}test_ds{{.*}}() 44 // CK1: [[SHAREDARGS1:%.+]] = alloca i8** 45 // CK1: [[SHAREDARGS2:%.+]] = alloca i8** 46 // CK1: call void @__kmpc_kernel_init 47 // CK1: call void @__kmpc_data_sharing_init_stack 48 // SEQ: [[SHARED_MEM_FLAG:%.+]] = load i16, i16* [[KERNEL_SHARED]], 49 // SEQ: [[SIZE:%.+]] = load i64, i64* [[KERNEL_SIZE]], 50 // SEQ: call void @__kmpc_get_team_static_memory(i16 0, i8* addrspacecast (i8 addrspace(3)* getelementptr inbounds ([[MEM_TY]], [[MEM_TY]] addrspace(3)* [[SHARED_GLOBAL_RD]], i32 0, i32 0, i32 0) to i8*), i64 [[SIZE]], i16 [[SHARED_MEM_FLAG]], i8** addrspacecast (i8* addrspace(3)* [[KERNEL_PTR]] to i8**)) 51 // SEQ: [[KERNEL_RD:%.+]] = load i8*, i8* addrspace(3)* [[KERNEL_PTR]], 52 // SEQ: [[GLOBALSTACK:%.+]] = getelementptr inbounds i8, i8* [[KERNEL_RD]], i64 0 53 // PAR: [[GLOBALSTACK:%.+]] = call i8* @__kmpc_data_sharing_push_stack(i{{32|64}} 8, i16 1) 54 // CK1: [[GLOBALSTACK2:%.+]] = bitcast i8* [[GLOBALSTACK]] to %struct._globalized_locals_ty* 55 // CK1: [[A:%.+]] = getelementptr inbounds %struct._globalized_locals_ty, %struct._globalized_locals_ty* [[GLOBALSTACK2]], i32 0, i32 0 56 // CK1: [[B:%.+]] = getelementptr inbounds %struct._globalized_locals_ty, %struct._globalized_locals_ty* [[GLOBALSTACK2]], i32 0, i32 1 57 // CK1: store i32 10, i32* [[A]] 58 // CK1: call void @__kmpc_kernel_prepare_parallel({{.*}}) 59 // CK1: call void @__kmpc_begin_sharing_variables(i8*** [[SHAREDARGS1]], i64 1) 60 // CK1: [[SHARGSTMP1:%.+]] = load i8**, i8*** [[SHAREDARGS1]] 61 // CK1: [[SHARGSTMP2:%.+]] = getelementptr inbounds i8*, i8** [[SHARGSTMP1]], i64 0 62 // CK1: [[SHAREDVAR:%.+]] = bitcast i32* [[A]] to i8* 63 // CK1: store i8* [[SHAREDVAR]], i8** [[SHARGSTMP2]] 64 // CK1: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* null, i32 0) 65 // CK1: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* null, i32 0) 66 // CK1: call void @__kmpc_end_sharing_variables() 67 // CK1: store i32 100, i32* [[B]] 68 // CK1: call void @__kmpc_kernel_prepare_parallel({{.*}}) 69 // CK1: call void @__kmpc_begin_sharing_variables(i8*** [[SHAREDARGS2]], i64 2) 70 // CK1: [[SHARGSTMP3:%.+]] = load i8**, i8*** [[SHAREDARGS2]] 71 // CK1: [[SHARGSTMP4:%.+]] = getelementptr inbounds i8*, i8** [[SHARGSTMP3]], i64 0 72 // CK1: [[SHAREDVAR1:%.+]] = bitcast i32* [[B]] to i8* 73 // CK1: store i8* [[SHAREDVAR1]], i8** [[SHARGSTMP4]] 74 // CK1: [[SHARGSTMP12:%.+]] = getelementptr inbounds i8*, i8** [[SHARGSTMP3]], i64 1 75 // CK1: [[SHAREDVAR2:%.+]] = bitcast i32* [[A]] to i8* 76 // CK1: store i8* [[SHAREDVAR2]], i8** [[SHARGSTMP12]] 77 // CK1: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* null, i32 0) 78 // CK1: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* null, i32 0) 79 // CK1: call void @__kmpc_end_sharing_variables() 80 // SEQ: [[SHARED_MEM_FLAG:%.+]] = load i16, i16* [[KERNEL_SHARED]], 81 // SEQ: call void @__kmpc_restore_team_static_memory(i16 0, i16 [[SHARED_MEM_FLAG]]) 82 // PAR: call void @__kmpc_data_sharing_pop_stack(i8* [[GLOBALSTACK]]) 83 // CK1: call void @__kmpc_kernel_deinit(i16 1) 84 85 /// ========= In the data sharing wrapper function ========= /// 86 87 // CK1: {{.*}}define internal void @__omp_outlined{{.*}}wrapper({{.*}}) 88 // CK1: [[SHAREDARGS4:%.+]] = alloca i8** 89 // CK1: call void @__kmpc_get_shared_variables(i8*** [[SHAREDARGS4]]) 90 // CK1: [[SHARGSTMP13:%.+]] = load i8**, i8*** [[SHAREDARGS4]] 91 // CK1: [[SHARGSTMP14:%.+]] = getelementptr inbounds i8*, i8** [[SHARGSTMP13]], i64 0 92 // CK1: [[SHARGSTMP15:%.+]] = bitcast i8** [[SHARGSTMP14]] to i32** 93 // CK1: [[SHARGSTMP16:%.+]] = load i32*, i32** [[SHARGSTMP15]] 94 // CK1: call void @__omp_outlined__{{.*}}({{.*}}, i32* [[SHARGSTMP16]]) 95 96 /// outlined function for the second parallel region /// 97 98 // CK1: define internal void @{{.+}}(i32* noalias %{{.+}}, i32* noalias %{{.+}}, i32* nonnull align {{[0-9]+}} dereferenceable{{.+}}, i32* nonnull align {{[0-9]+}} dereferenceable{{.+}}) 99 // CK1-NOT: call i8* @__kmpc_data_sharing_push_stack( 100 // CK1: [[C_ADDR:%.+]] = alloca i32, 101 // CK1: store i32* [[C_ADDR]], i32** % 102 // CK1i-NOT: call void @__kmpc_data_sharing_pop_stack( 103 104 /// ========= In the data sharing wrapper function ========= /// 105 106 // CK1: {{.*}}define internal void @__omp_outlined{{.*}}wrapper({{.*}}) 107 // CK1: [[SHAREDARGS3:%.+]] = alloca i8** 108 // CK1: call void @__kmpc_get_shared_variables(i8*** [[SHAREDARGS3]]) 109 // CK1: [[SHARGSTMP5:%.+]] = load i8**, i8*** [[SHAREDARGS3]] 110 // CK1: [[SHARGSTMP6:%.+]] = getelementptr inbounds i8*, i8** [[SHARGSTMP5]], i64 0 111 // CK1: [[SHARGSTMP7:%.+]] = bitcast i8** [[SHARGSTMP6]] to i32** 112 // CK1: [[SHARGSTMP8:%.+]] = load i32*, i32** [[SHARGSTMP7]] 113 // CK1: [[SHARGSTMP9:%.+]] = getelementptr inbounds i8*, i8** [[SHARGSTMP5]], i64 1 114 // CK1: [[SHARGSTMP10:%.+]] = bitcast i8** [[SHARGSTMP9]] to i32** 115 // CK1: [[SHARGSTMP11:%.+]] = load i32*, i32** [[SHARGSTMP10]] 116 // CK1: call void @__omp_outlined__{{.*}}({{.*}}, i32* [[SHARGSTMP8]], i32* [[SHARGSTMP11]]) 117 118 #endif 119 120