GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched > Struct Template Reference

GridwiseGemmSplitKMultipleD_xdl_cshuffle&lt; ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched &gt; Struct Template Reference#

Composable Kernel: ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched > Struct Template Reference
ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched > Struct Template Reference

#include <gridwise_gemm_split_k_multiple_d_xdl_cshuffle.hpp>

Public Types

using ThisThreadBlock = ThisThreadBlock<BlockSize>
using GridwiseGemmPipe = GridwiseGemmPipeline_v1<NumGemmKPrefetchStage, true, true>
using DefaultAGridDesc_AK0_M_AK1
using DefaultBGridDesc_BK0_N_BK1
using EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
using DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
using DefaultBlock2ETileMap
using DsGridPointer = decltype(MakeDsGridPointer())

Static Public Member Functions

__host__ static __device__ constexpr auto GetABlockDescriptor_AK0PerBlock_MPerBlock_AK1 ()
__host__ static __device__ constexpr auto GetABlockDescriptor_AKB_AK0PerBlock_MPerBlock_AK1 ()
__host__ static __device__ constexpr auto GetBBlockDescriptor_BK0PerBlock_NPerBlock_BK1 ()
__host__ static __device__ constexpr auto GetBBlockDescriptor_BKB_BK0PerBlock_NPerBlock_BK1 ()
__host__ static __device__ constexpr auto GetCShuffleBlockDescriptor_MBlock_MPerBlock_NBlock_NPerBlock ()
static constexpr auto MakeDsGridPointer ()
__host__ static __device__ constexpr index_t GetSharedMemoryNumberOfByte ()
__host__ static __device__ constexpr auto MakeDefaultAGridDescriptor_AKB_AK0_M_AK1 (const AGridDesc_M_K &a_grid_desc_m_k, const int split_k)
__host__ static __device__ constexpr auto MakeDefaultBGridDescriptor_BKB_BK0_N_BK1 (const BGridDesc_N_K &b_grid_desc_n_k, const int split_k)
template<typename EGridDescriptor_M_N>
__host__ static __device__ constexpr auto MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock (const EGridDescriptor_M_N &e_grid_desc_m_n)
template<typename DsGridDescriptor_M_N>
__host__ static __device__ constexpr auto MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock (const DsGridDescriptor_M_N &ds_grid_desc_m_n)
__host__ static __device__ constexpr auto MakeDefaultBlock2ETileMap (const EGridDesc_M_N &e_grid_desc_m_n, const int split_k)
template<typename AGridDesc_AKB_AK0_M_AK1, typename BGridDesc_BKB_BK0_N_BK1, typename Block2ETileMap>
__host__ static __device__ constexpr bool CheckValidity (const AGridDesc_AKB_AK0_M_AK1 &a_grid_desc_akb_ak0_m_ak1, const BGridDesc_BKB_BK0_N_BK1 &b_grid_desc_bkb_bk0_n_bk1, const DsGridDesc_M_N &ds_grid_desc_m_n, const EGridDesc_M_N &e_grid_desc_m_n, const Block2ETileMap &block_2_etile_map)
__host__ static __device__ constexpr bool CalculateHasMainKBlockLoop (index_t K)
template<bool HasMainKBlockLoop, typename AGridDesc_AKB_AK0_M_AK1, typename BGridDesc_BKB_BK0_N_BK1, typename Block2ETileMap>
static __device__ void Run (const ABDataType *__restrict__ p_a_grid, const ABDataType *__restrict__ p_b_grid, DsGridPointer p_ds_grid, EDataType *__restrict__ p_e_grid, void *__restrict__ p_shared, const AElementwiseOperation &a_element_op, const BElementwiseOperation &b_element_op, const CDEElementwiseOperation &cde_element_op, const AGridDesc_AKB_AK0_M_AK1 &a_grid_desc_akb_ak0_m_ak1, const BGridDesc_BKB_BK0_N_BK1 &b_grid_desc_bkb_bk0_n_bk1, const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock &ds_grid_desc_mblock_mperblock_nblock_nperblock, const EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock &e_grid_desc_mblock_mperblock_nblock_nperblock, const Block2ETileMap &block_2_etile_map)
template<bool HasMainKBlockLoop, typename AGridDesc_AKB_AK0_M_AK1, typename BGridDesc_BKB_BK0_N_BK1, typename Block2ETileMap>
static __device__ void Run0 (const ABDataType *__restrict__ p_a_grid, const ABDataType *__restrict__ p_b_grid, DsGridPointer p_ds_grid, EDataType *__restrict__ p_e_grid, void *__restrict__ p_shared, const AElementwiseOperation &a_element_op, const BElementwiseOperation &b_element_op, const CDEElementwiseOperation &cde_element_op, const AGridDesc_AKB_AK0_M_AK1 &a_grid_desc_akb_ak0_m_ak1, const BGridDesc_BKB_BK0_N_BK1 &b_grid_desc_bkb_bk0_n_bk1, const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock &ds_grid_desc_mblock_mperblock_nblock_nperblock, const EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock &e_grid_desc_mblock_mperblock_nblock_nperblock, const Block2ETileMap &block_2_etile_map)
template<bool HasMainKBlockLoop, typename AGridDesc_AKB_AK0_M_AK1, typename BGridDesc_BKB_BK0_N_BK1, typename Block2ETileMap>
static __device__ void Run1 (const ABDataType *__restrict__ p_a_grid, const ABDataType *__restrict__ p_b_grid, EDataType *__restrict__ p_e_grid, void *__restrict__ p_shared, const AElementwiseOperation &a_element_op, const BElementwiseOperation &b_element_op, const AGridDesc_AKB_AK0_M_AK1 &a_grid_desc_akb_ak0_m_ak1, const BGridDesc_BKB_BK0_N_BK1 &b_grid_desc_bkb_bk0_n_bk1, const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock &, const EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock &e_grid_desc_mblock_mperblock_nblock_nperblock, const Block2ETileMap &block_2_etile_map)

Static Public Attributes

static constexpr index_t NumDTensor = DsDataType::Size()
static constexpr auto I0 = Number<0>{}
static constexpr auto I1 = Number<1>{}
static constexpr auto I2 = Number<2>{}
static constexpr auto I3 = Number<3>{}
static constexpr auto I4 = Number<4>{}
static constexpr auto I5 = Number<5>{}
static constexpr auto I6 = Number<6>{}
static constexpr auto I7 = Number<7>{}
static constexpr auto AK1 = Number<AK1Value>{}
static constexpr auto BK1 = Number<BK1Value>{}
static constexpr auto AK0PerBlock = Number<KPerBlock / AK1Value>{}
static constexpr auto BK0PerBlock = Number<KPerBlock / BK1Value>{}

Member Typedef Documentation

◆ DefaultAGridDesc_AK0_M_AK1

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
using ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::DefaultAGridDesc_AK0_M_AK1
Initial value:
remove_cv_t< remove_reference_t< T > > remove_cvref_t
Definition type.hpp:297
__host__ static __device__ constexpr auto MakeDefaultAGridDescriptor_AKB_AK0_M_AK1(const AGridDesc_M_K &a_grid_desc_m_k, const int split_k)
Definition gridwise_gemm_split_k_multiple_d_xdl_cshuffle.hpp:193

◆ DefaultBGridDesc_BK0_N_BK1

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
using ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::DefaultBGridDesc_BK0_N_BK1
Initial value:
__host__ static __device__ constexpr auto MakeDefaultBGridDescriptor_BKB_BK0_N_BK1(const BGridDesc_N_K &b_grid_desc_n_k, const int split_k)
Definition gridwise_gemm_split_k_multiple_d_xdl_cshuffle.hpp:219

◆ DefaultBlock2ETileMap

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
using ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::DefaultBlock2ETileMap
Initial value:
remove_cvref_t<decltype(MakeDefaultBlock2ETileMap(EGridDesc_M_N{}, 1))>
__host__ static __device__ constexpr auto MakeDefaultBlock2ETileMap(const EGridDesc_M_N &e_grid_desc_m_n, const int split_k)
Definition gridwise_gemm_split_k_multiple_d_xdl_cshuffle.hpp:280

◆ DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
using ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
Initial value:
DsGridDesc_M_N{}))>
__host__ static __device__ constexpr auto MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(const DsGridDescriptor_M_N &ds_grid_desc_m_n)
Definition gridwise_gemm_split_k_multiple_d_xdl_cshuffle.hpp:268

◆ DsGridPointer

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
using ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::DsGridPointer = decltype(MakeDsGridPointer())

◆ EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
using ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
Initial value:
EGridDesc_M_N{}))>
__host__ static __device__ constexpr auto MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(const EGridDescriptor_M_N &e_grid_desc_m_n)
Definition gridwise_gemm_split_k_multiple_d_xdl_cshuffle.hpp:246

◆ GridwiseGemmPipe

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
using ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::GridwiseGemmPipe = GridwiseGemmPipeline_v1<NumGemmKPrefetchStage, true, true>

◆ ThisThreadBlock

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
using ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::ThisThreadBlock = ThisThreadBlock<BlockSize>

Member Function Documentation

◆ CalculateHasMainKBlockLoop()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
__host__ static __device__ constexpr bool ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::CalculateHasMainKBlockLoop ( index_t K)
inlinestaticconstexpr

◆ CheckValidity()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
template<typename AGridDesc_AKB_AK0_M_AK1, typename BGridDesc_BKB_BK0_N_BK1, typename Block2ETileMap>
__host__ static __device__ constexpr bool ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::CheckValidity ( const AGridDesc_AKB_AK0_M_AK1 & a_grid_desc_akb_ak0_m_ak1,
const BGridDesc_BKB_BK0_N_BK1 & b_grid_desc_bkb_bk0_n_bk1,
const DsGridDesc_M_N & ds_grid_desc_m_n,
const EGridDesc_M_N & e_grid_desc_m_n,
const Block2ETileMap & block_2_etile_map )
inlinestaticconstexpr

◆ GetABlockDescriptor_AK0PerBlock_MPerBlock_AK1()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
__host__ static __device__ constexpr auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::GetABlockDescriptor_AK0PerBlock_MPerBlock_AK1 ( )
inlinestaticconstexpr

◆ GetABlockDescriptor_AKB_AK0PerBlock_MPerBlock_AK1()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
__host__ static __device__ constexpr auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::GetABlockDescriptor_AKB_AK0PerBlock_MPerBlock_AK1 ( )
inlinestaticconstexpr

◆ GetBBlockDescriptor_BK0PerBlock_NPerBlock_BK1()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
__host__ static __device__ constexpr auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::GetBBlockDescriptor_BK0PerBlock_NPerBlock_BK1 ( )
inlinestaticconstexpr

◆ GetBBlockDescriptor_BKB_BK0PerBlock_NPerBlock_BK1()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
__host__ static __device__ constexpr auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::GetBBlockDescriptor_BKB_BK0PerBlock_NPerBlock_BK1 ( )
inlinestaticconstexpr

◆ GetCShuffleBlockDescriptor_MBlock_MPerBlock_NBlock_NPerBlock()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
__host__ static __device__ constexpr auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::GetCShuffleBlockDescriptor_MBlock_MPerBlock_NBlock_NPerBlock ( )
inlinestaticconstexpr

◆ GetSharedMemoryNumberOfByte()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
__host__ static __device__ constexpr index_t ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::GetSharedMemoryNumberOfByte ( )
inlinestaticconstexpr

◆ MakeDefaultAGridDescriptor_AKB_AK0_M_AK1()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
__host__ static __device__ constexpr auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::MakeDefaultAGridDescriptor_AKB_AK0_M_AK1 ( const AGridDesc_M_K & a_grid_desc_m_k,
const int split_k )
inlinestaticconstexpr

◆ MakeDefaultBGridDescriptor_BKB_BK0_N_BK1()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
__host__ static __device__ constexpr auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::MakeDefaultBGridDescriptor_BKB_BK0_N_BK1 ( const BGridDesc_N_K & b_grid_desc_n_k,
const int split_k )
inlinestaticconstexpr

◆ MakeDefaultBlock2ETileMap()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
__host__ static __device__ constexpr auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::MakeDefaultBlock2ETileMap ( const EGridDesc_M_N & e_grid_desc_m_n,
const int split_k )
inlinestaticconstexpr

◆ MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
template<typename DsGridDescriptor_M_N>
__host__ static __device__ constexpr auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock ( const DsGridDescriptor_M_N & ds_grid_desc_m_n)
inlinestaticconstexpr

◆ MakeDsGridPointer()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
constexpr auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::MakeDsGridPointer ( )
inlinestaticconstexpr

◆ MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
template<typename EGridDescriptor_M_N>
__host__ static __device__ constexpr auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock ( const EGridDescriptor_M_N & e_grid_desc_m_n)
inlinestaticconstexpr

◆ Run()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
template<bool HasMainKBlockLoop, typename AGridDesc_AKB_AK0_M_AK1, typename BGridDesc_BKB_BK0_N_BK1, typename Block2ETileMap>
__device__ void ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::Run ( const ABDataType *__restrict__ p_a_grid,
const ABDataType *__restrict__ p_b_grid,
DsGridPointer p_ds_grid,
EDataType *__restrict__ p_e_grid,
void *__restrict__ p_shared,
const AElementwiseOperation & a_element_op,
const BElementwiseOperation & b_element_op,
const CDEElementwiseOperation & cde_element_op,
const AGridDesc_AKB_AK0_M_AK1 & a_grid_desc_akb_ak0_m_ak1,
const BGridDesc_BKB_BK0_N_BK1 & b_grid_desc_bkb_bk0_n_bk1,
const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock & ds_grid_desc_mblock_mperblock_nblock_nperblock,
const EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock & e_grid_desc_mblock_mperblock_nblock_nperblock,
const Block2ETileMap & block_2_etile_map )
inlinestatic

◆ Run0()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
template<bool HasMainKBlockLoop, typename AGridDesc_AKB_AK0_M_AK1, typename BGridDesc_BKB_BK0_N_BK1, typename Block2ETileMap>
__device__ void ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::Run0 ( const ABDataType *__restrict__ p_a_grid,
const ABDataType *__restrict__ p_b_grid,
DsGridPointer p_ds_grid,
EDataType *__restrict__ p_e_grid,
void *__restrict__ p_shared,
const AElementwiseOperation & a_element_op,
const BElementwiseOperation & b_element_op,
const CDEElementwiseOperation & cde_element_op,
const AGridDesc_AKB_AK0_M_AK1 & a_grid_desc_akb_ak0_m_ak1,
const BGridDesc_BKB_BK0_N_BK1 & b_grid_desc_bkb_bk0_n_bk1,
const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock & ds_grid_desc_mblock_mperblock_nblock_nperblock,
const EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock & e_grid_desc_mblock_mperblock_nblock_nperblock,
const Block2ETileMap & block_2_etile_map )
inlinestatic

◆ Run1()

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
template<bool HasMainKBlockLoop, typename AGridDesc_AKB_AK0_M_AK1, typename BGridDesc_BKB_BK0_N_BK1, typename Block2ETileMap>
__device__ void ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::Run1 ( const ABDataType *__restrict__ p_a_grid,
const ABDataType *__restrict__ p_b_grid,
EDataType *__restrict__ p_e_grid,
void *__restrict__ p_shared,
const AElementwiseOperation & a_element_op,
const BElementwiseOperation & b_element_op,
const AGridDesc_AKB_AK0_M_AK1 & a_grid_desc_akb_ak0_m_ak1,
const BGridDesc_BKB_BK0_N_BK1 & b_grid_desc_bkb_bk0_n_bk1,
const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock & ,
const EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock & e_grid_desc_mblock_mperblock_nblock_nperblock,
const Block2ETileMap & block_2_etile_map )
inlinestatic

Member Data Documentation

◆ AK0PerBlock

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::AK0PerBlock = Number<KPerBlock / AK1Value>{}
staticconstexpr

◆ AK1

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::AK1 = Number<AK1Value>{}
staticconstexpr

◆ BK0PerBlock

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::BK0PerBlock = Number<KPerBlock / BK1Value>{}
staticconstexpr

◆ BK1

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::BK1 = Number<BK1Value>{}
staticconstexpr

◆ I0

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::I0 = Number<0>{}
staticconstexpr

◆ I1

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::I1 = Number<1>{}
staticconstexpr

◆ I2

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::I2 = Number<2>{}
staticconstexpr

◆ I3

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::I3 = Number<3>{}
staticconstexpr

◆ I4

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::I4 = Number<4>{}
staticconstexpr

◆ I5

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::I5 = Number<5>{}
staticconstexpr

◆ I6

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::I6 = Number<6>{}
staticconstexpr

◆ I7

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
auto ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::I7 = Number<7>{}
staticconstexpr

◆ NumDTensor

template<typename ABDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, InMemoryDataOperationEnum EGlobalMemoryDataOperation, typename AGridDesc_M_K, typename BGridDesc_N_K, typename DsGridDesc_M_N, typename EGridDesc_M_N, index_t NumGemmKPrefetchStage, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BThreadTransferSrcResetCoordinateAfterRun, index_t BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopScheduler LoopSched>
index_t ck::GridwiseGemmSplitKMultipleD_xdl_cshuffle< ABDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, EGlobalMemoryDataOperation, AGridDesc_M_K, BGridDesc_N_K, DsGridDesc_M_N, EGridDesc_M_N, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1Value, BK1Value, MPerXdl, NPerXdl, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, AThreadTransferSrcResetCoordinateAfterRun, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BThreadTransferSrcResetCoordinateAfterRun, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, LoopSched >::NumDTensor = DsDataType::Size()
staticconstexpr

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