DeviceGroupedGemmMultipleD_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
• DeviceGroupedGemmMultipleD_Dl

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

ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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_grouped_gemm_multiple_d_dl.hpp>

Inheritance diagram for ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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	Argument
struct	GemmKernelArg
struct	GroupedGemmBlock2ETileMap
struct	Invoker

Public Types
using	DeviceOp = DeviceGroupedGemmMultipleD_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 = 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 >
using	AGridDesc_K0_M0_M1_K1 = decltype(GridwiseGemm::MakeAGridDescriptor_K0_M0_M1_K1(AGridDesc_K0_M_K1{}))
using	BGridDesc_K0_N0_N1_K1 = decltype(GridwiseGemm::MakeBGridDescriptor_K0_N0_N1_K1(BGridDesc_K0_N_K1{}))
using	DsGridDesc_M0_M10_M11_N0_N10_N11 = decltype(GridwiseGemm::MakeDsGridDescriptor_M0_M10_M11_N0_N10_N11(DsGridDesc_M_N{}))
using	EGridDesc_M0_M10_M11_N0_N10_N11 = decltype(GridwiseGemm::MakeCGridDescriptor_M0_M10_M11_N0_N10_N11(EGridDesc_M_N{}))

Public Member Functions
bool	IsSupportedArgument (const BaseArgument *p_arg) override
std::unique_ptr< BaseArgument >	MakeArgumentPointer (std::vector< const void * > &p_As, std::vector< const void * > &p_Bs, std::vector< std::array< const void *, NumDTensor >> &p_Ds, std::vector< void * > &p_Es, std::vector< GemmDesc > &gemm_descs, AElementwiseOperation a_element_op, BElementwiseOperation b_element_op, CDEElementwiseOperation cde_element_op) override
std::unique_ptr< BaseInvoker >	MakeInvokerPointer () override
std::string	GetTypeString () const override
size_t	GetWorkSpaceSize (const BaseArgument *p_arg) const override
size_t	GetDeviceKernelArgSize (const BaseArgument *p_arg) const override
	Gets the device kernel argument size. More...
size_t	GetHostKernelArgSize (const BaseArgument *p_arg) const
void	SetDeviceKernelArgs (BaseArgument *p_arg, void *p_dev_kernel_args) const override
	Sets the device kernel arguments pointer and may copy data to device. More...
void	SetHostKernelArgsPointer (BaseArgument *p_arg, void *p_host_kernel_args) const
	Sets the host kernel arguments pointer and copies that data on the host side. This function can be utilised to use pinned memory for the host args and achieve fully async data copy. More...
Public Member Functions inherited from ck::tensor_operation::device::DeviceGroupedGemm< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation >
virtual void	SetDeviceKernelArgs (BaseArgument *p_arg, void *p_dev_kernel_args, const void *p_host_kernel_args) const
	Sets the device kernel arguments pointer and may copy data to device. More...
Public Member Functions inherited from ck::tensor_operation::device::BaseOperator
	BaseOperator ()=default
	BaseOperator (const BaseOperator &)=default
BaseOperator &	operator= (const BaseOperator &)=default
virtual std::string	GetTypeIdName () const
virtual std::optional< std::string >	GetObjectName () const
virtual std::optional< std::string >	GetTemplateInfo () const
virtual std::string	GetTypeIdHashCode () const
virtual 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 bool	IsSupportedArgument (const Argument &arg)
static auto	MakeArgument (std::vector< const void * > &p_As, std::vector< const void * > &p_Bs, std::vector< std::array< const void *, NumDTensor >> &p_Ds, std::vector< void * > &p_Es, std::vector< GemmDesc > gemm_descs, 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::DeviceGroupedGemm< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation >
static constexpr index_t	NumDTensor

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

◆ 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::DeviceGroupedGemmMultipleD_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::DeviceGroupedGemmMultipleD_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 = decltype(GridwiseGemm::MakeBGridDescriptor_K0_N0_N1_K1(BGridDesc_K0_N_K1{}))

◆ 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::DeviceGroupedGemmMultipleD_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))

◆ 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::DeviceGroupedGemmMultipleD_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 = DeviceGroupedGemmMultipleD_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::DeviceGroupedGemmMultipleD_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 = decltype(GridwiseGemm::MakeDsGridDescriptor_M0_M10_M11_N0_N10_N11(DsGridDesc_M_N{}))

◆ 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::DeviceGroupedGemmMultipleD_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::DeviceGroupedGemmMultipleD_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 = decltype(GridwiseGemm::MakeCGridDescriptor_M0_M10_M11_N0_N10_N11(EGridDesc_M_N{}))

◆ 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::DeviceGroupedGemmMultipleD_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::DeviceGroupedGemmMultipleD_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 = 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>

Member Function Documentation

◆ GetDeviceKernelArgSize()

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>

size_t ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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, >::GetDeviceKernelArgSize ( const BaseArgument *  p_arg ) const

inlineoverridevirtual

Gets the device kernel argument size.

Parameters

[in]

p_arg

The pointer to the Device op Argument.

Returns

The device kernel argument size.

Reimplemented from ck::tensor_operation::device::DeviceGroupedGemm< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation >.

◆ GetHostKernelArgSize()

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>

size_t ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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, >::GetHostKernelArgSize ( const BaseArgument *  p_arg ) const

inline

◆ 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::DeviceGroupedGemmMultipleD_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.

◆ GetWorkSpaceSize()

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>

size_t ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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, >::GetWorkSpaceSize ( const BaseArgument *  p_arg ) 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>

static bool ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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::DeviceGroupedGemmMultipleD_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.

◆ 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>

static auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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>

static auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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 ( std::vector< const void * > &  p_As, std::vector< const void * > &  p_Bs, std::vector< std::array< const void *, NumDTensor >> &  p_Ds, std::vector< void * > &  p_Es, std::vector< GemmDesc >  gemm_descs, 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::DeviceGroupedGemmMultipleD_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 ( std::vector< const void * > &  p_As, std::vector< const void * > &  p_Bs, std::vector< std::array< const void *, NumDTensor >> &  p_Ds, std::vector< void * > &  p_Es, std::vector< GemmDesc > &  gemm_descs, AElementwiseOperation  a_element_op, BElementwiseOperation  b_element_op, CDEElementwiseOperation  cde_element_op  )

inlineoverridevirtual

Implements ck::tensor_operation::device::DeviceGroupedGemm< 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>

static auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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>

static auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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 >

static auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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>

static auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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::DeviceGroupedGemmMultipleD_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::DeviceGroupedGemm< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation >.

◆ SetDeviceKernelArgs()

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>

void ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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, >::SetDeviceKernelArgs ( BaseArgument *  p_arg, void *  p_dev_kernel_args  ) const

inlineoverridevirtual

Sets the device kernel arguments pointer and may copy data to device.

Parameters

	p_arg	The pointer to the Argument we're going to update.
[in]	p_dev_kernel_args	The pointer to the device memory which contains kernel arguments.

Reimplemented from ck::tensor_operation::device::DeviceGroupedGemm< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation >.

◆ SetHostKernelArgsPointer()

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>

void ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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, >::SetHostKernelArgsPointer ( BaseArgument *  p_arg, void *  p_host_kernel_args  ) const

inline

Sets the host kernel arguments pointer and copies that data on the host side. This function can be utilised to use pinned memory for the host args and achieve fully async data copy.

Parameters

	p_arg	The pointer to the Argument we're going to update.
[in]	p_host_kernel_args	The pointer to the host memory where the kernel arguments will be copied

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>

constexpr auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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>

constexpr auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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>

constexpr auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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>

constexpr auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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>

constexpr auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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>

constexpr auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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>

constexpr auto ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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>

constexpr index_t ck::tensor_operation::device::DeviceGroupedGemmMultipleD_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:

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