Composable Kernel: ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched > Struct Template Reference

DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched > Struct Template Reference

DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched > Struct Template Reference#

Composable Kernel: ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched > Struct Template Reference

ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched > Struct Template Reference

#include <device_grouped_conv_fwd_multiple_d_xdl_large_tensor_cshuffle.hpp>

Inheritance diagram for ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >:

Classes
struct	Argument

struct	GemmArgs

struct	Invoker

Public Types
using	DeviceOp = DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor

using	ConvToGemmFwdTransformerIndexT = TransformConvFwdToGemm< NDimSpatial, ConvForwardSpecialization, true, ADataType, EDataType, I1, index_t >

using	ConvToGemmFwdTransformerLongIndexT = TransformConvFwdToGemm< NDimSpatial, ConvForwardSpecialization, true, ADataType, EDataType, I1, long_index_t >

using	AGridDesc_M_K = remove_cvref_t< decltype(MakeAGridDescriptor_M_K< ALayout >(dummy_conv_to_gemm_transformer))>

using	BGridDesc_N_K = remove_cvref_t< decltype(MakeBGridDescriptor_N_K< BLayout >(dummy_conv_to_gemm_transformer))>

using	EGridDesc_M_N = remove_cvref_t< decltype(MakeEGridDescriptor_M_N< ELayout >(dummy_conv_to_gemm_transformer))>

using	GridwiseGemm = GridwiseGemmMultipleD_xdl_cshuffle< GridwiseGemmTemplateParameters >

using	AGridDesc_AK0_M_AK1 = remove_cvref_t< decltype(GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1(AGridDesc_M_K{}))>

using	BGridDesc_BK0_N_BK1 = remove_cvref_t< decltype(GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1(BGridDesc_N_K{}))>

using	EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t< decltype(GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(EGridDesc_M_N{}))>

using	Block2ETileMap = remove_cvref_t< decltype(GridwiseGemm::MakeDefaultBlock2ETileMap(EGridDesc_M_N{}))>

Public Types inherited from ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, decltype(UnpackDataType< is_detected< is_tuple, ADataType >::value, Number< 0 >, ADataType >()), decltype(UnpackDataType< is_detected< is_tuple, ADataType >::value, Number< 0 >, ADataType >()) >
using	APointers = ck::conditional_t< isMultiA, std::array< const void , NumATensor > &, const void >

using	BPointers = ck::conditional_t< isMultiB, std::array< const void , NumBTensor > &, const void >

Public Member Functions
bool	IsSupportedArgument (const BaseArgument *p_arg) override

std::unique_ptr< BaseArgument >	MakeArgumentPointer (const void p_a, const void p_b, const std::array< const void , NumDTensor > &p_ds, void p_e, const std::array< index_t, NDimSpatial+3 > &a_g_n_c_wis_lengths, const std::array< index_t, NDimSpatial+3 > &a_g_n_c_wis_strides, const std::array< index_t, NDimSpatial+3 > &b_g_k_c_xs_lengths, const std::array< index_t, NDimSpatial+3 > &b_g_k_c_xs_strides, const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_lengths, const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_strides, const std::array< index_t, NDimSpatial+3 > &e_g_n_k_wos_lengths, const std::array< index_t, NDimSpatial+3 > &e_g_n_k_wos_strides, const std::array< index_t, NDimSpatial > &conv_filter_strides, const std::array< index_t, NDimSpatial > &conv_filter_dilations, const std::array< index_t, NDimSpatial > &input_left_pads, const std::array< index_t, NDimSpatial > &input_right_pads, const AElementwiseOperation &a_element_op, const BElementwiseOperation &b_element_op, const CDEElementwiseOperation &cde_element_op) override

std::unique_ptr< BaseArgument >	MakeArgumentPointer (const void p_a, const void p_b, const std::array< const void , NumDTensor > &p_ds, void p_e, const std::array< long_index_t, NDimSpatial+3 > &a_g_n_c_wis_lengths, const std::array< long_index_t, NDimSpatial+3 > &a_g_n_c_wis_strides, const std::array< long_index_t, NDimSpatial+3 > &b_g_k_c_xs_lengths, const std::array< long_index_t, NDimSpatial+3 > &b_g_k_c_xs_strides, const std::array< std::array< long_index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_lengths, const std::array< std::array< long_index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_strides, const std::array< long_index_t, NDimSpatial+3 > &e_g_n_k_wos_lengths, const std::array< long_index_t, NDimSpatial+3 > &e_g_n_k_wos_strides, const std::array< long_index_t, NDimSpatial > &conv_filter_strides, const std::array< long_index_t, NDimSpatial > &conv_filter_dilations, const std::array< long_index_t, NDimSpatial > &input_left_pads, const std::array< long_index_t, NDimSpatial > &input_right_pads, const AElementwiseOperation &a_element_op, const BElementwiseOperation &b_element_op, const CDEElementwiseOperation &cde_element_op) override

std::unique_ptr< BaseInvoker >	MakeInvokerPointer () override

std::string	GetTypeString () const override

Public Member Functions inherited from ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, decltype(UnpackDataType< is_detected< is_tuple, ADataType >::value, Number< 0 >, ADataType >()), decltype(UnpackDataType< is_detected< is_tuple, ADataType >::value, Number< 0 >, ADataType >()) >
virtual std::unique_ptr< BaseArgument >	MakeArgumentPointer (APointers p_a, BPointers p_b, const std::array< const void , NumDTensor > &p_ds, void p_e, const std::array< index_t, NDimSpatial+3 > &a_g_n_c_wis_lengths, const std::array< index_t, NDimSpatial+3 > &a_g_n_c_wis_strides, const std::array< index_t, NDimSpatial+3 > &b_g_k_c_xs_lengths, const std::array< index_t, NDimSpatial+3 > &b_g_k_c_xs_strides, const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_lengths, const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_strides, const std::array< index_t, NDimSpatial+3 > &e_g_n_k_wos_lengths, const std::array< index_t, NDimSpatial+3 > &e_g_n_k_wos_strides, const std::array< index_t, NDimSpatial > &conv_filter_strides, const std::array< index_t, NDimSpatial > &conv_filter_dilations, const std::array< index_t, NDimSpatial > &input_left_pads, const std::array< index_t, NDimSpatial > &input_right_pads, const AElementwiseOperation &a_element_op, const BElementwiseOperation &b_element_op, const CDEElementwiseOperation &cde_element_op)=0
	Make argument pointer for grouped conv fwd. More...

virtual std::unique_ptr< BaseArgument >	MakeArgumentPointer (APointers p_a, BPointers p_b, const std::array< const void , NumDTensor > &p_ds, void p_e, const std::array< long_index_t, NDimSpatial+3 > &a_g_n_c_wis_lengths, const std::array< long_index_t, NDimSpatial+3 > &a_g_n_c_wis_strides, const std::array< long_index_t, NDimSpatial+3 > &b_g_k_c_xs_lengths, const std::array< long_index_t, NDimSpatial+3 > &b_g_k_c_xs_strides, const std::array< std::array< long_index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_lengths, const std::array< std::array< long_index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_strides, const std::array< long_index_t, NDimSpatial+3 > &e_g_n_k_wos_lengths, const std::array< long_index_t, NDimSpatial+3 > &e_g_n_k_wos_strides, const std::array< long_index_t, NDimSpatial > &conv_filter_strides, const std::array< long_index_t, NDimSpatial > &conv_filter_dilations, const std::array< long_index_t, NDimSpatial > &input_left_pads, const std::array< long_index_t, NDimSpatial > &input_right_pads, const AElementwiseOperation &a_element_op, const BElementwiseOperation &b_element_op, const CDEElementwiseOperation &cde_element_op)=0

Public Member Functions inherited from ck::tensor_operation::device::BaseOperator
	BaseOperator ()=default

	BaseOperator (const BaseOperator &)=default

BaseOperator &	operator= (const BaseOperator &)=default

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
template<typename ALay >
static auto	MakeAGridDescriptor_M_K (const ConvToGemmFwdTransformerIndexT &conv_to_gemm_transformer)

template<typename BLay >
static auto	MakeBGridDescriptor_N_K (const ConvToGemmFwdTransformerIndexT &conv_to_gemm_transformer)

template<typename ELay >
static auto	MakeEGridDescriptor_M_N (const ConvToGemmFwdTransformerIndexT &conv_to_gemm_transformer)

static auto	GenerateConvToGemmTransforms (ConvToGemmFwdTransformerLongIndexT conv_to_gemm_transformer_base, const ADataType a_grid_ptr_base, EDataType c_grid_ptr_base)

static bool	IsSupportedArgument (const Argument &arg)

static auto	MakeArgument (const void p_a, const void p_b, const std::array< const void , NumDTensor > &p_ds, void p_e, const std::array< index_t, NDimSpatial+3 > &a_g_n_c_wis_lengths, const std::array< index_t, NDimSpatial+3 > &a_g_n_c_wis_strides, const std::array< index_t, NDimSpatial+3 > &b_g_k_c_xs_lengths, const std::array< index_t, NDimSpatial+3 > &b_g_k_c_xs_strides, const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_lengths, const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_strides, const std::array< index_t, NDimSpatial+3 > &e_g_n_k_wos_lengths, const std::array< index_t, NDimSpatial+3 > &e_g_n_k_wos_strides, const std::array< index_t, NDimSpatial > &conv_filter_strides, const std::array< index_t, NDimSpatial > &conv_filter_dilations, const std::array< index_t, NDimSpatial > &input_left_pads, const std::array< index_t, NDimSpatial > &input_right_pads, const AElementwiseOperation &a_element_op, const BElementwiseOperation &b_element_op, const CDEElementwiseOperation &cde_element_op)

static auto	MakeArgument (const void p_a, const void p_b, const std::array< const void , NumDTensor > &p_ds, void p_e, const std::array< long_index_t, NDimSpatial+3 > &a_g_n_c_wis_lengths, const std::array< long_index_t, NDimSpatial+3 > &a_g_n_c_wis_strides, const std::array< long_index_t, NDimSpatial+3 > &b_g_k_c_xs_lengths, const std::array< long_index_t, NDimSpatial+3 > &b_g_k_c_xs_strides, const std::array< std::array< long_index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_lengths, const std::array< std::array< long_index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_strides, const std::array< long_index_t, NDimSpatial+3 > &e_g_n_k_wos_lengths, const std::array< long_index_t, NDimSpatial+3 > &e_g_n_k_wos_strides, const std::array< long_index_t, NDimSpatial > &conv_filter_strides, const std::array< long_index_t, NDimSpatial > &conv_filter_dilations, const std::array< long_index_t, NDimSpatial > &input_left_pads, const std::array< long_index_t, NDimSpatial > &input_right_pads, const AElementwiseOperation &a_element_op, const BElementwiseOperation &b_element_op, const CDEElementwiseOperation &cde_element_op)

static auto	MakeInvoker ()

Static Public Attributes
static constexpr index_t	NumDTensor = DsDataType::Size()

static constexpr index_t	MaxGemmsNum = 32

static constexpr bool	DoElementwiseBeforeCShuffle

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	matrix_padder

constexpr static ConvToGemmFwdTransformerIndexT	dummy_conv_to_gemm_transformer

Static Public Attributes inherited from ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, decltype(UnpackDataType< is_detected< is_tuple, ADataType >::value, Number< 0 >, ADataType >()), decltype(UnpackDataType< is_detected< is_tuple, ADataType >::value, Number< 0 >, ADataType >()) >
static constexpr bool	isMultiA

static constexpr bool	isMultiB

static constexpr index_t	NumATensor

static constexpr index_t	NumBTensor

static constexpr index_t	NumDTensor

Member Typedef Documentation

◆ AGridDesc_AK0_M_AK1

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

using ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::AGridDesc_AK0_M_AK1 = remove_cvref_t<decltype(GridwiseGemm::MakeDefaultAGridDescriptor_AK0_M_AK1( AGridDesc_M_K{}))>

◆ AGridDesc_M_K

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

using ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::AGridDesc_M_K = remove_cvref_t<decltype(MakeAGridDescriptor_M_K<ALayout>(dummy_conv_to_gemm_transformer))>

◆ BGridDesc_BK0_N_BK1

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

using ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::BGridDesc_BK0_N_BK1 = remove_cvref_t<decltype(GridwiseGemm::MakeDefaultBGridDescriptor_BK0_N_BK1( BGridDesc_N_K{}))>

◆ BGridDesc_N_K

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

using ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::BGridDesc_N_K = remove_cvref_t<decltype(MakeBGridDescriptor_N_K<BLayout>(dummy_conv_to_gemm_transformer))>

◆ Block2ETileMap

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

using ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::Block2ETileMap = remove_cvref_t<decltype(GridwiseGemm::MakeDefaultBlock2ETileMap(EGridDesc_M_N{}))>

◆ ConvToGemmFwdTransformerIndexT

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

using ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::ConvToGemmFwdTransformerIndexT = TransformConvFwdToGemm<NDimSpatial, ConvForwardSpecialization, true , ADataType, EDataType, I1, index_t>

◆ ConvToGemmFwdTransformerLongIndexT

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

using ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::ConvToGemmFwdTransformerLongIndexT = TransformConvFwdToGemm<NDimSpatial, ConvForwardSpecialization, true , ADataType, EDataType, I1, long_index_t>

◆ DeviceOp

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

using ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::DeviceOp = DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor

◆ EGridDesc_M_N

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

using ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::EGridDesc_M_N = remove_cvref_t<decltype(MakeEGridDescriptor_M_N<ELayout>(dummy_conv_to_gemm_transformer))>

◆ EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

using ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock( EGridDesc_M_N{}))>

◆ GridwiseGemm

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

using ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::GridwiseGemm = GridwiseGemmMultipleD_xdl_cshuffle<GridwiseGemmTemplateParameters>

Member Function Documentation

◆ GenerateConvToGemmTransforms()

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

static auto ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::GenerateConvToGemmTransforms	(	ConvToGemmFwdTransformerLongIndexT	conv_to_gemm_transformer_base,
		const ADataType *	a_grid_ptr_base,
		EDataType *	c_grid_ptr_base
	)

inlinestatic

◆ GetTypeString()

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

std::string ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::GetTypeString ( ) const

inlineoverridevirtual

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

◆ IsSupportedArgument() [1/2]

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

static bool ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::IsSupportedArgument ( const Argument & arg )

inlinestatic

◆ IsSupportedArgument() [2/2]

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

bool ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::IsSupportedArgument ( const BaseArgument * p_arg )

inlineoverridevirtual

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

◆ MakeAGridDescriptor_M_K()

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

template<typename ALay >

static auto ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::MakeAGridDescriptor_M_K ( const ConvToGemmFwdTransformerIndexT & conv_to_gemm_transformer )

inlinestatic

◆ MakeArgument() [1/2]

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

static auto ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::MakeArgument	(	const void *	p_a,
		const void *	p_b,
		const std::array< const void *, NumDTensor > &	p_ds,
		void *	p_e,
		const std::array< index_t, NDimSpatial+3 > &	a_g_n_c_wis_lengths,
		const std::array< index_t, NDimSpatial+3 > &	a_g_n_c_wis_strides,
		const std::array< index_t, NDimSpatial+3 > &	b_g_k_c_xs_lengths,
		const std::array< index_t, NDimSpatial+3 > &	b_g_k_c_xs_strides,
		const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > &	ds_g_n_k_wos_lengths,
		const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > &	ds_g_n_k_wos_strides,
		const std::array< index_t, NDimSpatial+3 > &	e_g_n_k_wos_lengths,
		const std::array< index_t, NDimSpatial+3 > &	e_g_n_k_wos_strides,
		const std::array< index_t, NDimSpatial > &	conv_filter_strides,
		const std::array< index_t, NDimSpatial > &	conv_filter_dilations,
		const std::array< index_t, NDimSpatial > &	input_left_pads,
		const std::array< index_t, NDimSpatial > &	input_right_pads,
		const AElementwiseOperation &	a_element_op,
		const BElementwiseOperation &	b_element_op,
		const CDEElementwiseOperation &	cde_element_op
	)

inlinestatic

◆ MakeArgument() [2/2]

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

static auto ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::MakeArgument	(	const void *	p_a,
		const void *	p_b,
		const std::array< const void *, NumDTensor > &	p_ds,
		void *	p_e,
		const std::array< long_index_t, NDimSpatial+3 > &	a_g_n_c_wis_lengths,
		const std::array< long_index_t, NDimSpatial+3 > &	a_g_n_c_wis_strides,
		const std::array< long_index_t, NDimSpatial+3 > &	b_g_k_c_xs_lengths,
		const std::array< long_index_t, NDimSpatial+3 > &	b_g_k_c_xs_strides,
		const std::array< std::array< long_index_t, NDimSpatial+3 >, NumDTensor > &	ds_g_n_k_wos_lengths,
		const std::array< std::array< long_index_t, NDimSpatial+3 >, NumDTensor > &	ds_g_n_k_wos_strides,
		const std::array< long_index_t, NDimSpatial+3 > &	e_g_n_k_wos_lengths,
		const std::array< long_index_t, NDimSpatial+3 > &	e_g_n_k_wos_strides,
		const std::array< long_index_t, NDimSpatial > &	conv_filter_strides,
		const std::array< long_index_t, NDimSpatial > &	conv_filter_dilations,
		const std::array< long_index_t, NDimSpatial > &	input_left_pads,
		const std::array< long_index_t, NDimSpatial > &	input_right_pads,
		const AElementwiseOperation &	a_element_op,
		const BElementwiseOperation &	b_element_op,
		const CDEElementwiseOperation &	cde_element_op
	)

inlinestatic

◆ MakeArgumentPointer() [1/2]

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

std::unique_ptr<BaseArgument> ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::MakeArgumentPointer	(	const void *	p_a,
		const void *	p_b,
		const std::array< const void *, NumDTensor > &	p_ds,
		void *	p_e,
		const std::array< index_t, NDimSpatial+3 > &	a_g_n_c_wis_lengths,
		const std::array< index_t, NDimSpatial+3 > &	a_g_n_c_wis_strides,
		const std::array< index_t, NDimSpatial+3 > &	b_g_k_c_xs_lengths,
		const std::array< index_t, NDimSpatial+3 > &	b_g_k_c_xs_strides,
		const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > &	ds_g_n_k_wos_lengths,
		const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > &	ds_g_n_k_wos_strides,
		const std::array< index_t, NDimSpatial+3 > &	e_g_n_k_wos_lengths,
		const std::array< index_t, NDimSpatial+3 > &	e_g_n_k_wos_strides,
		const std::array< index_t, NDimSpatial > &	conv_filter_strides,
		const std::array< index_t, NDimSpatial > &	conv_filter_dilations,
		const std::array< index_t, NDimSpatial > &	input_left_pads,
		const std::array< index_t, NDimSpatial > &	input_right_pads,
		const AElementwiseOperation &	a_element_op,
		const BElementwiseOperation &	b_element_op,
		const CDEElementwiseOperation &	cde_element_op
	)

inlineoverride

◆ MakeArgumentPointer() [2/2]

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

std::unique_ptr<BaseArgument> ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::MakeArgumentPointer	(	const void *	p_a,
		const void *	p_b,
		const std::array< const void *, NumDTensor > &	p_ds,
		void *	p_e,
		const std::array< long_index_t, NDimSpatial+3 > &	a_g_n_c_wis_lengths,
		const std::array< long_index_t, NDimSpatial+3 > &	a_g_n_c_wis_strides,
		const std::array< long_index_t, NDimSpatial+3 > &	b_g_k_c_xs_lengths,
		const std::array< long_index_t, NDimSpatial+3 > &	b_g_k_c_xs_strides,
		const std::array< std::array< long_index_t, NDimSpatial+3 >, NumDTensor > &	ds_g_n_k_wos_lengths,
		const std::array< std::array< long_index_t, NDimSpatial+3 >, NumDTensor > &	ds_g_n_k_wos_strides,
		const std::array< long_index_t, NDimSpatial+3 > &	e_g_n_k_wos_lengths,
		const std::array< long_index_t, NDimSpatial+3 > &	e_g_n_k_wos_strides,
		const std::array< long_index_t, NDimSpatial > &	conv_filter_strides,
		const std::array< long_index_t, NDimSpatial > &	conv_filter_dilations,
		const std::array< long_index_t, NDimSpatial > &	input_left_pads,
		const std::array< long_index_t, NDimSpatial > &	input_right_pads,
		const AElementwiseOperation &	a_element_op,
		const BElementwiseOperation &	b_element_op,
		const CDEElementwiseOperation &	cde_element_op
	)

inlineoverride

◆ MakeBGridDescriptor_N_K()

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

template<typename BLay >

static auto ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::MakeBGridDescriptor_N_K ( const ConvToGemmFwdTransformerIndexT & conv_to_gemm_transformer )

inlinestatic

◆ MakeEGridDescriptor_M_N()

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

template<typename ELay >

static auto ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::MakeEGridDescriptor_M_N ( const ConvToGemmFwdTransformerIndexT & conv_to_gemm_transformer )

inlinestatic

◆ MakeInvoker()

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

static auto ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::MakeInvoker ( )

inlinestatic

◆ MakeInvokerPointer()

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

std::unique_ptr<BaseInvoker> ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::MakeInvokerPointer ( )

inlineoverridevirtual

Implements ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, decltype(UnpackDataType< is_detected< is_tuple, ADataType >::value, Number< 0 >, ADataType >()), decltype(UnpackDataType< is_detected< is_tuple, ADataType >::value, Number< 0 >, ADataType >()) >.

Member Data Documentation

◆ DoElementwiseBeforeCShuffle

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

constexpr bool ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::DoElementwiseBeforeCShuffle

staticconstexpr

Initial value:

=
        NumDTensor == 0 && is_same_v<EDataType, bhalf_t> &&
        !is_same_v<CDEElementwiseOperation, tensor_operation::element_wise::PassThrough>

◆ dummy_conv_to_gemm_transformer

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

constexpr static ConvToGemmFwdTransformerIndexT ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::dummy_conv_to_gemm_transformer

staticconstexpr

◆ I0

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

constexpr auto ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::I0 = Number<0>{}

staticconstexpr

◆ I1

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

constexpr auto ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::I1 = Number<1>{}

staticconstexpr

◆ I2

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

constexpr auto ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::I2 = Number<2>{}

staticconstexpr

◆ I3

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

constexpr auto ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::I3 = Number<3>{}

staticconstexpr

◆ matrix_padder

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

constexpr auto ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::matrix_padder

staticconstexpr

Initial value:

=

MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock}

◆ MaxGemmsNum

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

constexpr index_t ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::MaxGemmsNum = 32

staticconstexpr

◆ NumDTensor

template<index_t NDimSpatial, typename ALayout , typename BLayout , typename DsLayout , typename ELayout , typename ADataType , typename BDataType , typename AccDataType , typename CShuffleDataType , typename DsDataType , typename EDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CDEElementwiseOperation , ConvolutionForwardSpecialization ConvForwardSpecialization, GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, index_t MPerXDL, index_t NPerXDL, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1 , typename ABlockTransferThreadClusterArrangeOrder , typename ABlockTransferSrcAccessOrder , index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, index_t ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CDEBlockTransferScalarPerVector_NPerBlock, typename AComputeDataType = decltype(UnpackDataType<is_detected<is_tuple, ADataType>::value, Number<0>, ADataType>()), typename BComputeDataType = AComputeDataType, LoopScheduler LoopSched = make_default_loop_scheduler()>

constexpr index_t ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD_Xdl_CShuffle_Large_Tensor< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, ConvForwardSpecialization, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, AComputeDataType, BComputeDataType, LoopSched >::NumDTensor = DsDataType::Size()

staticconstexpr

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

/home/docs/checkouts/readthedocs.org/user_builds/advanced-micro-devices-composable-kernel/checkouts/docs-7.0.1/include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_d_xdl_large_tensor_cshuffle.hpp

Classes

Public Types

Public Member Functions

Static Public Member Functions

Static Public Attributes

Member Typedef Documentation

◆ AGridDesc_AK0_M_AK1

◆ AGridDesc_M_K

◆ BGridDesc_BK0_N_BK1

◆ BGridDesc_N_K

◆ Block2ETileMap

◆ ConvToGemmFwdTransformerIndexT

◆ ConvToGemmFwdTransformerLongIndexT

◆ DeviceOp

◆ EGridDesc_M_N

◆ EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock

◆ GridwiseGemm

Member Function Documentation

◆ GenerateConvToGemmTransforms()

◆ GetTypeString()

◆ IsSupportedArgument() [1/2]

◆ IsSupportedArgument() [2/2]

◆ MakeAGridDescriptor_M_K()

◆ MakeArgument() [1/2]

◆ MakeArgument() [2/2]

◆ MakeArgumentPointer() [1/2]

◆ MakeArgumentPointer() [2/2]

◆ MakeBGridDescriptor_N_K()

◆ MakeEGridDescriptor_M_N()

◆ MakeInvoker()

◆ MakeInvokerPointer()

Member Data Documentation

◆ DoElementwiseBeforeCShuffle

◆ dummy_conv_to_gemm_transformer

◆ I0

◆ I1

◆ I2

◆ I3

◆ matrix_padder

◆ MaxGemmsNum

◆ NumDTensor