GemmBiasTransKernelArg Struct Reference#
ck::tensor_operation::device::DeviceGroupedGemm_Xdl_Fixed_NK< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, NumPrefetch, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, PipelineVer, LoopSched, ComputeType, ALDSType, BLDSType >::GemmBiasTransKernelArg Struct Reference
#include <device_grouped_gemm_xdl_fixed_nk.hpp>
Public Attributes | |
| const void * | a_ptr_ |
| const void * | b_ptr_ |
| std::array< const void *, NumDTensor > | ds_ptr_ |
| void * | e_ptr_ |
| index_t | M_ |
| index_t | N_ |
| index_t | K_ |
| index_t | StrideA_ |
| index_t | StrideB_ |
| std::array< index_t, NumDTensor > | StrideDs_ |
| index_t | StrideE_ |
Member Data Documentation
◆ a_ptr_
template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, ck::index_t NumPrefetch, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t KPerBlock, ck::index_t AK1, ck::index_t BK1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEBlockTransferScalarPerVector_NPerBlock, PipelineVersion PipelineVer = PipelineVersion::v1, LoopScheduler LoopSched = make_default_loop_scheduler(), typename ComputeType = ADataType, typename ALDSType = ComputeType, typename BLDSType = ComputeType>
| const void* ck::tensor_operation::device::DeviceGroupedGemm_Xdl_Fixed_NK< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, NumPrefetch, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, PipelineVer, LoopSched, ComputeType, ALDSType, BLDSType >::GemmBiasTransKernelArg::a_ptr_ |
◆ b_ptr_
template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, ck::index_t NumPrefetch, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t KPerBlock, ck::index_t AK1, ck::index_t BK1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEBlockTransferScalarPerVector_NPerBlock, PipelineVersion PipelineVer = PipelineVersion::v1, LoopScheduler LoopSched = make_default_loop_scheduler(), typename ComputeType = ADataType, typename ALDSType = ComputeType, typename BLDSType = ComputeType>
| const void* ck::tensor_operation::device::DeviceGroupedGemm_Xdl_Fixed_NK< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, NumPrefetch, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, PipelineVer, LoopSched, ComputeType, ALDSType, BLDSType >::GemmBiasTransKernelArg::b_ptr_ |
◆ ds_ptr_
template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, ck::index_t NumPrefetch, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t KPerBlock, ck::index_t AK1, ck::index_t BK1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEBlockTransferScalarPerVector_NPerBlock, PipelineVersion PipelineVer = PipelineVersion::v1, LoopScheduler LoopSched = make_default_loop_scheduler(), typename ComputeType = ADataType, typename ALDSType = ComputeType, typename BLDSType = ComputeType>
| std::array<const void*, NumDTensor> ck::tensor_operation::device::DeviceGroupedGemm_Xdl_Fixed_NK< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, NumPrefetch, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, PipelineVer, LoopSched, ComputeType, ALDSType, BLDSType >::GemmBiasTransKernelArg::ds_ptr_ |
◆ e_ptr_
template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, ck::index_t NumPrefetch, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t KPerBlock, ck::index_t AK1, ck::index_t BK1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEBlockTransferScalarPerVector_NPerBlock, PipelineVersion PipelineVer = PipelineVersion::v1, LoopScheduler LoopSched = make_default_loop_scheduler(), typename ComputeType = ADataType, typename ALDSType = ComputeType, typename BLDSType = ComputeType>
| void* ck::tensor_operation::device::DeviceGroupedGemm_Xdl_Fixed_NK< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, NumPrefetch, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, PipelineVer, LoopSched, ComputeType, ALDSType, BLDSType >::GemmBiasTransKernelArg::e_ptr_ |
◆ K_
template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, ck::index_t NumPrefetch, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t KPerBlock, ck::index_t AK1, ck::index_t BK1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEBlockTransferScalarPerVector_NPerBlock, PipelineVersion PipelineVer = PipelineVersion::v1, LoopScheduler LoopSched = make_default_loop_scheduler(), typename ComputeType = ADataType, typename ALDSType = ComputeType, typename BLDSType = ComputeType>
| index_t ck::tensor_operation::device::DeviceGroupedGemm_Xdl_Fixed_NK< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, NumPrefetch, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, PipelineVer, LoopSched, ComputeType, ALDSType, BLDSType >::GemmBiasTransKernelArg::K_ |
◆ M_
template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, ck::index_t NumPrefetch, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t KPerBlock, ck::index_t AK1, ck::index_t BK1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEBlockTransferScalarPerVector_NPerBlock, PipelineVersion PipelineVer = PipelineVersion::v1, LoopScheduler LoopSched = make_default_loop_scheduler(), typename ComputeType = ADataType, typename ALDSType = ComputeType, typename BLDSType = ComputeType>
| index_t ck::tensor_operation::device::DeviceGroupedGemm_Xdl_Fixed_NK< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, NumPrefetch, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, PipelineVer, LoopSched, ComputeType, ALDSType, BLDSType >::GemmBiasTransKernelArg::M_ |
◆ N_
template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, ck::index_t NumPrefetch, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t KPerBlock, ck::index_t AK1, ck::index_t BK1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEBlockTransferScalarPerVector_NPerBlock, PipelineVersion PipelineVer = PipelineVersion::v1, LoopScheduler LoopSched = make_default_loop_scheduler(), typename ComputeType = ADataType, typename ALDSType = ComputeType, typename BLDSType = ComputeType>
| index_t ck::tensor_operation::device::DeviceGroupedGemm_Xdl_Fixed_NK< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, NumPrefetch, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, PipelineVer, LoopSched, ComputeType, ALDSType, BLDSType >::GemmBiasTransKernelArg::N_ |
◆ StrideA_
template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, ck::index_t NumPrefetch, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t KPerBlock, ck::index_t AK1, ck::index_t BK1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEBlockTransferScalarPerVector_NPerBlock, PipelineVersion PipelineVer = PipelineVersion::v1, LoopScheduler LoopSched = make_default_loop_scheduler(), typename ComputeType = ADataType, typename ALDSType = ComputeType, typename BLDSType = ComputeType>
| index_t ck::tensor_operation::device::DeviceGroupedGemm_Xdl_Fixed_NK< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, NumPrefetch, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, PipelineVer, LoopSched, ComputeType, ALDSType, BLDSType >::GemmBiasTransKernelArg::StrideA_ |
◆ StrideB_
template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, ck::index_t NumPrefetch, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t KPerBlock, ck::index_t AK1, ck::index_t BK1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEBlockTransferScalarPerVector_NPerBlock, PipelineVersion PipelineVer = PipelineVersion::v1, LoopScheduler LoopSched = make_default_loop_scheduler(), typename ComputeType = ADataType, typename ALDSType = ComputeType, typename BLDSType = ComputeType>
| index_t ck::tensor_operation::device::DeviceGroupedGemm_Xdl_Fixed_NK< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, NumPrefetch, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, PipelineVer, LoopSched, ComputeType, ALDSType, BLDSType >::GemmBiasTransKernelArg::StrideB_ |
◆ StrideDs_
template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, ck::index_t NumPrefetch, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t KPerBlock, ck::index_t AK1, ck::index_t BK1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEBlockTransferScalarPerVector_NPerBlock, PipelineVersion PipelineVer = PipelineVersion::v1, LoopScheduler LoopSched = make_default_loop_scheduler(), typename ComputeType = ADataType, typename ALDSType = ComputeType, typename BLDSType = ComputeType>
| std::array<index_t, NumDTensor> ck::tensor_operation::device::DeviceGroupedGemm_Xdl_Fixed_NK< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, NumPrefetch, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, PipelineVer, LoopSched, ComputeType, ALDSType, BLDSType >::GemmBiasTransKernelArg::StrideDs_ |
◆ StrideE_
template<typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename AccDataType, typename CShuffleDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, GemmSpecialization GemmSpec, ck::index_t NumPrefetch, ck::index_t BlockSize, ck::index_t MPerBlock, ck::index_t NPerBlock, ck::index_t KPerBlock, ck::index_t AK1, ck::index_t BK1, ck::index_t MPerXDL, ck::index_t NPerXDL, ck::index_t MXdlPerWave, ck::index_t NXdlPerWave, typename ABlockTransferThreadClusterLengths_K0_M_K1, typename ABlockTransferThreadClusterArrangeOrder, typename ABlockTransferSrcAccessOrder, ck::index_t ABlockTransferSrcVectorDim, ck::index_t ABlockTransferSrcScalarPerVector, ck::index_t ABlockTransferDstScalarPerVector_K1, bool ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_K0_N_K1, typename BBlockTransferThreadClusterArrangeOrder, typename BBlockTransferSrcAccessOrder, ck::index_t BBlockTransferSrcVectorDim, ck::index_t BBlockTransferSrcScalarPerVector, ck::index_t BBlockTransferDstScalarPerVector_K1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, index_t CDEBlockTransferScalarPerVector_NPerBlock, PipelineVersion PipelineVer = PipelineVersion::v1, LoopScheduler LoopSched = make_default_loop_scheduler(), typename ComputeType = ADataType, typename ALDSType = ComputeType, typename BLDSType = ComputeType>
| index_t ck::tensor_operation::device::DeviceGroupedGemm_Xdl_Fixed_NK< ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, GemmSpec, NumPrefetch, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_K0_N_K1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_K1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, PipelineVer, LoopSched, ComputeType, ALDSType, BLDSType >::GemmBiasTransKernelArg::StrideE_ |
The documentation for this struct was generated from the following file: