Composable Kernel: ck::tensor_operation::device::DeviceGemmXdlSplitKCShuffle_LdsDirectLoad< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferScalarPerVector, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferScalarPerVector, BBlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CBlockTransferScalarPerVector_NWaveNPerXDL, ComputeType, PipelineVer, LoopSched >::Argument Struct Reference

Argument Struct Reference

Argument Struct Reference#

Composable Kernel: ck::tensor_operation::device::DeviceGemmXdlSplitKCShuffle_LdsDirectLoad< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferScalarPerVector, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferScalarPerVector, BBlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CBlockTransferScalarPerVector_NWaveNPerXDL, ComputeType, PipelineVer, LoopSched >::Argument Struct Reference

Public Member Functions | Public Attributes | List of all members

ck::tensor_operation::device::DeviceGemmXdlSplitKCShuffle_LdsDirectLoad< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferScalarPerVector, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferScalarPerVector, BBlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CBlockTransferScalarPerVector_NWaveNPerXDL, ComputeType, PipelineVer, LoopSched >::Argument Struct Reference

#include <device_gemm_xdl_splitk_c_shuffle_lds_direct_load.hpp>

Inheritance diagram for ck::tensor_operation::device::DeviceGemmXdlSplitKCShuffle_LdsDirectLoad< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferScalarPerVector, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferScalarPerVector, BBlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CBlockTransferScalarPerVector_NWaveNPerXDL, ComputeType, PipelineVer, LoopSched >::Argument:

Public Member Functions
	Argument (const ADataType p_a_grid_, const BDataType p_b_grid_, CDataType *p_c_grid_, index_t M_, index_t N_, index_t K_, index_t StrideA_, index_t StrideB_, index_t StrideC_, index_t MPadded_, index_t NPadded_, index_t KPadded_, index_t K0Padded_, index_t k_batch_, AElementwiseOperation a_element_op_, BElementwiseOperation b_element_op_, CElementwiseOperation c_element_op_)

Public Member Functions inherited from ck::GridwiseGemm_xdlops_splitk_lds_direct_load< BlockSize, FloatA, FloatB, FloatAcc, FloatC, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, MPerBlock, NPerBlock, K0PerBlock, MPerXDL, NPerXDL, K1Value, MRepeat, NRepeat, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockLdsExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CBlockTransferScalarPerVector_NWaveNPerXDL, CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopSched, PipelineVer, ComputeType >::Argument
	Argument (const FloatA p_a_grid_, const FloatB p_b_grid_, FloatC *p_c_grid_, index_t M_, index_t N_, index_t K_, index_t StrideA_, index_t StrideB_, index_t StrideC_, index_t MPadded_, index_t NPadded_, index_t KPadded_, index_t K0Padded_, index_t k_batch_)

void	Print () const

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

	BaseArgument (const BaseArgument &)=default

BaseArgument &	operator= (const BaseArgument &)=default

virtual	~BaseArgument ()

Public Attributes
AElementwiseOperation	a_element_op

BElementwiseOperation	b_element_op

CElementwiseOperation	c_element_op

Public Attributes inherited from ck::GridwiseGemm_xdlops_splitk_lds_direct_load< BlockSize, FloatA, FloatB, FloatAcc, FloatC, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, MPerBlock, NPerBlock, K0PerBlock, MPerXDL, NPerXDL, K1Value, MRepeat, NRepeat, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockLdsExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CBlockTransferScalarPerVector_NWaveNPerXDL, CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopSched, PipelineVer, ComputeType >::Argument
const FloatA *	p_a_grid

const FloatB *	p_b_grid

FloatC *	p_c_grid

index_t	M

index_t	N

index_t	K

index_t	StrideA

index_t	StrideB

index_t	StrideC

index_t	MPadded

index_t	NPadded

index_t	KPadded

index_t	K0Padded

index_t	k_batch

Public Attributes inherited from ck::tensor_operation::device::BaseArgument
void *	p_workspace_ = nullptr

Constructor & Destructor Documentation

◆ Argument()

template<typename ADataType , typename BDataType , typename CDataType , typename AccDataType , typename ALayout , typename BLayout , typename CLayout , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, ck::index_t NumGemmKPrefetchStage, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1 , typename ABlockTransferSrcAccessOrder , ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferScalarPerVector, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1 , typename BBlockTransferSrcAccessOrder , ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferScalarPerVector, bool BBlockLdsAddExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename ComputeType = CDataType, PipelineVersion PipelineVer = PipelineVersion::v4, LoopScheduler LoopSched = make_default_loop_scheduler()>

ck::tensor_operation::device::DeviceGemmXdlSplitKCShuffle_LdsDirectLoad< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferScalarPerVector, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferScalarPerVector, BBlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CBlockTransferScalarPerVector_NWaveNPerXDL, ComputeType, PipelineVer, LoopSched >::Argument::Argument	(	const ADataType *	p_a_grid_,
		const BDataType *	p_b_grid_,
		CDataType *	p_c_grid_,
		index_t	M_,
		index_t	N_,
		index_t	K_,
		index_t	StrideA_,
		index_t	StrideB_,
		index_t	StrideC_,
		index_t	MPadded_,
		index_t	NPadded_,
		index_t	KPadded_,
		index_t	K0Padded_,
		index_t	k_batch_,
		AElementwiseOperation	a_element_op_,
		BElementwiseOperation	b_element_op_,
		CElementwiseOperation	c_element_op_
	)

inline

Member Data Documentation

◆ a_element_op

template<typename ADataType , typename BDataType , typename CDataType , typename AccDataType , typename ALayout , typename BLayout , typename CLayout , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, ck::index_t NumGemmKPrefetchStage, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1 , typename ABlockTransferSrcAccessOrder , ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferScalarPerVector, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1 , typename BBlockTransferSrcAccessOrder , ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferScalarPerVector, bool BBlockLdsAddExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename ComputeType = CDataType, PipelineVersion PipelineVer = PipelineVersion::v4, LoopScheduler LoopSched = make_default_loop_scheduler()>

AElementwiseOperation ck::tensor_operation::device::DeviceGemmXdlSplitKCShuffle_LdsDirectLoad< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferScalarPerVector, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferScalarPerVector, BBlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CBlockTransferScalarPerVector_NWaveNPerXDL, ComputeType, PipelineVer, LoopSched >::Argument::a_element_op

◆ b_element_op

template<typename ADataType , typename BDataType , typename CDataType , typename AccDataType , typename ALayout , typename BLayout , typename CLayout , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, ck::index_t NumGemmKPrefetchStage, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1 , typename ABlockTransferSrcAccessOrder , ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferScalarPerVector, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1 , typename BBlockTransferSrcAccessOrder , ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferScalarPerVector, bool BBlockLdsAddExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename ComputeType = CDataType, PipelineVersion PipelineVer = PipelineVersion::v4, LoopScheduler LoopSched = make_default_loop_scheduler()>

BElementwiseOperation ck::tensor_operation::device::DeviceGemmXdlSplitKCShuffle_LdsDirectLoad< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferScalarPerVector, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferScalarPerVector, BBlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CBlockTransferScalarPerVector_NWaveNPerXDL, ComputeType, PipelineVer, LoopSched >::Argument::b_element_op

◆ c_element_op

template<typename ADataType , typename BDataType , typename CDataType , typename AccDataType , typename ALayout , typename BLayout , typename CLayout , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, ck::index_t NumGemmKPrefetchStage, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t K0PerBlock, ck::index_t K1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1 , typename ABlockTransferSrcAccessOrder , ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferScalarPerVector, bool ABlockLdsAddExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1 , typename BBlockTransferSrcAccessOrder , ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferScalarPerVector, bool BBlockLdsAddExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename ComputeType = CDataType, PipelineVersion PipelineVer = PipelineVersion::v4, LoopScheduler LoopSched = make_default_loop_scheduler()>

CElementwiseOperation ck::tensor_operation::device::DeviceGemmXdlSplitKCShuffle_LdsDirectLoad< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferScalarPerVector, ABlockLdsAddExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferScalarPerVector, BBlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CBlockTransferScalarPerVector_NWaveNPerXDL, ComputeType, PipelineVer, LoopSched >::Argument::c_element_op

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_gemm_xdl_splitk_c_shuffle_lds_direct_load.hpp