Composable Kernel: ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, > Struct Template Reference

DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, > Struct Template Reference

DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, > Struct Template Reference#

Composable Kernel: ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, > Struct Template Reference

ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, > Struct Template Reference

#include <device_batched_gemm_multiple_d_dl.hpp>

Inheritance diagram for ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >:

Classes
struct	ComputePtrOffsetOfStridedBatch
struct	Argument
struct	Invoker

Public Types
using	DeviceOp = DeviceBatchedGemmMultipleD_Dl
using	AGridDesc_K0_M_K1 = decltype(MakeAGridDescriptor_K0_M_K1(1, 1, 1))
using	BGridDesc_K0_N_K1 = decltype(MakeBGridDescriptor_K0_N_K1(1, 1, 1))
using	DsGridDesc_M_N = decltype(MakeDsGridDescriptor_M_N({}, {}, {}))
using	EGridDesc_M_N = decltype(MakeEGridDescriptor_M_N<ELayout>(1, 1, 1))
using	GridwiseGemm
using	AGridDesc_K0_M0_M1_K1
using	BGridDesc_K0_N0_N1_K1
using	DsGridDesc_M0_M10_M11_N0_N10_N11
using	EGridDesc_M0_M10_M11_N0_N10_N11
using	DefaultBlock2CTileMap

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, index_t M, index_t N, index_t K, index_t Batch, index_t StrideA, index_t StrideB, const std::array< ck::index_t, NumDTensor > &StrideDs, index_t StrideE, index_t BatchStrideA, index_t BatchStrideB, const std::array< ck::index_t, NumDTensor > &BatchStrideDs, index_t BatchStrideE, AElementwiseOperation a_element_op, BElementwiseOperation b_element_op, 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::BaseOperator
	BaseOperator ()=default
	BaseOperator (const BaseOperator &)=default
BaseOperator &	operator= (const BaseOperator &)=default
virtual std::string	GetInstanceString () const
virtual std::string	GetTypeIdName () const
virtual std::optional< std::string >	GetObjectName () const
virtual std::optional< std::string >	GetTemplateInfo () const
virtual std::string	GetTypeIdHashCode () const
virtual size_t	GetWorkSpaceSize (const BaseArgument *) const
virtual void	SetWorkSpacePointer (BaseArgument p_arg, void p_workspace, const StreamConfig &=StreamConfig{}) const
virtual	~BaseOperator ()

Static Public Member Functions
static auto	MakeAGridDescriptor_K0_M_K1 (index_t M, index_t K, index_t StrideA)
static auto	MakeBGridDescriptor_K0_N_K1 (index_t K, index_t N, index_t StrideB)
template<typename ELay>
static auto	MakeEGridDescriptor_M_N (index_t M, index_t N, index_t StrideE)
static auto	MakeDsGridDescriptor_M_N (const std::array< index_t, NumDTensor > &MRaws, const std::array< index_t, NumDTensor > &NRaws, const std::array< index_t, NumDTensor > &DsStride)
static constexpr bool	IsValidCompilationParameter ()
static bool	IsSupportedArgument (const Argument &arg)
static auto	MakeArgument (const void p_a, const void p_b, std::array< const void , NumDTensor > p_ds, void p_e, index_t M, index_t N, index_t K, index_t Batch, index_t StrideA, index_t StrideB, std::array< ck::index_t, NumDTensor > StrideDs, index_t StrideE, index_t BatchStrideA, index_t BatchStrideB, const std::array< ck::index_t, NumDTensor > &BatchStrideDs, index_t BatchStrideE, AElementwiseOperation a_element_op, BElementwiseOperation b_element_op, CDEElementwiseOperation cde_element_op)
static auto	MakeInvoker ()

Static Public Attributes
static constexpr index_t	NumDTensor = DsDataType::Size()
static constexpr auto	I0 = Number<0>{}
static constexpr auto	I1 = Number<1>{}
static constexpr auto	I2 = Number<2>{}
static constexpr auto	I3 = Number<3>{}
static constexpr auto	I4 = Number<4>{}
static constexpr auto	I5 = Number<5>{}
static constexpr auto	K1Number = Number<K1>{}
Static Public Attributes inherited from ck::tensor_operation::device::DeviceBatchedGemmMultiD< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation >
static constexpr index_t	NumDTensor = DsDataType::Size()

Member Typedef Documentation

◆ AGridDesc_K0_M0_M1_K1

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::AGridDesc_K0_M0_M1_K1

Initial value:

decltype(GridwiseGemm::MakeAGridDescriptor_K0_M0_M1_K1(AGridDesc_K0_M_K1{}))

ck::GridwiseGemmDlMultipleD_km_kn_mn< BlockSize, ADataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, InMemoryDataOperationEnum::Set, AGridDesc_K0_M_K1, BGridDesc_K0_N_K1, EGridDesc_M_N, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector >::MakeAGridDescriptor_K0_M0_M1_K1

__host__ static __device__ constexpr auto MakeAGridDescriptor_K0_M0_M1_K1(const AGridDesc_K0_M_K1 &a_grid_desc_k0_m_k1)

Definition gridwise_gemm_dl_multiple_d.hpp:158

ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl::AGridDesc_K0_M_K1

decltype(MakeAGridDescriptor_K0_M_K1(1, 1, 1)) AGridDesc_K0_M_K1

Definition device_batched_gemm_multiple_d_dl.hpp:324

◆ AGridDesc_K0_M_K1

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::AGridDesc_K0_M_K1 = decltype(MakeAGridDescriptor_K0_M_K1(1, 1, 1))

◆ BGridDesc_K0_N0_N1_K1

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::BGridDesc_K0_N0_N1_K1

Initial value:

decltype(GridwiseGemm::MakeBGridDescriptor_K0_N0_N1_K1(BGridDesc_K0_N_K1{}))

ck::GridwiseGemmDlMultipleD_km_kn_mn< BlockSize, ADataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, InMemoryDataOperationEnum::Set, AGridDesc_K0_M_K1, BGridDesc_K0_N_K1, EGridDesc_M_N, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector >::MakeBGridDescriptor_K0_N0_N1_K1

__host__ static __device__ constexpr auto MakeBGridDescriptor_K0_N0_N1_K1(const BGridDesc_K0_N_K1 &b_grid_desc_k0_n_k1)

Definition gridwise_gemm_dl_multiple_d.hpp:178

ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl::BGridDesc_K0_N_K1

decltype(MakeBGridDescriptor_K0_N_K1(1, 1, 1)) BGridDesc_K0_N_K1

Definition device_batched_gemm_multiple_d_dl.hpp:325

◆ BGridDesc_K0_N_K1

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::BGridDesc_K0_N_K1 = decltype(MakeBGridDescriptor_K0_N_K1(1, 1, 1))

◆ DefaultBlock2CTileMap

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::DefaultBlock2CTileMap

Initial value:

decltype(GridwiseGemm::MakeDefaultBlock2CTileMap(EGridDesc_M_N{}))

ck::GridwiseGemmDlMultipleD_km_kn_mn< BlockSize, ADataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, InMemoryDataOperationEnum::Set, AGridDesc_K0_M_K1, BGridDesc_K0_N_K1, EGridDesc_M_N, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector >::MakeDefaultBlock2CTileMap

__host__ static __device__ constexpr auto MakeDefaultBlock2CTileMap(const EGridDesc_M_N &c_grid_desc_m_n)

Definition gridwise_gemm_dl_multiple_d.hpp:242

ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl::EGridDesc_M_N

decltype(MakeEGridDescriptor_M_N< ELayout >(1, 1, 1)) EGridDesc_M_N

Definition device_batched_gemm_multiple_d_dl.hpp:327

◆ DeviceOp

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::DeviceOp = DeviceBatchedGemmMultipleD_Dl

◆ DsGridDesc_M0_M10_M11_N0_N10_N11

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::DsGridDesc_M0_M10_M11_N0_N10_N11

Initial value:

decltype(GridwiseGemm::MakeDsGridDescriptor_M0_M10_M11_N0_N10_N11(DsGridDesc_M_N{}))

ck::GridwiseGemmDlMultipleD_km_kn_mn< BlockSize, ADataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, InMemoryDataOperationEnum::Set, AGridDesc_K0_M_K1, BGridDesc_K0_N_K1, EGridDesc_M_N, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector >::MakeDsGridDescriptor_M0_M10_M11_N0_N10_N11

__host__ static __device__ constexpr auto MakeDsGridDescriptor_M0_M10_M11_N0_N10_N11(const DsGridDesc_M_N &ds_grid_desc_m_n)

Definition gridwise_gemm_dl_multiple_d.hpp:234

ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl::DsGridDesc_M_N

decltype(MakeDsGridDescriptor_M_N({}, {}, {})) DsGridDesc_M_N

Definition device_batched_gemm_multiple_d_dl.hpp:326

◆ DsGridDesc_M_N

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::DsGridDesc_M_N = decltype(MakeDsGridDescriptor_M_N({}, {}, {}))

◆ EGridDesc_M0_M10_M11_N0_N10_N11

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::EGridDesc_M0_M10_M11_N0_N10_N11

Initial value:

decltype(GridwiseGemm::MakeCGridDescriptor_M0_M10_M11_N0_N10_N11(EGridDesc_M_N{}))

ck::GridwiseGemmDlMultipleD_km_kn_mn< BlockSize, ADataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, InMemoryDataOperationEnum::Set, AGridDesc_K0_M_K1, BGridDesc_K0_N_K1, EGridDesc_M_N, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector >::MakeCGridDescriptor_M0_M10_M11_N0_N10_N11

__host__ static __device__ constexpr auto MakeCGridDescriptor_M0_M10_M11_N0_N10_N11(const CGridDesc_M_N_ &c_grid_desc_m_n)

Definition gridwise_gemm_dl_multiple_d.hpp:200

◆ EGridDesc_M_N

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::EGridDesc_M_N = decltype(MakeEGridDescriptor_M_N<ELayout>(1, 1, 1))

◆ GridwiseGemm

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::GridwiseGemm

Member Function Documentation

◆ GetTypeString()

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

std::string ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::GetTypeString ( ) const

inlineoverridevirtual

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

◆ IsSupportedArgument() [1/2]

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

bool ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::IsSupportedArgument ( const Argument & arg )

inlinestatic

◆ IsSupportedArgument() [2/2]

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

bool ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::IsSupportedArgument ( const BaseArgument * p_arg )

inlineoverridevirtual

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

◆ IsValidCompilationParameter()

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

constexpr bool ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::IsValidCompilationParameter ( )

inlinestaticconstexpr

◆ MakeAGridDescriptor_K0_M_K1()

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::MakeAGridDescriptor_K0_M_K1	(	index_t	M,
		index_t	K,
		index_t	StrideA )

inlinestatic

◆ MakeArgument()

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::MakeArgument	(	const void *	p_a,
		const void *	p_b,
		std::array< const void *, NumDTensor >	p_ds,
		void *	p_e,
		index_t	M,
		index_t	N,
		index_t	K,
		index_t	Batch,
		index_t	StrideA,
		index_t	StrideB,
		std::array< ck::index_t, NumDTensor >	StrideDs,
		index_t	StrideE,
		index_t	BatchStrideA,
		index_t	BatchStrideB,
		const std::array< ck::index_t, NumDTensor > &	BatchStrideDs,
		index_t	BatchStrideE,
		AElementwiseOperation	a_element_op,
		BElementwiseOperation	b_element_op,
		CDEElementwiseOperation	cde_element_op )

inlinestatic

◆ MakeArgumentPointer()

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

std::unique_ptr< BaseArgument > ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::MakeArgumentPointer	(	const void *	p_a,
		const void *	p_b,
		const std::array< const void *, NumDTensor > &	p_ds,
		void *	p_e,
		index_t	M,
		index_t	N,
		index_t	K,
		index_t	Batch,
		index_t	StrideA,
		index_t	StrideB,
		const std::array< ck::index_t, NumDTensor > &	StrideDs,
		index_t	StrideE,
		index_t	BatchStrideA,
		index_t	BatchStrideB,
		const std::array< ck::index_t, NumDTensor > &	BatchStrideDs,
		index_t	BatchStrideE,
		AElementwiseOperation	a_element_op,
		BElementwiseOperation	b_element_op,
		CDEElementwiseOperation	cde_element_op )

inlineoverridevirtual

Implements ck::tensor_operation::device::DeviceBatchedGemmMultiD< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation >.

◆ MakeBGridDescriptor_K0_N_K1()

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::MakeBGridDescriptor_K0_N_K1	(	index_t	K,
		index_t	N,
		index_t	StrideB )

inlinestatic

◆ MakeDsGridDescriptor_M_N()

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::MakeDsGridDescriptor_M_N	(	const std::array< index_t, NumDTensor > &	MRaws,
		const std::array< index_t, NumDTensor > &	NRaws,
		const std::array< index_t, NumDTensor > &	DsStride )

inlinestatic

◆ MakeEGridDescriptor_M_N()

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

template<typename ELay>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::MakeEGridDescriptor_M_N	(	index_t	M,
		index_t	N,
		index_t	StrideE )

inlinestatic

◆ MakeInvoker()

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::MakeInvoker ( )

inlinestatic

◆ MakeInvokerPointer()

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

std::unique_ptr< BaseInvoker > ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::MakeInvokerPointer ( )

inlineoverridevirtual

Implements ck::tensor_operation::device::DeviceBatchedGemmMultiD< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation >.

Member Data Documentation

◆ I0

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::I0 = Number<0>{}

staticconstexpr

◆ I1

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::I1 = Number<1>{}

staticconstexpr

◆ I2

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::I2 = Number<2>{}

staticconstexpr

◆ I3

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::I3 = Number<3>{}

staticconstexpr

◆ I4

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::I4 = Number<4>{}

staticconstexpr

◆ I5

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::I5 = Number<5>{}

staticconstexpr

◆ K1Number

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::K1Number = Number<K1>{}

staticconstexpr

◆ NumDTensor

template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t K1, index_t M1PerThread, index_t N1PerThread, index_t KPerThread, typename M1N1ThreadClusterM1Xs, typename M1N1ThreadClusterN1Xs, typename ABlockTransferThreadSliceLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterLengths_K0_M0_M1_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, typename ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, typename ABlockTransferSrcVectorTensorContiguousDimOrder, typename ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, typename BBlockTransferThreadSliceLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterLengths_K0_N0_N1_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, typename BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, typename BBlockTransferSrcVectorTensorContiguousDimOrder, typename BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, typename CThreadTransferSrcDstAccessOrder, index_t CThreadTransferSrcDstVectorDim, index_t CThreadTransferDstScalarPerVector, enable_if_t< is_same_v< AElementwiseOperation, ck::tensor_operation::element_wise::PassThrough > &&is_same_v< BElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

index_t ck::tensor_operation::device::DeviceBatchedGemmMultipleD_Dl< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::NumDTensor = DsDataType::Size()

staticconstexpr

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

device_batched_gemm_multiple_d_dl.hpp

Classes

Public Types

Public Member Functions

Static Public Member Functions

Static Public Attributes

Member Typedef Documentation

◆ AGridDesc_K0_M0_M1_K1

◆ AGridDesc_K0_M_K1

◆ BGridDesc_K0_N0_N1_K1

◆ BGridDesc_K0_N_K1

◆ DefaultBlock2CTileMap

◆ DeviceOp

◆ DsGridDesc_M0_M10_M11_N0_N10_N11

◆ DsGridDesc_M_N

◆ EGridDesc_M0_M10_M11_N0_N10_N11

◆ EGridDesc_M_N

◆ GridwiseGemm

Member Function Documentation

◆ GetTypeString()

◆ IsSupportedArgument() [1/2]

◆ IsSupportedArgument() [2/2]

◆ IsValidCompilationParameter()

◆ MakeAGridDescriptor_K0_M_K1()

◆ MakeArgument()

◆ MakeArgumentPointer()

◆ MakeBGridDescriptor_K0_N_K1()

◆ MakeDsGridDescriptor_M_N()

◆ MakeEGridDescriptor_M_N()

◆ MakeInvoker()

◆ MakeInvokerPointer()

Member Data Documentation

◆ I0

◆ I1

◆ I2

◆ I3

◆ I4

◆ I5

◆ K1Number

◆ NumDTensor