DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl > Struct Template Reference

• ck
• tensor_operation
• device
• DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K

Classes | Public Types | Public Member Functions | Static Public Member Functions | Static Public Attributes | List of all members

ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl > Struct Template Reference

#include <device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp>

Inheritance diagram for ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >:

Classes
struct	Argument
struct	Invoker

Public Types
using	DeviceOp = DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K
using	ADataType = InDataType
using	BDataType = WeiDataType
using	CDataType = OutDataType
using	ABDataType = InDataType
using	ABCGridDescs
using	AGridDesc_K0_M_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I0])>
using	BGridDesc_K0_N_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I1])>
using	CGridDesc_M_N = remove_cvref_t<decltype(ABCGridDescs{}[I2])>
using	Block2CTileMap = BlockToCTileMap_M00_N0_M01<MPerBlock, NPerBlock, CGridDesc_M_N>
template<index_t NXdlPerWave_>
using	GridwiseGemmBase
using	GridwiseGemm64 = GridwiseGemmBase<math::max(NXdlPerWave64, 1)>
using	GridwiseGemm32 = GridwiseGemmBase<math::max(NXdlPerWave32, 1)>

Public Member Functions
bool	IsSupportedArgument (const BaseArgument *p_arg) override
std::unique_ptr< BaseArgument >	MakeArgumentPointer (const void *p_in_grid, const void *p_wei_grid, void *p_out_grid, ck::index_t N, ck::index_t K, ck::index_t C, std::vector< ck::index_t > input_spatial_lengths, std::vector< ck::index_t > filter_spatial_lengths, std::vector< ck::index_t > output_spatial_lengths, std::vector< ck::index_t > conv_filter_strides, std::vector< ck::index_t > conv_filter_dilations, std::vector< ck::index_t > input_left_pads, std::vector< ck::index_t > input_right_pads, InElementwiseOperation in_element_op, WeiElementwiseOperation wei_element_op, OutElementwiseOperation out_element_op) override
std::unique_ptr< BaseInvoker >	MakeInvokerPointer () override
std::string	GetTypeString () const override
Public Member Functions inherited from ck::tensor_operation::device::BaseOperator
	BaseOperator ()=default
	BaseOperator (const BaseOperator &)=default
BaseOperator &	operator= (const BaseOperator &)=default
virtual std::string	GetInstanceString () const
virtual std::string	GetTypeIdName () const
virtual std::optional< std::string >	GetObjectName () const
virtual std::optional< std::string >	GetTemplateInfo () const
virtual std::string	GetTypeIdHashCode () const
virtual size_t	GetWorkSpaceSize (const BaseArgument *) const
virtual void	SetWorkSpacePointer (BaseArgument *p_arg, void *p_workspace, const StreamConfig &=StreamConfig{}) const
virtual	~BaseOperator ()

Static Public Member Functions
static auto	MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N (ck::index_t N, ck::index_t K, ck::index_t C, std::vector< ck::index_t > input_spatial_lengths, std::vector< ck::index_t > filter_spatial_lengths, std::vector< ck::index_t > output_spatial_lengths, std::vector< ck::index_t > conv_filter_strides, std::vector< ck::index_t > conv_filter_dilations, std::vector< ck::index_t > input_left_pads, std::vector< ck::index_t > input_right_pads)
static constexpr bool	IsValidCompilationParameter ()
static bool	IsSupportedArgument (const Argument &arg)
static auto	MakeArgument (const InDataType *p_in_grid, const WeiDataType *p_wei_grid, OutDataType *p_out_grid, ck::index_t N, ck::index_t K, ck::index_t C, std::vector< ck::index_t > input_spatial_lengths, std::vector< ck::index_t > filter_spatial_lengths, std::vector< ck::index_t > output_spatial_lengths, std::vector< ck::index_t > conv_filter_strides, std::vector< ck::index_t > conv_filter_dilations, std::vector< ck::index_t > input_left_pads, std::vector< ck::index_t > input_right_pads, InElementwiseOperation in_element_op, WeiElementwiseOperation wei_element_op, OutElementwiseOperation out_element_op)
static auto	MakeInvoker ()

Static Public Attributes
static GET_NXDL_PER_WAVE_IMPL constexpr auto	NXdlPerWave64 = GetNXdlPerWave<true>()
static constexpr auto	NXdlPerWave32 = GetNXdlPerWave<false>()
static constexpr index_t	NDimSpatial = 2
static constexpr auto	I0 = Number<0>{}
static constexpr auto	I1 = Number<1>{}
static constexpr auto	I2 = Number<2>{}
static constexpr auto	I3 = Number<3>{}
static constexpr auto	I4 = Number<4>{}
static constexpr auto	I5 = Number<5>{}
static constexpr auto	K1Number = Number<K1>{}
static constexpr auto	GemmK1Number = K1Number

Member Typedef Documentation

◆ ABCGridDescs

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::ABCGridDescs

Initial value:

decltype(MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(

1, 1, 1, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}, {1, 1}))

ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K::MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N

static auto MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(ck::index_t N, ck::index_t K, ck::index_t C, std::vector< ck::index_t > input_spatial_lengths, std::vector< ck::index_t > filter_spatial_lengths, std::vector< ck::index_t > output_spatial_lengths, std::vector< ck::index_t > conv_filter_strides, std::vector< ck::index_t > conv_filter_dilations, std::vector< ck::index_t > input_left_pads, std::vector< ck::index_t > input_right_pads)

Definition device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp:99

◆ ABDataType

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::ABDataType = InDataType

◆ ADataType

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::ADataType = InDataType

◆ AGridDesc_K0_M_K1

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::AGridDesc_K0_M_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I0])>

◆ BDataType

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::BDataType = WeiDataType

◆ BGridDesc_K0_N_K1

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::BGridDesc_K0_N_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I1])>

◆ Block2CTileMap

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::Block2CTileMap = BlockToCTileMap_M00_N0_M01<MPerBlock, NPerBlock, CGridDesc_M_N>

◆ CDataType

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::CDataType = OutDataType

◆ CGridDesc_M_N

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::CGridDesc_M_N = remove_cvref_t<decltype(ABCGridDescs{}[I2])>

◆ DeviceOp

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::DeviceOp = DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K

◆ GridwiseGemm32

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::GridwiseGemm32 = GridwiseGemmBase<math::max(NXdlPerWave32, 1)>

◆ GridwiseGemm64

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::GridwiseGemm64 = GridwiseGemmBase<math::max(NXdlPerWave64, 1)>

◆ GridwiseGemmBase

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

template<index_t NXdlPerWave_>

using ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::GridwiseGemmBase

Member Function Documentation

◆ GetTypeString()

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

std::string ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::GetTypeString ( ) const

inlineoverridevirtual

Reimplemented from ck::tensor_operation::device::BaseOperator.

◆ IsSupportedArgument() [1/2]

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

bool ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::IsSupportedArgument ( const Argument & arg )

inlinestatic

◆ IsSupportedArgument() [2/2]

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

bool ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::IsSupportedArgument ( const BaseArgument * p_arg )

inlineoverridevirtual

Reimplemented from ck::tensor_operation::device::BaseOperator.

◆ IsValidCompilationParameter()

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

constexpr bool ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::IsValidCompilationParameter ( )

inlinestaticconstexpr

◆ MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N()

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N ( ck::index_t N, ck::index_t K, ck::index_t C, std::vector< ck::index_t > input_spatial_lengths, std::vector< ck::index_t > filter_spatial_lengths, std::vector< ck::index_t > output_spatial_lengths, std::vector< ck::index_t > conv_filter_strides, std::vector< ck::index_t > conv_filter_dilations, std::vector< ck::index_t > input_left_pads, std::vector< ck::index_t > input_right_pads )

inlinestatic

◆ MakeArgument()

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::MakeArgument ( const InDataType * p_in_grid, const WeiDataType * p_wei_grid, OutDataType * p_out_grid, ck::index_t N, ck::index_t K, ck::index_t C, std::vector< ck::index_t > input_spatial_lengths, std::vector< ck::index_t > filter_spatial_lengths, std::vector< ck::index_t > output_spatial_lengths, std::vector< ck::index_t > conv_filter_strides, std::vector< ck::index_t > conv_filter_dilations, std::vector< ck::index_t > input_left_pads, std::vector< ck::index_t > input_right_pads, InElementwiseOperation in_element_op, WeiElementwiseOperation wei_element_op, OutElementwiseOperation out_element_op )

inlinestatic

◆ MakeArgumentPointer()

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

std::unique_ptr< BaseArgument > ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::MakeArgumentPointer ( const void * p_in_grid, const void * p_wei_grid, void * p_out_grid, ck::index_t N, ck::index_t K, ck::index_t C, std::vector< ck::index_t > input_spatial_lengths, std::vector< ck::index_t > filter_spatial_lengths, std::vector< ck::index_t > output_spatial_lengths, std::vector< ck::index_t > conv_filter_strides, std::vector< ck::index_t > conv_filter_dilations, std::vector< ck::index_t > input_left_pads, std::vector< ck::index_t > input_right_pads, InElementwiseOperation in_element_op, WeiElementwiseOperation wei_element_op, OutElementwiseOperation out_element_op )

inlineoverridevirtual

Implements ck::tensor_operation::device::DeviceConvFwd< 2, ck::tensor_layout::convolution::NHWC, ck::tensor_layout::convolution::KYXC, ck::tensor_layout::convolution::NHWK, InDataType, WeiDataType, OutDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation >.

◆ MakeInvoker()

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::MakeInvoker ( )

inlinestatic

◆ MakeInvokerPointer()

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

std::unique_ptr< BaseInvoker > ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::MakeInvokerPointer ( )

inlineoverridevirtual

Implements ck::tensor_operation::device::DeviceConvFwd< 2, ck::tensor_layout::convolution::NHWC, ck::tensor_layout::convolution::KYXC, ck::tensor_layout::convolution::NHWK, InDataType, WeiDataType, OutDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation >.

Member Data Documentation

◆ GemmK1Number

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::GemmK1Number = K1Number

staticconstexpr

◆ I0

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::I0 = Number<0>{}

staticconstexpr

◆ I1

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::I1 = Number<1>{}

staticconstexpr

◆ I2

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::I2 = Number<2>{}

staticconstexpr

◆ I3

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::I3 = Number<3>{}

staticconstexpr

◆ I4

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::I4 = Number<4>{}

staticconstexpr

◆ I5

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::I5 = Number<5>{}

staticconstexpr

◆ K1Number

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::K1Number = Number<K1>{}

staticconstexpr

◆ NDimSpatial

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

index_t ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::NDimSpatial = 2

staticconstexpr

◆ NXdlPerWave32

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::NXdlPerWave32 = GetNXdlPerWave<false>()

staticconstexpr

◆ NXdlPerWave64

template<typename InDataType, typename WeiDataType, typename OutDataType, typename AccDataType, typename InElementwiseOperation, typename WeiElementwiseOperation, typename OutElementwiseOperation, ConvolutionForwardSpecialization ConvForwardSpecialization, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsAddExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, index_t CBlockTransferScalarPerVector_NWaveNPerXdl>

GET_NXDL_PER_WAVE_IMPL constexpr auto ck::tensor_operation::device::DeviceConv2dFwdXdl_C_Shuffle_Input_N_Hi_Wi_C_Weight_K_Y_X_C_Output_N_Ho_Wo_K< InDataType, WeiDataType, OutDataType, AccDataType, InElementwiseOperation, WeiElementwiseOperation, OutElementwiseOperation, ConvForwardSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsAddExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CBlockTransferClusterLengths_MBlock_MXdlPerWave_MWaveMPerXdl_NBlock_NXdlPerWave_NWaveNPerXdl, CBlockTransferScalarPerVector_NWaveNPerXdl >::NXdlPerWave64 = GetNXdlPerWave<true>()

staticconstexpr

---

The documentation for this struct was generated from the following file:

• device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp