Skip to content

GitLab

Unverified Commit 4634b120 authored by Shaojie WANG's avatar Shaojie WANG Committed by GitHub
Browse files

fix Issue 291 (#294) 

* rename for typeconvert functor

* refine code
parent 15c89e81
Hide whitespace changes
Inline Side-by-side
Showing with 71 additions and 128 deletions
include/ck/tensor_operation/gpu/device/device_convnd_backward_weight_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp
View file @ 4634b120
......@@ -433,7 +433,7 @@ struct DeviceConvndBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
using namespace ck;
const index_t Di = input_spatial_lengths[0];
const index_t Hi = input_spatial_lengths[2];
const index_t Hi = input_spatial_lengths[1];
const index_t Wi = input_spatial_lengths[2];
const index_t Do = output_spatial_lengths[0];
......@@ -671,11 +671,14 @@ struct DeviceConvndBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
return PadDescriptor_M0_1d(desc, gridSize, blockSize);
}
using TypeConvertFunctor =
using TypeConvertFp32ToBf16Functor =
ck::tensor_operation::element_wise::UnaryTypeConvert<ck::bhalf_t, float>;
using GridDesc_M0 = decltype(MakeDescriptor_M0<1>({1}, {1}, 1, 1));
using GridwiseUEltwise =
GridwiseUnaryElementwise_1D<AccDataType, InDataType, GridDesc_M0, TypeConvertFunctor, 4>;
using GridDesc_M0 = decltype(MakeDescriptor_M0<1>({1}, {1}, 1, 1));
using GridwiseUEltwise = GridwiseUnaryElementwise_1D<AccDataType,
InDataType,
GridDesc_M0,
TypeConvertFp32ToBf16Functor,
4>;
using ABCGridDescs = decltype(GetABCGridDesc<NumDimSpatial>());
......@@ -979,33 +982,32 @@ struct DeviceConvndBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
const auto K0 = arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I1);
const bool has_main_k0_block_loop = GridwiseGemm::CalculateHasMainK0BlockLoop(K0);
float ave_time = 0;
const auto Run = [&](const auto& kernel) {
const bool has_main_k0_block_loop = GridwiseGemm::CalculateHasMainK0BlockLoop(K0);
const auto run_conv = [&](const auto& kernel) {
hipGetErrorString(hipMemset(
arg.p_c_grid_,
0,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_.GetElementSpaceSize() *
sizeof(CDataType)));
ave_time =
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_kbatch_k0_m_k1_,
arg.b_grid_desc_kbatch_k0_n_k1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.block_2_ctile_map_);
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_c_grid_,
arg.a_grid_desc_kbatch_k0_m_k1_,
arg.b_grid_desc_kbatch_k0_n_k1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.block_2_ctile_map_);
};
// run kernel for bf16 with splitk
......@@ -1016,22 +1018,21 @@ struct DeviceConvndBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_.GetElementSpaceSize() *
sizeof(AccDataType)));
ave_time =
launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
static_cast<AccDataType*>(arg.p_workspace_),
arg.a_grid_desc_kbatch_k0_m_k1_,
arg.b_grid_desc_kbatch_k0_n_k1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.block_2_ctile_map_);
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
static_cast<AccDataType*>(arg.p_workspace_),
arg.a_grid_desc_kbatch_k0_m_k1_,
arg.b_grid_desc_kbatch_k0_n_k1_,
arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.a_element_op_,
arg.b_element_op_,
arg.c_element_op_,
arg.block_2_ctile_map_);
};
// kernel for type conversion
......@@ -1059,7 +1060,7 @@ struct DeviceConvndBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
// run kernel for type conversion
void* p_c_grid_tmp_ = static_cast<void*>(arg.p_c_grid_);
InDataType* p_c_grid_tmp_bf16_ = static_cast<InDataType*>(p_c_grid_tmp_);
const auto Run_type_convert = [&](const auto& kernel) {
const auto run_type_convert = [&](const auto& kernel) {
float elapsed_time =
launch_and_time_kernel(stream_config,
kernel,
......@@ -1070,14 +1071,15 @@ struct DeviceConvndBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
p_c_grid_tmp_bf16_,
a_grid_desc_m0_,
b_grid_desc_m0_,
TypeConvertFunctor{});
TypeConvertFp32ToBf16Functor{});
return elapsed_time;
};
if constexpr(std::is_same<InDataType, ck::bhalf_t>::value)
{
if(has_main_k0_block_loop)
{
auto launch_kernel = [&](auto has_main_k_block_loop) {
constexpr bool has_main_loop = has_main_k_block_loop.value;
if(kbatch == 1)
{
const auto kernel = kernel_gemm_xdlops_bwd_weight<
......@@ -1092,9 +1094,9 @@ struct DeviceConvndBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
InElementwiseOperation,
WeiElementwiseOperation,
remove_reference_t<DeviceOp::Block2CTileMap>,
true>;
has_main_loop>;
Run(kernel);
return run_conv(kernel);
}
else
{
......@@ -1103,7 +1105,7 @@ struct DeviceConvndBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
AccDataType,
InDataType,
GridDesc_M0,
TypeConvertFunctor>;
TypeConvertFp32ToBf16Functor>;
const auto kernel_conv = kernel_gemm_xdlops_bwd_weight<
GridwiseGemmAtomicAddFloatBf16Splitk,
......@@ -1117,56 +1119,28 @@ struct DeviceConvndBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
InElementwiseOperation,
WeiElementwiseOperation,
remove_reference_t<DeviceOp::Block2CTileMap>,
true>;
has_main_loop>;
run_bf16_splitk(kernel_conv);
ave_time += Run_type_convert(kernel_type_convert);
float elapsed_time = 0;
elapsed_time += run_bf16_splitk(kernel_conv);
elapsed_time += run_type_convert(kernel_type_convert);
return elapsed_time;
}
};
if(has_main_k0_block_loop)
{
ave_time = launch_kernel(integral_constant<bool, true>{});
}
else
{
if(kbatch == 1)
{
const auto kernel = kernel_gemm_xdlops_bwd_weight<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
OutElementwiseOperation,
InElementwiseOperation,
WeiElementwiseOperation,
remove_reference_t<DeviceOp::Block2CTileMap>,
false>;
Run(kernel);
}
else
{
const auto kernel = kernel_gemm_xdlops_bwd_weight<
GridwiseGemmAtomicAddFloatBf16Splitk,
ADataType, // TODO: distiguish A/B datatype
AccDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
OutElementwiseOperation,
InElementwiseOperation,
WeiElementwiseOperation,
remove_reference_t<DeviceOp::Block2CTileMap>,
false>;
run_bf16_splitk(kernel);
}
ave_time = launch_kernel(integral_constant<bool, false>{});
}
}
else
{
if(has_main_k0_block_loop)
{
auto launch_kernel = [&](auto has_main_k_block_loop) {
constexpr bool has_main_loop = has_main_k_block_loop.value;
if(kbatch == 1)
{
const auto kernel = kernel_gemm_xdlops_bwd_weight<
......@@ -1181,9 +1155,9 @@ struct DeviceConvndBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
InElementwiseOperation,
WeiElementwiseOperation,
remove_reference_t<DeviceOp::Block2CTileMap>,
true>;
has_main_loop>;
Run(kernel);
return run_conv(kernel);
}
else
{
......@@ -1199,49 +1173,18 @@ struct DeviceConvndBwdWeightXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_
InElementwiseOperation,
WeiElementwiseOperation,
remove_reference_t<DeviceOp::Block2CTileMap>,
true>;
has_main_loop>;
Run(kernel);
return run_conv(kernel);
}
};
if(has_main_k0_block_loop)
{
ave_time = launch_kernel(integral_constant<bool, true>{});
}
else
{
if(kbatch == 1)
{
const auto kernel = kernel_gemm_xdlops_bwd_weight<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
OutElementwiseOperation,
InElementwiseOperation,
WeiElementwiseOperation,
remove_reference_t<DeviceOp::Block2CTileMap>,
false>;
Run(kernel);
}
else
{
const auto kernel = kernel_gemm_xdlops_bwd_weight<
GridwiseGemmAtomicAdd,
ADataType, // TODO: distiguish A/B datatype
CDataType,
remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
remove_reference_t<
DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
OutElementwiseOperation,
InElementwiseOperation,
WeiElementwiseOperation,
remove_reference_t<DeviceOp::Block2CTileMap>,
false>;
Run(kernel);
}
ave_time = launch_kernel(integral_constant<bool, false>{});
}
}
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment