DeviceBatchedGemmSoftmaxGemmPermute< NumDimG, NumDimM, NumDimN, NumDimK, NumDimO, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc1BiasDataType, AElementwiseOperation, B0ElementwiseOperation, C0DEElementwiseOperation, B1ElementwiseOperation, C1DEElementwiseOperation, MaskingSpec > Struct Template Reference

DeviceBatchedGemmSoftmaxGemmPermute&lt; NumDimG, NumDimM, NumDimN, NumDimK, NumDimO, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc1BiasDataType, AElementwiseOperation, B0ElementwiseOperation, C0DEElementwiseOperation, B1ElementwiseOperation, C1DEElementwiseOperation, MaskingSpec &gt; Struct Template Reference#

Composable Kernel: ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute< NumDimG, NumDimM, NumDimN, NumDimK, NumDimO, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc1BiasDataType, AElementwiseOperation, B0ElementwiseOperation, C0DEElementwiseOperation, B1ElementwiseOperation, C1DEElementwiseOperation, MaskingSpec > Struct Template Reference
Public Member Functions | Static Public Attributes | List of all members
ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute< NumDimG, NumDimM, NumDimN, NumDimK, NumDimO, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc1BiasDataType, AElementwiseOperation, B0ElementwiseOperation, C0DEElementwiseOperation, B1ElementwiseOperation, C1DEElementwiseOperation, MaskingSpec > Struct Template Referenceabstract

#include <device_batched_gemm_softmax_gemm_permute.hpp>

Inheritance diagram for ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute< NumDimG, NumDimM, NumDimN, NumDimK, NumDimO, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc1BiasDataType, AElementwiseOperation, B0ElementwiseOperation, C0DEElementwiseOperation, B1ElementwiseOperation, C1DEElementwiseOperation, MaskingSpec >:
ck::tensor_operation::device::BaseOperator

Public Member Functions

virtual std::unique_ptr< BaseArgument > MakeArgumentPointer (const void *p_a, const void *p_b0, const void *p_b1, void *p_c, const std::array< void *, NumAcc0Bias > p_acc0_biases, const std::array< void *, NumAcc1Bias > p_acc1_biases, const std::vector< index_t > &a_gs_ms_ks_lengths, const std::vector< index_t > &a_gs_ms_ks_strides, const std::vector< index_t > &b_gs_ns_ks_lengths, const std::vector< index_t > &b_gs_ns_ks_strides, const std::vector< index_t > &b1_gs_gemm1ns_gemm1ks_lengths, const std::vector< index_t > &b1_gs_gemm1ns_gemm1ks_strides, const std::vector< index_t > &c_gs_ms_gemm1ns_lengths, const std::vector< index_t > &c_gs_ms_gemm1ns_strides, const std::array< std::vector< index_t >, NumAcc0Bias > acc0_biases_gs_ms_ns_lengths, const std::array< std::vector< index_t >, NumAcc0Bias > acc0_biases_gs_ms_ns_strides, const std::array< std::vector< index_t >, NumAcc1Bias > acc1_biases_gs_ms_gemm1ns_lengths, const std::array< std::vector< index_t >, NumAcc1Bias > acc1_biases_gs_ms_gemm1ns_strides, AElementwiseOperation a_element_op, B0ElementwiseOperation b0_element_op, C0DEElementwiseOperation c0de_element_op, B1ElementwiseOperation b1_element_op, C1DEElementwiseOperation c1de_element_op)=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 NumAcc0Bias = Acc0BiasDataType::Size()
static constexpr index_t NumAcc1Bias = Acc1BiasDataType::Size()

Member Function Documentation

◆ MakeArgumentPointer()

template<index_t NumDimG, index_t NumDimM, index_t NumDimN, index_t NumDimK, index_t NumDimO, typename ADataType, typename B0DataType, typename B1DataType, typename CDataType, typename Acc0BiasDataType, typename Acc1BiasDataType, typename AElementwiseOperation, typename B0ElementwiseOperation, typename C0DEElementwiseOperation, typename B1ElementwiseOperation, typename C1DEElementwiseOperation, MaskingSpecialization MaskingSpec>
virtual std::unique_ptr< BaseArgument > ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute< NumDimG, NumDimM, NumDimN, NumDimK, NumDimO, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc1BiasDataType, AElementwiseOperation, B0ElementwiseOperation, C0DEElementwiseOperation, B1ElementwiseOperation, C1DEElementwiseOperation, MaskingSpec >::MakeArgumentPointer ( const void * p_a,
const void * p_b0,
const void * p_b1,
void * p_c,
const std::array< void *, NumAcc0Bias > p_acc0_biases,
const std::array< void *, NumAcc1Bias > p_acc1_biases,
const std::vector< index_t > & a_gs_ms_ks_lengths,
const std::vector< index_t > & a_gs_ms_ks_strides,
const std::vector< index_t > & b_gs_ns_ks_lengths,
const std::vector< index_t > & b_gs_ns_ks_strides,
const std::vector< index_t > & b1_gs_gemm1ns_gemm1ks_lengths,
const std::vector< index_t > & b1_gs_gemm1ns_gemm1ks_strides,
const std::vector< index_t > & c_gs_ms_gemm1ns_lengths,
const std::vector< index_t > & c_gs_ms_gemm1ns_strides,
const std::array< std::vector< index_t >, NumAcc0Bias > acc0_biases_gs_ms_ns_lengths,
const std::array< std::vector< index_t >, NumAcc0Bias > acc0_biases_gs_ms_ns_strides,
const std::array< std::vector< index_t >, NumAcc1Bias > acc1_biases_gs_ms_gemm1ns_lengths,
const std::array< std::vector< index_t >, NumAcc1Bias > acc1_biases_gs_ms_gemm1ns_strides,
AElementwiseOperation a_element_op,
B0ElementwiseOperation b0_element_op,
C0DEElementwiseOperation c0de_element_op,
B1ElementwiseOperation b1_element_op,
C1DEElementwiseOperation c1de_element_op )
pure virtual

Implemented in ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute_Wmma_CShuffle< NumDimG, NumDimM, NumDimL, NumDimK, NumDimN, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc0DataType, Acc1BiasDataType, Acc1DataType, CShuffleDataType, AElementwiseOperation, B0ElementwiseOperation, AccElementwiseOperation, B1ElementwiseOperation, CElementwiseOperation, GemmSpec, ASpec, B0Spec, B1Spec, CSpec, NumPrefetch, BlockSize, MPerBlock, LPerBlock, KPerBlock, AK1, BK1, NPerBlock, LTilePerBlock, L1, MPerWmma, LPerWmma, NPerWmma, MRepeat, LRepeat, NRepeat, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, B0BlockTransferThreadClusterLengths_K0_L_K1, B0BlockTransferThreadClusterArrangeOrder, B0BlockTransferSrcAccessOrder, B0BlockTransferSrcVectorDim, B0BlockTransferSrcScalarPerVector, B0BlockTransferDstScalarPerVector_K1, B0BlockLdsAddExtraL, B1BlockTransferThreadClusterLengths_L0_N_L1, B1BlockTransferThreadClusterArrangeOrder, B1BlockTransferSrcAccessOrder, B1BlockTransferSrcVectorDim, B1BlockTransferSrcScalarPerVector, B1BlockTransferDstScalarPerVector_L1, B1BlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CShuffleBlockTransferScalarPerVector_NPerBlock, MaskingSpec, LoopSched, PipelineVer >, ck::tensor_operation::device::DeviceGroupedQueryAttentionForward_Wmma< NumDimG, NumDimM, NumDimL, NumDimK, NumDimN, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc0DataType, Acc1BiasDataType, Acc1DataType, CShuffleDataType, AElementwiseOperation, B0ElementwiseOperation, AccElementwiseOperation, B1ElementwiseOperation, CElementwiseOperation, GemmSpec, ASpec, B0Spec, B1Spec, CSpec, NumPrefetch, QueryGroupNumber, BlockSize, MPerBlock, LPerBlock, KPerBlock, AK1, BK1, NPerBlock, LTilePerBlock, L1, MPerWmma, LPerWmma, NPerWmma, MRepeat, LRepeat, NRepeat, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, B0BlockTransferThreadClusterLengths_K0_L_K1, B0BlockTransferThreadClusterArrangeOrder, B0BlockTransferSrcAccessOrder, B0BlockTransferSrcVectorDim, B0BlockTransferSrcScalarPerVector, B0BlockTransferDstScalarPerVector_K1, B0BlockLdsAddExtraL, B1BlockTransferThreadClusterLengths_L0_N_L1, B1BlockTransferThreadClusterArrangeOrder, B1BlockTransferSrcAccessOrder, B1BlockTransferSrcVectorDim, B1BlockTransferSrcScalarPerVector, B1BlockTransferDstScalarPerVector_L1, B1BlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CShuffleBlockTransferScalarPerVector_NPerBlock, MaskingSpec, LoopSched, PipelineVer >, and ck::tensor_operation::device::DeviceMultiQueryAttentionForward_Wmma< NumDimG, NumDimM, NumDimL, NumDimK, NumDimN, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc0DataType, Acc1BiasDataType, Acc1DataType, CShuffleDataType, AElementwiseOperation, B0ElementwiseOperation, AccElementwiseOperation, B1ElementwiseOperation, CElementwiseOperation, GemmSpec, ASpec, B0Spec, B1Spec, CSpec, NumPrefetch, BlockSize, MPerBlock, LPerBlock, KPerBlock, AK1, BK1, NPerBlock, LTilePerBlock, L1, MPerWmma, LPerWmma, NPerWmma, MRepeat, LRepeat, NRepeat, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, B0BlockTransferThreadClusterLengths_K0_L_K1, B0BlockTransferThreadClusterArrangeOrder, B0BlockTransferSrcAccessOrder, B0BlockTransferSrcVectorDim, B0BlockTransferSrcScalarPerVector, B0BlockTransferDstScalarPerVector_K1, B0BlockLdsAddExtraL, B1BlockTransferThreadClusterLengths_L0_N_L1, B1BlockTransferThreadClusterArrangeOrder, B1BlockTransferSrcAccessOrder, B1BlockTransferSrcVectorDim, B1BlockTransferSrcScalarPerVector, B1BlockTransferDstScalarPerVector_L1, B1BlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CShuffleBlockTransferScalarPerVector_NPerBlock, MaskingSpec, LoopSched, PipelineVer >.

◆ MakeInvokerPointer()

template<index_t NumDimG, index_t NumDimM, index_t NumDimN, index_t NumDimK, index_t NumDimO, typename ADataType, typename B0DataType, typename B1DataType, typename CDataType, typename Acc0BiasDataType, typename Acc1BiasDataType, typename AElementwiseOperation, typename B0ElementwiseOperation, typename C0DEElementwiseOperation, typename B1ElementwiseOperation, typename C1DEElementwiseOperation, MaskingSpecialization MaskingSpec>
virtual std::unique_ptr< BaseInvoker > ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute< NumDimG, NumDimM, NumDimN, NumDimK, NumDimO, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc1BiasDataType, AElementwiseOperation, B0ElementwiseOperation, C0DEElementwiseOperation, B1ElementwiseOperation, C1DEElementwiseOperation, MaskingSpec >::MakeInvokerPointer ( )
pure virtual

Implemented in ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute_Wmma_CShuffle< NumDimG, NumDimM, NumDimL, NumDimK, NumDimN, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc0DataType, Acc1BiasDataType, Acc1DataType, CShuffleDataType, AElementwiseOperation, B0ElementwiseOperation, AccElementwiseOperation, B1ElementwiseOperation, CElementwiseOperation, GemmSpec, ASpec, B0Spec, B1Spec, CSpec, NumPrefetch, BlockSize, MPerBlock, LPerBlock, KPerBlock, AK1, BK1, NPerBlock, LTilePerBlock, L1, MPerWmma, LPerWmma, NPerWmma, MRepeat, LRepeat, NRepeat, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, B0BlockTransferThreadClusterLengths_K0_L_K1, B0BlockTransferThreadClusterArrangeOrder, B0BlockTransferSrcAccessOrder, B0BlockTransferSrcVectorDim, B0BlockTransferSrcScalarPerVector, B0BlockTransferDstScalarPerVector_K1, B0BlockLdsAddExtraL, B1BlockTransferThreadClusterLengths_L0_N_L1, B1BlockTransferThreadClusterArrangeOrder, B1BlockTransferSrcAccessOrder, B1BlockTransferSrcVectorDim, B1BlockTransferSrcScalarPerVector, B1BlockTransferDstScalarPerVector_L1, B1BlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CShuffleBlockTransferScalarPerVector_NPerBlock, MaskingSpec, LoopSched, PipelineVer >, ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute_Xdl_CShuffle< NumDimG, NumDimM, NumDimN, NumDimK, NumDimO, ADataType, BDataType, B1DataType, CDataType, D0sDataType, D1sDataType, GemmAccDataType, CShuffleDataType, AElementwiseOperation, BElementwiseOperation, C0DEElementwiseOperation, B1ElementwiseOperation, C1DEElementwiseOperation, GemmSpec, ASpec, BSpec, B1Spec, CSpec, NumGemmKPrefetchStage, BlockSize, MPerBlock, NPerBlock, KPerBlock, Gemm1NPerBlock, Gemm1KPerBlock, AK1, BK1, B1K1, MPerXDL, NPerXDL, MXdlPerWave, NXdlPerWave, Gemm1NXdlPerWave, ABlockTransferThreadClusterLengths_AK0_M_AK1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_AK1, ABlockLdsExtraM, BBlockTransferThreadClusterLengths_BK0_N_BK1, BBlockTransferThreadClusterArrangeOrder, BBlockTransferSrcAccessOrder, BBlockTransferSrcVectorDim, BBlockTransferSrcScalarPerVector, BBlockTransferDstScalarPerVector_BK1, BBlockLdsExtraN, B1BlockTransferThreadClusterLengths_BK0_N_BK1, B1BlockTransferThreadClusterArrangeOrder, B1BlockTransferSrcAccessOrder, B1BlockTransferSrcVectorDim, B1BlockTransferSrcScalarPerVector, B1BlockTransferDstScalarPerVector_BK1, B1BlockLdsExtraN, CShuffleMXdlPerWavePerShuffle, CShuffleNXdlPerWavePerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CShuffleBlockTransferScalarPerVector_NPerBlock, MaskingSpec, D0sTransferSrcScalarPerVector, LoopSched >, ck::tensor_operation::device::DeviceGroupedQueryAttentionForward_Wmma< NumDimG, NumDimM, NumDimL, NumDimK, NumDimN, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc0DataType, Acc1BiasDataType, Acc1DataType, CShuffleDataType, AElementwiseOperation, B0ElementwiseOperation, AccElementwiseOperation, B1ElementwiseOperation, CElementwiseOperation, GemmSpec, ASpec, B0Spec, B1Spec, CSpec, NumPrefetch, QueryGroupNumber, BlockSize, MPerBlock, LPerBlock, KPerBlock, AK1, BK1, NPerBlock, LTilePerBlock, L1, MPerWmma, LPerWmma, NPerWmma, MRepeat, LRepeat, NRepeat, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, B0BlockTransferThreadClusterLengths_K0_L_K1, B0BlockTransferThreadClusterArrangeOrder, B0BlockTransferSrcAccessOrder, B0BlockTransferSrcVectorDim, B0BlockTransferSrcScalarPerVector, B0BlockTransferDstScalarPerVector_K1, B0BlockLdsAddExtraL, B1BlockTransferThreadClusterLengths_L0_N_L1, B1BlockTransferThreadClusterArrangeOrder, B1BlockTransferSrcAccessOrder, B1BlockTransferSrcVectorDim, B1BlockTransferSrcScalarPerVector, B1BlockTransferDstScalarPerVector_L1, B1BlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CShuffleBlockTransferScalarPerVector_NPerBlock, MaskingSpec, LoopSched, PipelineVer >, and ck::tensor_operation::device::DeviceMultiQueryAttentionForward_Wmma< NumDimG, NumDimM, NumDimL, NumDimK, NumDimN, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc0DataType, Acc1BiasDataType, Acc1DataType, CShuffleDataType, AElementwiseOperation, B0ElementwiseOperation, AccElementwiseOperation, B1ElementwiseOperation, CElementwiseOperation, GemmSpec, ASpec, B0Spec, B1Spec, CSpec, NumPrefetch, BlockSize, MPerBlock, LPerBlock, KPerBlock, AK1, BK1, NPerBlock, LTilePerBlock, L1, MPerWmma, LPerWmma, NPerWmma, MRepeat, LRepeat, NRepeat, ABlockTransferThreadClusterLengths_K0_M_K1, ABlockTransferThreadClusterArrangeOrder, ABlockTransferSrcAccessOrder, ABlockTransferSrcVectorDim, ABlockTransferSrcScalarPerVector, ABlockTransferDstScalarPerVector_K1, ABlockLdsAddExtraM, B0BlockTransferThreadClusterLengths_K0_L_K1, B0BlockTransferThreadClusterArrangeOrder, B0BlockTransferSrcAccessOrder, B0BlockTransferSrcVectorDim, B0BlockTransferSrcScalarPerVector, B0BlockTransferDstScalarPerVector_K1, B0BlockLdsAddExtraL, B1BlockTransferThreadClusterLengths_L0_N_L1, B1BlockTransferThreadClusterArrangeOrder, B1BlockTransferSrcAccessOrder, B1BlockTransferSrcVectorDim, B1BlockTransferSrcScalarPerVector, B1BlockTransferDstScalarPerVector_L1, B1BlockLdsAddExtraN, CShuffleMRepeatPerShuffle, CShuffleNRepeatPerShuffle, CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock, CShuffleBlockTransferScalarPerVector_NPerBlock, MaskingSpec, LoopSched, PipelineVer >.

Member Data Documentation

◆ NumAcc0Bias

template<index_t NumDimG, index_t NumDimM, index_t NumDimN, index_t NumDimK, index_t NumDimO, typename ADataType, typename B0DataType, typename B1DataType, typename CDataType, typename Acc0BiasDataType, typename Acc1BiasDataType, typename AElementwiseOperation, typename B0ElementwiseOperation, typename C0DEElementwiseOperation, typename B1ElementwiseOperation, typename C1DEElementwiseOperation, MaskingSpecialization MaskingSpec>
index_t ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute< NumDimG, NumDimM, NumDimN, NumDimK, NumDimO, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc1BiasDataType, AElementwiseOperation, B0ElementwiseOperation, C0DEElementwiseOperation, B1ElementwiseOperation, C1DEElementwiseOperation, MaskingSpec >::NumAcc0Bias = Acc0BiasDataType::Size()
staticconstexpr

◆ NumAcc1Bias

template<index_t NumDimG, index_t NumDimM, index_t NumDimN, index_t NumDimK, index_t NumDimO, typename ADataType, typename B0DataType, typename B1DataType, typename CDataType, typename Acc0BiasDataType, typename Acc1BiasDataType, typename AElementwiseOperation, typename B0ElementwiseOperation, typename C0DEElementwiseOperation, typename B1ElementwiseOperation, typename C1DEElementwiseOperation, MaskingSpecialization MaskingSpec>
index_t ck::tensor_operation::device::DeviceBatchedGemmSoftmaxGemmPermute< NumDimG, NumDimM, NumDimN, NumDimK, NumDimO, ADataType, B0DataType, B1DataType, CDataType, Acc0BiasDataType, Acc1BiasDataType, AElementwiseOperation, B0ElementwiseOperation, C0DEElementwiseOperation, B1ElementwiseOperation, C1DEElementwiseOperation, MaskingSpec >::NumAcc1Bias = Acc1BiasDataType::Size()
staticconstexpr
The documentation for this struct was generated from the following file: