DeviceGroupedConvBwdDataMultipleD< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, AComputeType, BComputeType > Struct Template Reference

DeviceGroupedConvBwdDataMultipleD&lt; NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, AComputeType, BComputeType &gt; Struct Template Reference#

Composable Kernel: ck::tensor_operation::device::DeviceGroupedConvBwdDataMultipleD< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, AComputeType, BComputeType > Struct Template Reference
Public Member Functions | Static Public Attributes | List of all members
ck::tensor_operation::device::DeviceGroupedConvBwdDataMultipleD< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, AComputeType, BComputeType > Struct Template Referenceabstract

#include <device_grouped_conv_bwd_data_multiple_d.hpp>

Inheritance diagram for ck::tensor_operation::device::DeviceGroupedConvBwdDataMultipleD< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, AComputeType, BComputeType >:
ck::tensor_operation::device::BaseOperator

Public Member Functions

virtual std::unique_ptr< BaseArgument > MakeArgumentPointer (const void *p_a, const void *p_b, const std::array< const void *, NumDTensor > &p_ds, void *p_e, const std::array< index_t, NDimSpatial+3 > &a_g_n_k_wos_lengths, const std::array< index_t, NDimSpatial+3 > &a_g_n_k_wos_strides, const std::array< index_t, NDimSpatial+3 > &b_g_k_c_xs_lengths, const std::array< index_t, NDimSpatial+3 > &b_g_k_c_xs_strides, const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_lengths, const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > &ds_g_n_k_wos_strides, const std::array< index_t, NDimSpatial+3 > &e_g_n_c_wis_lengths, const std::array< index_t, NDimSpatial+3 > &e_g_n_c_wis_strides, const std::array< index_t, NDimSpatial > &conv_filter_strides, const std::array< index_t, NDimSpatial > &conv_filter_dilations, const std::array< index_t, NDimSpatial > &input_left_pads, const std::array< index_t, NDimSpatial > &input_right_pads, const AElementwiseOperation &a_element_op, const BElementwiseOperation &b_element_op, const CDEElementwiseOperation &cde_element_op, const ck::index_t split_k=1)=0
virtual std::unique_ptr< BaseInvoker > MakeInvokerPointer ()=0
Public Member Functions inherited from ck::tensor_operation::device::BaseOperator
 BaseOperator ()=default
 BaseOperator (const BaseOperator &)=default
BaseOperator & operator= (const BaseOperator &)=default
virtual bool IsSupportedArgument (const BaseArgument *)
virtual std::string GetTypeString () const
virtual std::string GetInstanceString () const
virtual std::string GetTypeIdName () const
virtual std::optional< std::string > GetObjectName () const
virtual std::optional< std::string > GetTemplateInfo () const
virtual std::string GetTypeIdHashCode () const
virtual size_t GetWorkSpaceSize (const BaseArgument *) const
virtual void SetWorkSpacePointer (BaseArgument *p_arg, void *p_workspace, const StreamConfig &=StreamConfig{}) const
virtual ~BaseOperator ()

Static Public Attributes

static constexpr index_t NumDTensor = DsDataType::Size()

Member Function Documentation

◆ MakeArgumentPointer()

template<ck::index_t NDimSpatial, typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, typename AComputeType = ADataType, typename BComputeType = AComputeType>
virtual std::unique_ptr< BaseArgument > ck::tensor_operation::device::DeviceGroupedConvBwdDataMultipleD< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, AComputeType, BComputeType >::MakeArgumentPointer ( const void * p_a,
const void * p_b,
const std::array< const void *, NumDTensor > & p_ds,
void * p_e,
const std::array< index_t, NDimSpatial+3 > & a_g_n_k_wos_lengths,
const std::array< index_t, NDimSpatial+3 > & a_g_n_k_wos_strides,
const std::array< index_t, NDimSpatial+3 > & b_g_k_c_xs_lengths,
const std::array< index_t, NDimSpatial+3 > & b_g_k_c_xs_strides,
const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > & ds_g_n_k_wos_lengths,
const std::array< std::array< index_t, NDimSpatial+3 >, NumDTensor > & ds_g_n_k_wos_strides,
const std::array< index_t, NDimSpatial+3 > & e_g_n_c_wis_lengths,
const std::array< index_t, NDimSpatial+3 > & e_g_n_c_wis_strides,
const std::array< index_t, NDimSpatial > & conv_filter_strides,
const std::array< index_t, NDimSpatial > & conv_filter_dilations,
const std::array< index_t, NDimSpatial > & input_left_pads,
const std::array< index_t, NDimSpatial > & input_right_pads,
const AElementwiseOperation & a_element_op,
const BElementwiseOperation & b_element_op,
const CDEElementwiseOperation & cde_element_op,
const ck::index_t split_k = 1 )
pure virtual

Implemented in ck::tensor_operation::device::DeviceGroupedConvBwdDataMultipleD_Wmma_CShuffle< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, ConvBackwardDataSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerWMMA, NPerWMMA, MRepeat, NRepeat, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CDEShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, NumGemmKPrefetchStage, LoopSched, PipelineVer >, and ck::tensor_operation::device::DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, ConvBackwardDataSpecialization, DoPadGemmM, DoPadGemmN, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, LoopSched, AComputeType, BComputeType, MaxTransposeTransferInScalarPerVector, MaxTransposeTransferOutScalarPerVector >.

◆ MakeInvokerPointer()

template<ck::index_t NDimSpatial, typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, typename AComputeType = ADataType, typename BComputeType = AComputeType>
virtual std::unique_ptr< BaseInvoker > ck::tensor_operation::device::DeviceGroupedConvBwdDataMultipleD< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, AComputeType, BComputeType >::MakeInvokerPointer ( )
pure virtual

Implemented in ck::tensor_operation::device::DeviceGroupedConvBwdDataMultipleD_Wmma_CShuffle< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, ConvBackwardDataSpecialization, BlockSize, MPerBlock, NPerBlock, K0PerBlock, K1, MPerWMMA, NPerWMMA, MRepeat, NRepeat, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CDEShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVector_NPerBlock, NumGemmKPrefetchStage, LoopSched, PipelineVer >, and ck::tensor_operation::device::DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, AccDataType, CShuffleDataType, DsDataType, EDataType, AElementwiseOp, BElementwiseOp, CDEElementwiseOp, ConvBackwardDataSpecialization, DoPadGemmM, DoPadGemmN, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEBlockTransferScalarPerVector_NPerBlock, LoopSched, AComputeType, BComputeType, MaxTransposeTransferInScalarPerVector, MaxTransposeTransferOutScalarPerVector >.

Member Data Documentation

◆ NumDTensor

template<ck::index_t NDimSpatial, typename ALayout, typename BLayout, typename DsLayout, typename ELayout, typename ADataType, typename BDataType, typename DsDataType, typename EDataType, typename AElementwiseOperation, typename BElementwiseOperation, typename CDEElementwiseOperation, typename AComputeType = ADataType, typename BComputeType = AComputeType>
index_t ck::tensor_operation::device::DeviceGroupedConvBwdDataMultipleD< NDimSpatial, ALayout, BLayout, DsLayout, ELayout, ADataType, BDataType, DsDataType, EDataType, AElementwiseOperation, BElementwiseOperation, CDEElementwiseOperation, AComputeType, BComputeType >::NumDTensor = DsDataType::Size()
staticconstexpr
The documentation for this struct was generated from the following file: