Argument Struct Reference

Argument Struct Reference#

Composable Kernel: ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Argument Struct Reference
Public Member Functions | Public Attributes | List of all members
ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Argument Struct Reference

#include <gridwise_gemm_xdl_cshuffle_v3_multi_d.hpp>

Inheritance diagram for ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Argument:
ck::tensor_operation::device::BaseArgument ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Problem

Public Member Functions

__host__ Argument ()=default
__host__ Argument (const ADataType *p_a_grid_, const BDataType *p_b_grid_, std::array< const void *, NumDTensor > p_ds_grid_, CDataType *p_c_grid_, index_t M_, index_t N_, index_t K_, index_t StrideA_, index_t StrideB_, std::array< index_t, NumDTensor > StrideDs_, index_t StrideC_, index_t k_batch_, AElementwiseOperation a_element_op_, BElementwiseOperation b_element_op_, CElementwiseOperation c_element_op_)
Public Member Functions inherited from ck::tensor_operation::device::BaseArgument
 BaseArgument ()=default
 BaseArgument (const BaseArgument &)=default
BaseArgument & operator= (const BaseArgument &)=default
virtual ~BaseArgument ()
Public Member Functions inherited from ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Problem
__host__ __device__ Problem ()=default
__host__ __device__ Problem (index_t M_, index_t N_, index_t K_, index_t StrideA_, index_t StrideB_, std::array< index_t, NumDTensor > StrideDs_, index_t StrideC_, index_t KBatch_)
__host__ void Print () const

Public Attributes

const ADataType * p_a_grid
const BDataType * p_b_grid
DsGridPointer p_ds_grid
CDataType * p_c_grid
AElementwiseOperation a_element_op
BElementwiseOperation b_element_op
CElementwiseOperation c_element_op
Public Attributes inherited from ck::tensor_operation::device::BaseArgument
void * p_workspace_ = nullptr
Public Attributes inherited from ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Problem
index_t M
index_t N
index_t K
index_t StrideA
index_t StrideB
std::array< index_t, NumDTensor > StrideDs
index_t StrideC
index_t KBatch
index_t MPadded
index_t NPadded
index_t KRead
index_t KPadded
index_t AK0
index_t BK0
index_t MBlock
index_t NBlock

Constructor & Destructor Documentation

◆ Argument() [1/2]

template<typename ALayout, typename BLayout, typename DsLayout, typename CLayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename CDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1Value, index_t BK1Value, index_t MPerXdl, index_t NPerXdl, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, bool AThreadTransferSrcResetCoordinateAfterRun, index_t ABlockLdsExtraMCustom, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraNCustom, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, typename CDEShuffleBlockTransferScalarPerVectors, BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v4, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ADataType, typename LDSTypeB = BDataType, bool DoElementwiseBeforeCShuffle = false, bool DirectLoad = false>
__host__ ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Argument::Argument ( )
default

◆ Argument() [2/2]

template<typename ALayout, typename BLayout, typename DsLayout, typename CLayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename CDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1Value, index_t BK1Value, index_t MPerXdl, index_t NPerXdl, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, bool AThreadTransferSrcResetCoordinateAfterRun, index_t ABlockLdsExtraMCustom, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraNCustom, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, typename CDEShuffleBlockTransferScalarPerVectors, BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v4, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ADataType, typename LDSTypeB = BDataType, bool DoElementwiseBeforeCShuffle = false, bool DirectLoad = false>
__host__ ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Argument::Argument ( const ADataType * p_a_grid_,
const BDataType * p_b_grid_,
std::array< const void *, NumDTensor > p_ds_grid_,
CDataType * p_c_grid_,
index_t M_,
index_t N_,
index_t K_,
index_t StrideA_,
index_t StrideB_,
std::array< index_t, NumDTensor > StrideDs_,
index_t StrideC_,
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 ALayout, typename BLayout, typename DsLayout, typename CLayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename CDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1Value, index_t BK1Value, index_t MPerXdl, index_t NPerXdl, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, bool AThreadTransferSrcResetCoordinateAfterRun, index_t ABlockLdsExtraMCustom, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraNCustom, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, typename CDEShuffleBlockTransferScalarPerVectors, BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v4, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ADataType, typename LDSTypeB = BDataType, bool DoElementwiseBeforeCShuffle = false, bool DirectLoad = false>
AElementwiseOperation ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Argument::a_element_op

◆ b_element_op

template<typename ALayout, typename BLayout, typename DsLayout, typename CLayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename CDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1Value, index_t BK1Value, index_t MPerXdl, index_t NPerXdl, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, bool AThreadTransferSrcResetCoordinateAfterRun, index_t ABlockLdsExtraMCustom, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraNCustom, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, typename CDEShuffleBlockTransferScalarPerVectors, BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v4, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ADataType, typename LDSTypeB = BDataType, bool DoElementwiseBeforeCShuffle = false, bool DirectLoad = false>
BElementwiseOperation ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Argument::b_element_op

◆ c_element_op

template<typename ALayout, typename BLayout, typename DsLayout, typename CLayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename CDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1Value, index_t BK1Value, index_t MPerXdl, index_t NPerXdl, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, bool AThreadTransferSrcResetCoordinateAfterRun, index_t ABlockLdsExtraMCustom, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraNCustom, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, typename CDEShuffleBlockTransferScalarPerVectors, BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v4, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ADataType, typename LDSTypeB = BDataType, bool DoElementwiseBeforeCShuffle = false, bool DirectLoad = false>
CElementwiseOperation ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Argument::c_element_op

◆ p_a_grid

template<typename ALayout, typename BLayout, typename DsLayout, typename CLayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename CDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1Value, index_t BK1Value, index_t MPerXdl, index_t NPerXdl, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, bool AThreadTransferSrcResetCoordinateAfterRun, index_t ABlockLdsExtraMCustom, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraNCustom, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, typename CDEShuffleBlockTransferScalarPerVectors, BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v4, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ADataType, typename LDSTypeB = BDataType, bool DoElementwiseBeforeCShuffle = false, bool DirectLoad = false>
const ADataType* ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Argument::p_a_grid

◆ p_b_grid

template<typename ALayout, typename BLayout, typename DsLayout, typename CLayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename CDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1Value, index_t BK1Value, index_t MPerXdl, index_t NPerXdl, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, bool AThreadTransferSrcResetCoordinateAfterRun, index_t ABlockLdsExtraMCustom, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraNCustom, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, typename CDEShuffleBlockTransferScalarPerVectors, BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v4, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ADataType, typename LDSTypeB = BDataType, bool DoElementwiseBeforeCShuffle = false, bool DirectLoad = false>
const BDataType* ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Argument::p_b_grid

◆ p_c_grid

template<typename ALayout, typename BLayout, typename DsLayout, typename CLayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename CDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1Value, index_t BK1Value, index_t MPerXdl, index_t NPerXdl, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, bool AThreadTransferSrcResetCoordinateAfterRun, index_t ABlockLdsExtraMCustom, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraNCustom, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, typename CDEShuffleBlockTransferScalarPerVectors, BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v4, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ADataType, typename LDSTypeB = BDataType, bool DoElementwiseBeforeCShuffle = false, bool DirectLoad = false>
CDataType* ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Argument::p_c_grid

◆ p_ds_grid

template<typename ALayout, typename BLayout, typename DsLayout, typename CLayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename CDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CElementwiseOperation, tensor_operation::device::GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1Value, index_t BK1Value, index_t MPerXdl, index_t NPerXdl, index_t MXdlPerWave, index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_AK0_M_AK1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, index_t ABlockTransferSrcVectorDim, index_t ABlockTransferSrcScalarPerVector, index_t ABlockTransferDstScalarPerVector_AK1, bool AThreadTransferSrcResetCoordinateAfterRun, index_t ABlockLdsExtraMCustom, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraNCustom, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, typename CDEShuffleBlockTransferScalarPerVectors, BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v4, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ADataType, typename LDSTypeB = BDataType, bool DoElementwiseBeforeCShuffle = false, bool DirectLoad = false>
DsGridPointer ck::GridwiseGemmMultiD_xdl_cshuffle_v3< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraMCustom, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraNCustom, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB, DoElementwiseBeforeCShuffle, DirectLoad >::Argument::p_ds_grid
The documentation for this struct was generated from the following file: