DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB > Struct Template Reference

DeviceMoeGemm&lt; ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB &gt; Struct Template Reference#

Composable Kernel: ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB > Struct Template Reference
Classes | Public Types | Public Member Functions | Static Public Member Functions | Static Public Attributes | List of all members
ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB > Struct Template Reference

#include <device_moe_gemm.hpp>

Inheritance diagram for ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >:
ck::tensor_operation::device::DeviceGemmMultipleDSplitKBPreShuffle< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation > ck::tensor_operation::device::BaseOperator

Classes

struct  Invoker
 

Public Types

using GridwiseGemm = GridwiseMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, GemmAccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, false, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, false, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >
 
using Argument = typename GridwiseGemm::Argument
 

Public Member Functions

int GetPreShuffleParameters () override
 
bool IsSupportedArgument (const BaseArgument *p_arg) override
 
std::unique_ptr< BaseArgument > MakeArgumentPointer (const void *p_a, const void *p_b, std::array< const void *, NumDTensor > p_ds, void *p_c, index_t M, index_t N, index_t K, index_t StrideA, index_t StrideB, std::array< ck::index_t, NumDTensor > StrideDs, index_t StrideC, index_t KBatch, AElementwiseOperation a_element_op, BElementwiseOperation b_element_op, CElementwiseOperation c_element_op) override
 
std::unique_ptr< BaseInvoker > MakeInvokerPointer () override
 
std::string GetTypeString () const override
 
- Public Member Functions inherited from ck::tensor_operation::device::BaseOperator
 BaseOperator ()=default
 
 BaseOperator (const BaseOperator &)=default
 
BaseOperator & operator= (const BaseOperator &)=default
 
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 Member Functions

static constexpr bool IsValidCompilationParameter ()
 
static bool IsSupportedArgument (const Argument &arg)
 
static auto MakeArgument (const void *p_sorted_token_ids, const void *p_sorted_expert_ids, const void *p_max_token_id, const void *p_a, const void *p_b, std::array< const void *, NumDTensor > p_ds, void *p_c, index_t NumTokens, index_t TopK, index_t M, index_t N, index_t K, index_t StrideA, index_t StrideB, std::array< index_t, NumDTensor > StrideDs, index_t StrideC, index_t KBatch, AElementwiseOperation a_element_op, BElementwiseOperation b_element_op, CElementwiseOperation c_element_op)
 
static auto MakeInvoker ()
 

Static Public Attributes

static constexpr index_t NumDTensor = DsDataType::Size()
 
static constexpr index_t APackedSize
 
static constexpr index_t BPackedSize
 
- Static Public Attributes inherited from ck::tensor_operation::device::DeviceGemmMultipleDSplitKBPreShuffle< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation >
static constexpr index_t NumDTensor
 

Member Typedef Documentation

◆ Argument

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
using ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::Argument = typename GridwiseGemm::Argument

◆ GridwiseGemm

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
using ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::GridwiseGemm = GridwiseMoeGemm<ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, GemmAccDataType, CShuffleDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, BlockSize, MPerBlock, NPerBlock, KPerBlock, AK1, BK1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, false, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, false, BBlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB>

Member Function Documentation

◆ GetPreShuffleParameters()

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
int ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::GetPreShuffleParameters ( )
inlineoverridevirtual

Implements ck::tensor_operation::device::DeviceGemmMultipleDSplitKBPreShuffle< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation >.

◆ GetTypeString()

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
std::string ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::GetTypeString ( ) const
inlineoverridevirtual

Reimplemented from ck::tensor_operation::device::BaseOperator.

◆ IsSupportedArgument() [1/2]

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
static bool ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::IsSupportedArgument ( const Argument &  arg)
inlinestatic

◆ IsSupportedArgument() [2/2]

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
bool ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::IsSupportedArgument ( const BaseArgument *  p_arg)
inlineoverridevirtual

Reimplemented from ck::tensor_operation::device::BaseOperator.

◆ IsValidCompilationParameter()

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
static constexpr bool ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::IsValidCompilationParameter ( )
inlinestaticconstexpr

◆ MakeArgument()

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
static auto ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::MakeArgument ( const void *  p_sorted_token_ids,
const void *  p_sorted_expert_ids,
const void *  p_max_token_id,
const void *  p_a,
const void *  p_b,
std::array< const void *, NumDTensor >  p_ds,
void *  p_c,
index_t  NumTokens,
index_t  TopK,
index_t  M,
index_t  N,
index_t  K,
index_t  StrideA,
index_t  StrideB,
std::array< index_t, NumDTensor >  StrideDs,
index_t  StrideC,
index_t  KBatch,
AElementwiseOperation  a_element_op,
BElementwiseOperation  b_element_op,
CElementwiseOperation  c_element_op 
)
inlinestatic

◆ MakeArgumentPointer()

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
std::unique_ptr<BaseArgument> ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::MakeArgumentPointer ( const void *  p_a,
const void *  p_b,
std::array< const void *, NumDTensor >  p_ds,
void *  p_c,
index_t  M,
index_t  N,
index_t  K,
index_t  StrideA,
index_t  StrideB,
std::array< ck::index_t, NumDTensor >  StrideDs,
index_t  StrideC,
index_t  KBatch,
AElementwiseOperation  a_element_op,
BElementwiseOperation  b_element_op,
CElementwiseOperation  c_element_op 
)
inlineoverridevirtual

Implements ck::tensor_operation::device::DeviceGemmMultipleDSplitKBPreShuffle< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation >.

◆ MakeInvoker()

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
static auto ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::MakeInvoker ( )
inlinestatic

◆ MakeInvokerPointer()

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
std::unique_ptr<BaseInvoker> ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::MakeInvokerPointer ( )
inlineoverridevirtual

Implements ck::tensor_operation::device::DeviceGemmMultipleDSplitKBPreShuffle< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation >.

Member Data Documentation

◆ APackedSize

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
constexpr index_t ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::APackedSize
staticconstexpr
Initial value:
= []() {
if constexpr(is_same_v<remove_cvref_t<ADataType>, pk_i4_t>)
return 2;
else
return 1;
}()
ck::is_same_v
constexpr bool is_same_v
Definition: type.hpp:283

◆ BPackedSize

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
constexpr index_t ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::BPackedSize
staticconstexpr
Initial value:
= []() {
if constexpr(is_same_v<remove_cvref_t<BDataType>, pk_i4_t>)
return 2;
else
return 1;
}()

◆ NumDTensor

template<typename ALayout , typename BLayout , typename DsLayout , typename CLayout , typename ADataType , typename BDataType , typename DsDataType , typename CDataType , typename GemmAccDataType , typename CShuffleDataType , typename AElementwiseOperation , typename BElementwiseOperation , typename CElementwiseOperation , GemmSpecialization GemmSpec, index_t BlockSize, index_t MPerBlock, index_t NPerBlock, index_t KPerBlock, index_t AK1, index_t BK1, 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 ABlockLdsExtraM, typename BBlockTransferThreadClusterLengths_BK0_N_BK1 , typename BBlockTransferThreadClusterArrangeOrder , typename BBlockTransferSrcAccessOrder , index_t BBlockTransferSrcVectorDim, index_t BBlockTransferSrcScalarPerVector, index_t BBlockTransferDstScalarPerVector_BK1, bool BBlockLdsExtraN, index_t CShuffleMXdlPerWavePerShuffle, index_t CShuffleNXdlPerWavePerShuffle, typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock , typename CDEShuffleBlockTransferScalarPerVectors , BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave, BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1, index_t ActivationOP = 0, bool NSwizzle = false, bool IsInputGemm = true, bool MulRoutedWeight = true, bool PerTokenQuant = true, typename IndexType = index_t, typename ComputeTypeA = CDataType, typename ComputeTypeB = ComputeTypeA, typename LDSTypeA = ComputeTypeA, typename LDSTypeB = ComputeTypeB>
constexpr index_t ck::tensor_operation::device::DeviceMoeGemm< ALayout, BLayout, DsLayout, CLayout, ADataType, BDataType, DsDataType, CDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, CElementwiseOperation, GemmSpec, 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, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CDEShuffleBlockTransferScalarPerVectors, BlkGemmPipeSched, BlkGemmPipelineVer, ActivationOP, NSwizzle, IsInputGemm, MulRoutedWeight, PerTokenQuant, IndexType, ComputeTypeA, ComputeTypeB, LDSTypeA, LDSTypeB >::NumDTensor = DsDataType::Size()
staticconstexpr
The documentation for this struct was generated from the following file:
  • /home/docs/checkouts/readthedocs.org/user_builds/advanced-micro-devices-composable-kernel/checkouts/docs-7.0.1/include/ck/tensor_operation/gpu/device/impl/device_moe_gemm.hpp