Composable Kernel: 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 Struct Reference

Argument Struct Reference

Argument Struct Reference#

Composable Kernel: 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 Struct Reference

Public Member Functions | Public Attributes | List of all members

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 Struct Reference

#include <gridwise_gemm_xdlops_splitk_lds_direct_load.hpp>

Inheritance diagram for 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:

Public Member Functions
	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
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<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

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_ )

inline

Member Function Documentation

◆ Print()

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

void 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::Print ( ) const

inline

Member Data Documentation

◆ K

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

index_t 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::K

◆ K0Padded

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

index_t 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::K0Padded

◆ k_batch

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

index_t 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::k_batch

◆ KPadded

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

index_t 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::KPadded

◆ M

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

index_t 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::M

◆ MPadded

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

index_t 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::MPadded

◆ N

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

index_t 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::N

◆ NPadded

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

index_t 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::NPadded

◆ p_a_grid

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

const FloatA* 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::p_a_grid

◆ p_b_grid

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

const FloatB* 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::p_b_grid

◆ p_c_grid

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

FloatC* 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::p_c_grid

◆ StrideA

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

index_t 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::StrideA

◆ StrideB

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

index_t 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::StrideB

◆ StrideC

template<index_t BlockSize, typename FloatA, typename FloatB, typename FloatAcc, typename FloatC, typename ALayout, typename BLayout, typename CLayout, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t NumGemmKPrefetchStage, index_t MPerBlock, index_t NPerBlock, index_t K0PerBlock, index_t MPerXdl, index_t NPerXdl, index_t K1Value, index_t MRepeat, index_t NRepeat, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, bool BBlockLdsExtraN, index_t CShuffleMRepeatPerShuffle, index_t CShuffleNRepeatPerShuffle, index_t CBlockTransferScalarPerVector_NWaveNPerXDL, typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, LoopScheduler LoopSched = make_default_loop_scheduler(), PipelineVersion PipelineVer = PipelineVersion::v4, typename ComputeType = FloatC>

index_t 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::StrideC

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

gridwise_gemm_xdlops_splitk_lds_direct_load.hpp