DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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
• DeviceGemmDl

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

ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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_gemm_dl.hpp>

Inheritance diagram for ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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	Invoker

Public Types
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	CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1))
using	GridwiseGemm
using	AGridDesc_K0_M0_M1_K1
using	BGridDesc_K0_N0_N1_K1
using	CGridDesc_M0_M10_M11_N0_N10_N11
using	DefaultBlock2CTileMap

Public Member Functions
bool	IsSupportedArgument (const BaseArgument *p_arg) override
std::unique_ptr< BaseArgument >	MakeArgumentPointer (const void *p_a, const void *p_b, void *p_c, index_t M, index_t N, index_t K, index_t StrideA, index_t StrideB, index_t StrideC, AElementwiseOperation a_element_op, BElementwiseOperation b_element_op, CElementwiseOperation c_element_op) override
std::unique_ptr< BaseInvoker >	MakeInvokerPointer () override
virtual std::string	GetTypeString () const override
Public Member Functions inherited from ck::tensor_operation::device::BaseOperator
	BaseOperator ()=default
	BaseOperator (const BaseOperator &)=default
BaseOperator &	operator= (const BaseOperator &)=default
virtual std::string	GetInstanceString () const
virtual std::string	GetTypeIdName () const
virtual std::optional< std::string >	GetObjectName () const
virtual std::optional< std::string >	GetTemplateInfo () const
virtual std::string	GetTypeIdHashCode () const
virtual size_t	GetWorkSpaceSize (const BaseArgument *) const
virtual void	SetWorkSpacePointer (BaseArgument *p_arg, void *p_workspace, const StreamConfig &=StreamConfig{}) const
virtual	~BaseOperator ()

Static Public Member Functions
static auto	MakeAGridDescriptor_K0_M_K1 (index_t M, index_t K, index_t StrideA)
static auto	MakeBGridDescriptor_K0_N_K1 (index_t K, index_t N, index_t StrideB)
static auto	MakeCGridDescriptor_M_N (index_t M, index_t N, index_t StrideC)
static constexpr bool	IsValidCompilationParameter ()
static bool	IsSupportedArgument (const Argument &arg)
static auto	MakeArgument (const ADataType *p_a, const BDataType *p_b, CDataType *p_c, index_t M, index_t N, index_t K, index_t StrideA, index_t StrideB, index_t StrideC, AElementwiseOperation a_element_op, BElementwiseOperation b_element_op, CElementwiseOperation c_element_op)
static auto	MakeInvoker ()

Static Public Attributes
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>{}

Member Typedef Documentation

◆ AGridDesc_K0_M0_M1_K1

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

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

Initial value:

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

ck::GridwiseGemmDl_km_kn_mn_v1r3< BlockSize, ADataType, AccDataType, CDataType, InMemoryDataOperationEnum::Set, AGridDesc_K0_M_K1, BGridDesc_K0_N_K1, CGridDesc_M_N, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector >::MakeAGridDescriptor_K0_M0_M1_K1

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

Definition gridwise_gemm_dl_v1r3.hpp:168

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

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

Definition device_batched_gemm_multiple_d_dl.hpp:324

◆ AGridDesc_K0_M_K1

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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 ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

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

Initial value:

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

ck::GridwiseGemmDl_km_kn_mn_v1r3< BlockSize, ADataType, AccDataType, CDataType, InMemoryDataOperationEnum::Set, AGridDesc_K0_M_K1, BGridDesc_K0_N_K1, CGridDesc_M_N, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector >::MakeBGridDescriptor_K0_N0_N1_K1

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

Definition gridwise_gemm_dl_v1r3.hpp:188

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

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

Definition device_batched_gemm_multiple_d_dl.hpp:325

◆ BGridDesc_K0_N_K1

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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))

◆ CGridDesc_M0_M10_M11_N0_N10_N11

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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, >::CGridDesc_M0_M10_M11_N0_N10_N11

Initial value:

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

ck::GridwiseGemmDl_km_kn_mn_v1r3< BlockSize, ADataType, AccDataType, CDataType, InMemoryDataOperationEnum::Set, AGridDesc_K0_M_K1, BGridDesc_K0_N_K1, CGridDesc_M_N, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector >::MakeCGridDescriptor_M0_M10_M11_N0_N10_N11

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

Definition gridwise_gemm_dl_v1r3.hpp:208

ck::tensor_operation::device::DeviceGemmDl::CGridDesc_M_N

decltype(MakeCGridDescriptor_M_N(1, 1, 1)) CGridDesc_M_N

Definition device_gemm_dl.hpp:203

◆ CGridDesc_M_N

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

using ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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, >::CGridDesc_M_N = decltype(MakeCGridDescriptor_M_N(1, 1, 1))

◆ DefaultBlock2CTileMap

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

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

Initial value:

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

ck::GridwiseGemmDl_km_kn_mn_v1r3< BlockSize, ADataType, AccDataType, CDataType, InMemoryDataOperationEnum::Set, AGridDesc_K0_M_K1, BGridDesc_K0_N_K1, CGridDesc_M_N, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector >::MakeDefaultBlock2CTileMap

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

Definition gridwise_gemm_dl_v1r3.hpp:241

◆ GridwiseGemm

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

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

Member Function Documentation

◆ GetTypeString()

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

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

inlineoverridevirtual

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

◆ IsSupportedArgument() [1/2]

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

bool ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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 ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

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

inlineoverridevirtual

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

◆ IsValidCompilationParameter()

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

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

inlinestaticconstexpr

◆ MakeAGridDescriptor_K0_M_K1()

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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 ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::MakeArgument ( const ADataType * p_a, const BDataType * p_b, CDataType * p_c, index_t M, index_t N, index_t K, index_t StrideA, index_t StrideB, index_t StrideC, AElementwiseOperation a_element_op, BElementwiseOperation b_element_op, CElementwiseOperation c_element_op )

inlinestatic

◆ MakeArgumentPointer()

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

std::unique_ptr< BaseArgument > ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, M1PerThread, N1PerThread, KPerThread, M1N1ThreadClusterM1Xs, M1N1ThreadClusterN1Xs, ABlockTransferThreadSliceLengths_K0_M0_M1_K1, ABlockTransferThreadClusterLengths_K0_M0_M1_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorTensorLengths_K0_M0_M1_K1, ABlockTransferSrcVectorTensorContiguousDimOrder, ABlockTransferDstVectorTensorLengths_K0_M0_M1_K1, BBlockTransferThreadSliceLengths_K0_N0_N1_K1, BBlockTransferThreadClusterLengths_K0_N0_N1_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorTensorLengths_K0_N0_N1_K1, BBlockTransferSrcVectorTensorContiguousDimOrder, BBlockTransferDstVectorTensorLengths_K0_N0_N1_K1, CThreadTransferSrcDstAccessOrder, CThreadTransferSrcDstVectorDim, CThreadTransferDstScalarPerVector, >::MakeArgumentPointer ( const void * p_a, const void * p_b, void * p_c, index_t M, index_t N, index_t K, index_t StrideA, index_t StrideB, index_t StrideC, AElementwiseOperation a_element_op, BElementwiseOperation b_element_op, CElementwiseOperation c_element_op )

inlineoverridevirtual

Implements ck::tensor_operation::device::DeviceGemm< ALayout, BLayout, CLayout, ADataType, BDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation >.

◆ MakeBGridDescriptor_K0_N_K1()

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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

◆ MakeCGridDescriptor_M_N()

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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, >::MakeCGridDescriptor_M_N ( index_t M, index_t N, index_t StrideC )

inlinestatic

◆ MakeInvoker()

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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 ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

std::unique_ptr< BaseInvoker > ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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::DeviceGemm< ALayout, BLayout, CLayout, ADataType, BDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation >.

Member Data Documentation

◆ I0

template<typename ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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 ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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 ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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 ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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 ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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 ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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 ADataType, typename BDataType, typename CDataType, typename AccDataType, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, 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 > &&is_same_v< CElementwiseOperation, ck::tensor_operation::element_wise::PassThrough >, bool > = false>

auto ck::tensor_operation::device::DeviceGemmDl< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, 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

---

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

• device_gemm_dl.hpp