diff options
Diffstat (limited to 'test/OpenMP/nvptx_target_parallel_reduction_codegen.cpp')
| -rw-r--r-- | test/OpenMP/nvptx_target_parallel_reduction_codegen.cpp | 160 |
1 files changed, 76 insertions, 84 deletions
diff --git a/test/OpenMP/nvptx_target_parallel_reduction_codegen.cpp b/test/OpenMP/nvptx_target_parallel_reduction_codegen.cpp index 02676272b0fed..c23f275bd1a4e 100644 --- a/test/OpenMP/nvptx_target_parallel_reduction_codegen.cpp +++ b/test/OpenMP/nvptx_target_parallel_reduction_codegen.cpp @@ -9,7 +9,7 @@ #define HEADER // Check for the data transfer medium in shared memory to transfer the reduction list to the first warp. -// CHECK-DAG: [[TRANSFER_STORAGE:@.+]] = common addrspace([[SHARED_ADDRSPACE:[0-9]+]]) global [32 x i64] +// CHECK-DAG: [[TRANSFER_STORAGE:@.+]] = common addrspace([[SHARED_ADDRSPACE:[0-9]+]]) global [32 x i32] // Check that the execution mode of all 3 target regions is set to Spmd Mode. // CHECK-DAG: {{@__omp_offloading_.+l27}}_exec_mode = weak constant i8 0 @@ -54,13 +54,13 @@ int bar(int n){ // CHECK: define {{.*}}void {{@__omp_offloading_.+template.+l27}}( // - // CHECK: call void @__kmpc_spmd_kernel_init( + // CHECK: call void @__kmpc_spmd_kernel_init(i32 {{.+}}, i16 1, i16 0) // CHECK: call void @__kmpc_data_sharing_init_stack_spmd // CHECK: br label {{%?}}[[EXECUTE:.+]] // // CHECK: [[EXECUTE]] // CHECK: {{call|invoke}} void [[PFN:@.+]](i32* - // CHECK: call void @__kmpc_spmd_kernel_deinit() + // CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1) // // // define internal void [[PFN]]( @@ -72,19 +72,17 @@ int bar(int n){ // CHECK: [[E_CAST:%.+]] = bitcast double* [[E]] to i8* // CHECK: store i8* [[E_CAST]], i8** [[PTR1]], align // CHECK: [[ARG_RL:%.+]] = bitcast [[RLT]]* [[RL]] to i8* - // CHECK: [[RET:%.+]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait(i32 {{.+}}, i32 1, i{{32|64}} {{4|8}}, i8* [[ARG_RL]], void (i8*, i16, i16, i16)* [[SHUFFLE_REDUCE_FN:@.+]], void (i8*, i32)* [[WARP_COPY_FN:@.+]]) - // CHECK: switch i32 [[RET]], label {{%?}}[[DEFAULTLABEL:.+]] [ - // CHECK: i32 1, label {{%?}}[[REDLABEL:.+]] + // CHECK: [[RET:%.+]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @{{.+}}, i32 {{.+}}, i32 1, i{{32|64}} {{4|8}}, i8* [[ARG_RL]], void (i8*, i16, i16, i16)* [[SHUFFLE_REDUCE_FN:@.+]], void (i8*, i32)* [[WARP_COPY_FN:@.+]]) + // CHECK: [[CMP:%.+]] = icmp eq i32 [[RET]], 1 + // CHECK: br i1 [[CMP]], label - // CHECK: [[REDLABEL]] // CHECK: [[E_INV:%.+]] = load double, double* [[E_IN:%.+]], align // CHECK: [[EV:%.+]] = load double, double* [[E]], align // CHECK: [[ADD:%.+]] = fadd double [[E_INV]], [[EV]] // CHECK: store double [[ADD]], double* [[E_IN]], align // CHECK: call void @__kmpc_nvptx_end_reduce_nowait( - // CHECK: br label %[[DEFAULTLABEL]] + // CHECK: br label // - // CHECK: [[DEFAULTLABEL]] // CHECK: ret // @@ -187,18 +185,24 @@ int bar(int n){ // CHECK-DAG: [[LANEID:%.+]] = and i32 {{.+}}, 31 // CHECK-DAG: [[WARPID:%.+]] = ashr i32 {{.+}}, 5 // CHECK-DAG: [[RED_LIST:%.+]] = bitcast i8* {{.+}} to [[RLT]]* + // CHECK: store i32 0, i32* [[CNT_ADDR:%.+]], + // CHECK: br label + // CHECK: [[CNT:%.+]] = load i32, i32* [[CNT_ADDR]], + // CHECK: [[DONE_COPY:%.+]] = icmp ult i32 [[CNT]], 2 + // CHECK: br i1 [[DONE_COPY]], label + // CHECK: call void @__kmpc_barrier(%struct.ident_t* @ // CHECK: [[IS_WARP_MASTER:%.+]] = icmp eq i32 [[LANEID]], 0 // CHECK: br i1 [[IS_WARP_MASTER]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]] // // [[DO_COPY]] // CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 0 // CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]], - // CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to double* + // CHECK: [[BASE_ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i32* + // CHECK: [[ELT:%.+]] = getelementptr i32, i32* [[BASE_ELT]], i32 [[CNT]] // - // CHECK: [[MEDIUM_ELT64:%.+]] = getelementptr inbounds [32 x i64], [32 x i64] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]] - // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i64 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT64]] to double addrspace([[SHARED_ADDRSPACE]])* - // CHECK: [[ELT_VAL:%.+]] = load double, double* [[ELT]], align - // CHECK: store double [[ELT_VAL]], double addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align + // CHECK: [[MEDIUM_ELT:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]] + // CHECK: [[ELT_VAL:%.+]] = load i32, i32* [[ELT]], + // CHECK: store volatile i32 [[ELT_VAL]], i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], // CHECK: br label {{%?}}[[COPY_CONT:.+]] // // CHECK: [[COPY_ELSE]] @@ -206,29 +210,30 @@ int bar(int n){ // // Barrier after copy to shared memory storage medium. // CHECK: [[COPY_CONT]] - // CHECK: [[WS:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize() - // CHECK: [[ACTIVE_THREADS:%.+]] = mul nsw i32 [[ACTIVE_WARPS:%.+]], [[WS]] - // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]]) + // CHECK: call void @__kmpc_barrier(%struct.ident_t* @ + // CHECK: [[ACTIVE_WARPS:%.+]] = load i32, i32* // // Read into warp 0. // CHECK: [[IS_W0_ACTIVE_THREAD:%.+]] = icmp ult i32 [[TID:%.+]], [[ACTIVE_WARPS]] // CHECK: br i1 [[IS_W0_ACTIVE_THREAD]], label {{%?}}[[DO_READ:.+]], label {{%?}}[[READ_ELSE:.+]] // // CHECK: [[DO_READ]] - // CHECK: [[MEDIUM_ELT64:%.+]] = getelementptr inbounds [32 x i64], [32 x i64] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]] - // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i64 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT64]] to double addrspace([[SHARED_ADDRSPACE]])* + // CHECK: [[MEDIUM_ELT:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]] // CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 0 // CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]], - // CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to double* - // CHECK: [[MEDIUM_ELT_VAL:%.+]] = load double, double addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align - // CHECK: store double [[MEDIUM_ELT_VAL]], double* [[ELT]], align + // CHECK: [[ELT_BASE:%.+]] = bitcast i8* [[ELT_VOID]] to i32* + // CHECK: [[ELT:%.+]] = getelementptr i32, i32* [[ELT_BASE]], i32 [[CNT]] + // CHECK: [[MEDIUM_ELT_VAL:%.+]] = load volatile i32, i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], + // CHECK: store i32 [[MEDIUM_ELT_VAL]], i32* [[ELT]], // CHECK: br label {{%?}}[[READ_CONT:.+]] // // CHECK: [[READ_ELSE]] // CHECK: br label {{%?}}[[READ_CONT]] // // CHECK: [[READ_CONT]] - // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]]) + // CHECK: [[NEXT:%.+]] = add nsw i32 [[CNT]], 1 + // CHECK: store i32 [[NEXT]], i32* [[CNT_ADDR]], + // CHECK: br label // CHECK: ret @@ -242,13 +247,13 @@ int bar(int n){ // CHECK: define {{.*}}void {{@__omp_offloading_.+template.+l32}}( // - // CHECK: call void @__kmpc_spmd_kernel_init( + // CHECK: call void @__kmpc_spmd_kernel_init(i32 {{.+}}, i16 1, i16 0) // CHECK: call void @__kmpc_data_sharing_init_stack_spmd // CHECK: br label {{%?}}[[EXECUTE:.+]] // // CHECK: [[EXECUTE]] // CHECK: {{call|invoke}} void [[PFN1:@.+]](i32* - // CHECK: call void @__kmpc_spmd_kernel_deinit() + // CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1) // // // define internal void [[PFN1]]( @@ -267,11 +272,9 @@ int bar(int n){ // CHECK: [[D_CAST:%.+]] = bitcast float* [[D]] to i8* // CHECK: store i8* [[D_CAST]], i8** [[PTR2]], align // CHECK: [[ARG_RL:%.+]] = bitcast [[RLT]]* [[RL]] to i8* - // CHECK: [[RET:%.+]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait(i32 {{.+}}, i32 2, i{{32|64}} {{8|16}}, i8* [[ARG_RL]], void (i8*, i16, i16, i16)* [[SHUFFLE_REDUCE_FN:@.+]], void (i8*, i32)* [[WARP_COPY_FN:@.+]]) - // CHECK: switch i32 [[RET]], label {{%?}}[[DEFAULTLABEL:.+]] [ - // CHECK: i32 1, label {{%?}}[[REDLABEL:.+]] - - // CHECK: [[REDLABEL]] + // CHECK: [[RET:%.+]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @{{.+}}, i32 {{.+}}, i32 2, i{{32|64}} {{8|16}}, i8* [[ARG_RL]], void (i8*, i16, i16, i16)* [[SHUFFLE_REDUCE_FN:@.+]], void (i8*, i32)* [[WARP_COPY_FN:@.+]]) + // CHECK: [[CMP:%.+]] = icmp eq i32 [[RET]], 1 + // CHECK: br i1 [[CMP]], label // CHECK: [[C_INV8:%.+]] = load i8, i8* [[C_IN:%.+]], align // CHECK: [[C_INV:%.+]] = sext i8 [[C_INV8]] to i32 // CHECK: [[CV8:%.+]] = load i8, i8* [[C]], align @@ -284,9 +287,8 @@ int bar(int n){ // CHECK: [[MUL:%.+]] = fmul float [[D_INV]], [[DV]] // CHECK: store float [[MUL]], float* [[D_IN]], align // CHECK: call void @__kmpc_nvptx_end_reduce_nowait( - // CHECK: br label %[[DEFAULTLABEL]] + // CHECK: br label // - // CHECK: [[DEFAULTLABEL]] // CHECK: ret // @@ -425,6 +427,7 @@ int bar(int n){ // CHECK-DAG: [[LANEID:%.+]] = and i32 {{.+}}, 31 // CHECK-DAG: [[WARPID:%.+]] = ashr i32 {{.+}}, 5 // CHECK-DAG: [[RED_LIST:%.+]] = bitcast i8* {{.+}} to [[RLT]]* + // CHECK: call void @__kmpc_barrier(%struct.ident_t* @ // CHECK: [[IS_WARP_MASTER:%.+]] = icmp eq i32 [[LANEID]], 0 // CHECK: br i1 [[IS_WARP_MASTER]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]] // @@ -432,10 +435,10 @@ int bar(int n){ // CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 0 // CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]], // - // CHECK: [[MEDIUM_ELT64:%.+]] = getelementptr inbounds [32 x i64], [32 x i64] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]] - // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i64 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT64]] to i8 addrspace([[SHARED_ADDRSPACE]])* + // CHECK: [[MEDIUM_ELT64:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]] + // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT64]] to i8 addrspace([[SHARED_ADDRSPACE]])* // CHECK: [[ELT_VAL:%.+]] = load i8, i8* [[ELT_VOID]], align - // CHECK: store i8 [[ELT_VAL]], i8 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align + // CHECK: store volatile i8 [[ELT_VAL]], i8 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align // CHECK: br label {{%?}}[[COPY_CONT:.+]] // // CHECK: [[COPY_ELSE]] @@ -443,20 +446,19 @@ int bar(int n){ // // Barrier after copy to shared memory storage medium. // CHECK: [[COPY_CONT]] - // CHECK: [[WS:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize() - // CHECK: [[ACTIVE_THREADS:%.+]] = mul nsw i32 [[ACTIVE_WARPS:%.+]], [[WS]] - // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]]) + // CHECK: call void @__kmpc_barrier(%struct.ident_t* @ + // CHECK: [[ACTIVE_WARPS:%.+]] = load i32, i32* // // Read into warp 0. // CHECK: [[IS_W0_ACTIVE_THREAD:%.+]] = icmp ult i32 [[TID:%.+]], [[ACTIVE_WARPS]] // CHECK: br i1 [[IS_W0_ACTIVE_THREAD]], label {{%?}}[[DO_READ:.+]], label {{%?}}[[READ_ELSE:.+]] // // CHECK: [[DO_READ]] - // CHECK: [[MEDIUM_ELT64:%.+]] = getelementptr inbounds [32 x i64], [32 x i64] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]] - // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i64 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT64]] to i8 addrspace([[SHARED_ADDRSPACE]])* + // CHECK: [[MEDIUM_ELT32:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]] + // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT32]] to i8 addrspace([[SHARED_ADDRSPACE]])* // CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 0 // CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]], - // CHECK: [[MEDIUM_ELT_VAL:%.+]] = load i8, i8 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align + // CHECK: [[MEDIUM_ELT_VAL:%.+]] = load volatile i8, i8 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align // CHECK: store i8 [[MEDIUM_ELT_VAL]], i8* [[ELT_VOID]], align // CHECK: br label {{%?}}[[READ_CONT:.+]] // @@ -464,19 +466,18 @@ int bar(int n){ // CHECK: br label {{%?}}[[READ_CONT]] // // CHECK: [[READ_CONT]] - // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]]) + // CHECK: call void @__kmpc_barrier(%struct.ident_t* @ // CHECK: [[IS_WARP_MASTER:%.+]] = icmp eq i32 [[LANEID]], 0 // CHECK: br i1 [[IS_WARP_MASTER]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]] // // [[DO_COPY]] // CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST]], i{{32|64}} 0, i{{32|64}} 1 // CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]], - // CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to float* + // CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i32* // - // CHECK: [[MEDIUM_ELT64:%.+]] = getelementptr inbounds [32 x i64], [32 x i64] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]] - // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i64 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT64]] to float addrspace([[SHARED_ADDRSPACE]])* - // CHECK: [[ELT_VAL:%.+]] = load float, float* [[ELT]], align - // CHECK: store float [[ELT_VAL]], float addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align + // CHECK: [[MEDIUM_ELT:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]] + // CHECK: [[ELT_VAL:%.+]] = load i32, i32* [[ELT]], align + // CHECK: store volatile i32 [[ELT_VAL]], i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align // CHECK: br label {{%?}}[[COPY_CONT:.+]] // // CHECK: [[COPY_ELSE]] @@ -484,29 +485,26 @@ int bar(int n){ // // Barrier after copy to shared memory storage medium. // CHECK: [[COPY_CONT]] - // CHECK: [[WS:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize() - // CHECK: [[ACTIVE_THREADS:%.+]] = mul nsw i32 [[ACTIVE_WARPS:%.+]], [[WS]] - // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]]) + // CHECK: call void @__kmpc_barrier(%struct.ident_t* @ + // CHECK: [[ACTIVE_WARPS:%.+]] = load i32, i32* // // Read into warp 0. // CHECK: [[IS_W0_ACTIVE_THREAD:%.+]] = icmp ult i32 [[TID:%.+]], [[ACTIVE_WARPS]] // CHECK: br i1 [[IS_W0_ACTIVE_THREAD]], label {{%?}}[[DO_READ:.+]], label {{%?}}[[READ_ELSE:.+]] // // CHECK: [[DO_READ]] - // CHECK: [[MEDIUM_ELT64:%.+]] = getelementptr inbounds [32 x i64], [32 x i64] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]] - // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i64 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT64]] to float addrspace([[SHARED_ADDRSPACE]])* + // CHECK: [[MEDIUM_ELT:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]] // CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 1 // CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]], - // CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to float* - // CHECK: [[MEDIUM_ELT_VAL:%.+]] = load float, float addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align - // CHECK: store float [[MEDIUM_ELT_VAL]], float* [[ELT]], align + // CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i32* + // CHECK: [[MEDIUM_ELT_VAL:%.+]] = load volatile i32, i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align + // CHECK: store i32 [[MEDIUM_ELT_VAL]], i32* [[ELT]], align // CHECK: br label {{%?}}[[READ_CONT:.+]] // // CHECK: [[READ_ELSE]] // CHECK: br label {{%?}}[[READ_CONT]] // // CHECK: [[READ_CONT]] - // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]]) // CHECK: ret @@ -520,13 +518,13 @@ int bar(int n){ // CHECK: define {{.*}}void {{@__omp_offloading_.+template.+l38}}( // - // CHECK: call void @__kmpc_spmd_kernel_init( + // CHECK: call void @__kmpc_spmd_kernel_init(i32 {{.+}}, i16 1, i16 0) // CHECK: call void @__kmpc_data_sharing_init_stack_spmd // CHECK: br label {{%?}}[[EXECUTE:.+]] // // CHECK: [[EXECUTE]] // CHECK: {{call|invoke}} void [[PFN2:@.+]](i32* - // CHECK: call void @__kmpc_spmd_kernel_deinit() + // CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1) // // // define internal void [[PFN2]]( @@ -559,11 +557,10 @@ int bar(int n){ // CHECK: [[B_CAST:%.+]] = bitcast i16* [[B]] to i8* // CHECK: store i8* [[B_CAST]], i8** [[PTR2]], align // CHECK: [[ARG_RL:%.+]] = bitcast [[RLT]]* [[RL]] to i8* - // CHECK: [[RET:%.+]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait(i32 {{.+}}, i32 2, i{{32|64}} {{8|16}}, i8* [[ARG_RL]], void (i8*, i16, i16, i16)* [[SHUFFLE_REDUCE_FN:@.+]], void (i8*, i32)* [[WARP_COPY_FN:@.+]]) - // CHECK: switch i32 [[RET]], label {{%?}}[[DEFAULTLABEL:.+]] [ - // CHECK: i32 1, label {{%?}}[[REDLABEL:.+]] + // CHECK: [[RET:%.+]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(%struct.ident_t* @{{.+}}, i32 {{.+}}, i32 2, i{{32|64}} {{8|16}}, i8* [[ARG_RL]], void (i8*, i16, i16, i16)* [[SHUFFLE_REDUCE_FN:@.+]], void (i8*, i32)* [[WARP_COPY_FN:@.+]]) + // CHECK: [[CMP:%.+]] = icmp eq i32 [[RET]], 1 + // CHECK: br i1 [[CMP]], label - // CHECK: [[REDLABEL]] // CHECK: [[A_INV:%.+]] = load i32, i32* [[A_IN:%.+]], align // CHECK: [[AV:%.+]] = load i32, i32* [[A]], align // CHECK: [[OR:%.+]] = or i32 [[A_INV]], [[AV]] @@ -587,9 +584,8 @@ int bar(int n){ // CHECK: [[B_MAX:%.+]] = phi i16 [ [[MAX1]], %[[DO_MAX]] ], [ [[MAX2]], %[[MAX_ELSE]] ] // CHECK: store i16 [[B_MAX]], i16* [[B_IN]], align // CHECK: call void @__kmpc_nvptx_end_reduce_nowait( - // CHECK: br label %[[DEFAULTLABEL]] + // CHECK: br label // - // CHECK: [[DEFAULTLABEL]] // CHECK: ret // @@ -744,6 +740,7 @@ int bar(int n){ // CHECK-DAG: [[LANEID:%.+]] = and i32 {{.+}}, 31 // CHECK-DAG: [[WARPID:%.+]] = ashr i32 {{.+}}, 5 // CHECK-DAG: [[RED_LIST:%.+]] = bitcast i8* {{.+}} to [[RLT]]* + // CHECK: call void @__kmpc_barrier(%struct.ident_t* @ // CHECK: [[IS_WARP_MASTER:%.+]] = icmp eq i32 [[LANEID]], 0 // CHECK: br i1 [[IS_WARP_MASTER]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]] // @@ -752,10 +749,9 @@ int bar(int n){ // CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]], // CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i32* // - // CHECK: [[MEDIUM_ELT64:%.+]] = getelementptr inbounds [32 x i64], [32 x i64] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]] - // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i64 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT64]] to i32 addrspace([[SHARED_ADDRSPACE]])* + // CHECK: [[MEDIUM_ELT:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]] // CHECK: [[ELT_VAL:%.+]] = load i32, i32* [[ELT]], align - // CHECK: store i32 [[ELT_VAL]], i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align + // CHECK: store volatile i32 [[ELT_VAL]], i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align // CHECK: br label {{%?}}[[COPY_CONT:.+]] // // CHECK: [[COPY_ELSE]] @@ -763,21 +759,19 @@ int bar(int n){ // // Barrier after copy to shared memory storage medium. // CHECK: [[COPY_CONT]] - // CHECK: [[WS:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize() - // CHECK: [[ACTIVE_THREADS:%.+]] = mul nsw i32 [[ACTIVE_WARPS:%.+]], [[WS]] - // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]]) + // CHECK: call void @__kmpc_barrier(%struct.ident_t* @ + // CHECK: [[ACTIVE_WARPS:%.+]] = load i32, i32* // // Read into warp 0. // CHECK: [[IS_W0_ACTIVE_THREAD:%.+]] = icmp ult i32 [[TID:%.+]], [[ACTIVE_WARPS]] // CHECK: br i1 [[IS_W0_ACTIVE_THREAD]], label {{%?}}[[DO_READ:.+]], label {{%?}}[[READ_ELSE:.+]] // // CHECK: [[DO_READ]] - // CHECK: [[MEDIUM_ELT64:%.+]] = getelementptr inbounds [32 x i64], [32 x i64] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]] - // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i64 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT64]] to i32 addrspace([[SHARED_ADDRSPACE]])* + // CHECK: [[MEDIUM_ELT:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]] // CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 0 // CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]], // CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i32* - // CHECK: [[MEDIUM_ELT_VAL:%.+]] = load i32, i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align + // CHECK: [[MEDIUM_ELT_VAL:%.+]] = load volatile i32, i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align // CHECK: store i32 [[MEDIUM_ELT_VAL]], i32* [[ELT]], align // CHECK: br label {{%?}}[[READ_CONT:.+]] // @@ -785,7 +779,7 @@ int bar(int n){ // CHECK: br label {{%?}}[[READ_CONT]] // // CHECK: [[READ_CONT]] - // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]]) + // CHECK: call void @__kmpc_barrier(%struct.ident_t* @ // CHECK: [[IS_WARP_MASTER:%.+]] = icmp eq i32 [[LANEID]], 0 // CHECK: br i1 [[IS_WARP_MASTER]], label {{%?}}[[DO_COPY:.+]], label {{%?}}[[COPY_ELSE:.+]] // @@ -794,10 +788,10 @@ int bar(int n){ // CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]], // CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i16* // - // CHECK: [[MEDIUM_ELT64:%.+]] = getelementptr inbounds [32 x i64], [32 x i64] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]] - // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i64 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT64]] to i16 addrspace([[SHARED_ADDRSPACE]])* + // CHECK: [[MEDIUM_ELT32:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[WARPID]] + // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT32]] to i16 addrspace([[SHARED_ADDRSPACE]])* // CHECK: [[ELT_VAL:%.+]] = load i16, i16* [[ELT]], align - // CHECK: store i16 [[ELT_VAL]], i16 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align + // CHECK: store volatile i16 [[ELT_VAL]], i16 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align // CHECK: br label {{%?}}[[COPY_CONT:.+]] // // CHECK: [[COPY_ELSE]] @@ -805,21 +799,20 @@ int bar(int n){ // // Barrier after copy to shared memory storage medium. // CHECK: [[COPY_CONT]] - // CHECK: [[WS:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.warpsize() - // CHECK: [[ACTIVE_THREADS:%.+]] = mul nsw i32 [[ACTIVE_WARPS:%.+]], [[WS]] - // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]]) + // CHECK: call void @__kmpc_barrier(%struct.ident_t* @ + // CHECK: [[ACTIVE_WARPS:%.+]] = load i32, i32* // // Read into warp 0. // CHECK: [[IS_W0_ACTIVE_THREAD:%.+]] = icmp ult i32 [[TID:%.+]], [[ACTIVE_WARPS]] // CHECK: br i1 [[IS_W0_ACTIVE_THREAD]], label {{%?}}[[DO_READ:.+]], label {{%?}}[[READ_ELSE:.+]] // // CHECK: [[DO_READ]] - // CHECK: [[MEDIUM_ELT64:%.+]] = getelementptr inbounds [32 x i64], [32 x i64] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]] - // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i64 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT64]] to i16 addrspace([[SHARED_ADDRSPACE]])* + // CHECK: [[MEDIUM_ELT32:%.+]] = getelementptr inbounds [32 x i32], [32 x i32] addrspace([[SHARED_ADDRSPACE]])* [[TRANSFER_STORAGE]], i64 0, i32 [[TID]] + // CHECK: [[MEDIUM_ELT:%.+]] = bitcast i32 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT32]] to i16 addrspace([[SHARED_ADDRSPACE]])* // CHECK: [[ELT_REF:%.+]] = getelementptr inbounds [[RLT]], [[RLT]]* [[RED_LIST:%.+]], i{{32|64}} 0, i{{32|64}} 1 // CHECK: [[ELT_VOID:%.+]] = load i8*, i8** [[ELT_REF]], // CHECK: [[ELT:%.+]] = bitcast i8* [[ELT_VOID]] to i16* - // CHECK: [[MEDIUM_ELT_VAL:%.+]] = load i16, i16 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align + // CHECK: [[MEDIUM_ELT_VAL:%.+]] = load volatile i16, i16 addrspace([[SHARED_ADDRSPACE]])* [[MEDIUM_ELT]], align // CHECK: store i16 [[MEDIUM_ELT_VAL]], i16* [[ELT]], align // CHECK: br label {{%?}}[[READ_CONT:.+]] // @@ -827,7 +820,6 @@ int bar(int n){ // CHECK: br label {{%?}}[[READ_CONT]] // // CHECK: [[READ_CONT]] - // CHECK: call void @llvm.nvvm.barrier(i32 1, i32 [[ACTIVE_THREADS]]) // CHECK: ret #endif |