Argument Struct Reference

Argument Struct Reference#

Composable Kernel: ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument Struct Reference

#include <device_reduce_threadwise.hpp>

Inheritance diagram for ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument:

Public Member Functions
	Argument (const std::array< index_t, Rank > inLengths, const std::array< index_t, Rank > inStrides, const std::array< index_t, NumDstDim > outLengths, const std::array< index_t, NumDstDim > outStrides, const std::array< int, NumReduceDim > reduceDims, double alpha, double beta, const InDataType in_dev, OutDataType out_dev, IndexDataType *out_index_dev, const InElementwiseOperation in_elementwise_op, const AccElementwiseOperation acc_elementwise_op)
Public Member Functions inherited from ck::tensor_operation::device::BaseArgument
	BaseArgument ()=default
	BaseArgument (const BaseArgument &)=default
BaseArgument &	operator= (const BaseArgument &)=default
virtual	~BaseArgument ()

Public Attributes
std::array< index_t, Rank >	inLengths_
std::array< index_t, Rank >	inStrides_
std::array< index_t, NumDstDim >	outLengths_
std::array< index_t, NumDstDim >	outStrides_
AccDataType	alpha_
AccDataType	beta_
const InDataType *	in_dev_
OutDataType *	out_dev_
IndexDataType *	out_index_dev_
InElementwiseOperation	in_elementwise_op_
AccElementwiseOperation	acc_elementwise_op_
index_t	invariant_lowest_length
index_t	reduce_lowest_length
long_index_t	invariant_total_length
long_index_t	reduce_total_length
int	numBlockTileIteration
size_t	gridSize
Public Attributes inherited from ck::tensor_operation::device::BaseArgument
void *	p_workspace_ = nullptr

Constructor & Destructor Documentation

◆ Argument()

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::Argument	(	const std::array< index_t, Rank >	inLengths,
		const std::array< index_t, Rank >	inStrides,
		const std::array< index_t, NumDstDim >	outLengths,
		const std::array< index_t, NumDstDim >	outStrides,
		const std::array< int, NumReduceDim >	reduceDims,
		double	alpha,
		double	beta,
		const InDataType *	in_dev,
		OutDataType *	out_dev,
		IndexDataType *	out_index_dev,
		const InElementwiseOperation	in_elementwise_op,
		const AccElementwiseOperation	acc_elementwise_op )

inline

Member Data Documentation

◆ acc_elementwise_op_

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

AccElementwiseOperation ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::acc_elementwise_op_

◆ alpha_

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

AccDataType ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::alpha_

◆ beta_

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

AccDataType ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::beta_

◆ gridSize

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

size_t ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::gridSize

◆ in_dev_

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

const InDataType* ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::in_dev_

◆ in_elementwise_op_

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

InElementwiseOperation ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::in_elementwise_op_

◆ inLengths_

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

std::array<index_t, Rank> ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::inLengths_

◆ inStrides_

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

std::array<index_t, Rank> ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::inStrides_

◆ invariant_lowest_length

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

index_t ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::invariant_lowest_length

◆ invariant_total_length

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

long_index_t ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::invariant_total_length

◆ numBlockTileIteration

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

int ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::numBlockTileIteration

◆ out_dev_

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

OutDataType* ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::out_dev_

◆ out_index_dev_

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

IndexDataType* ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::out_index_dev_

◆ outLengths_

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

std::array<index_t, NumDstDim> ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::outLengths_

◆ outStrides_

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

std::array<index_t, NumDstDim> ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::outStrides_

◆ reduce_lowest_length

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

index_t ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::reduce_lowest_length

◆ reduce_total_length

template<typename InDataType, typename AccDataType, typename OutDataType, index_t Rank, index_t NumReduceDim, typename ReduceOperation, typename InElementwiseOperation, typename AccElementwiseOperation, bool PropagateNan, bool OutputIndex, bool TransformIndexKtoGlobal, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>

long_index_t ck::tensor_operation::device::DeviceReduceThreadWise< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex, TransformIndexKtoGlobal, HaveIndexInputIfOutputIndex, BlockSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::Argument::reduce_total_length

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

device_reduce_threadwise.hpp

Argument Struct Reference

Argument Struct Reference#

Public Member Functions

Public Attributes

Constructor & Destructor Documentation

◆ Argument()

Member Data Documentation

◆ acc_elementwise_op_

◆ alpha_

◆ beta_

◆ gridSize

◆ in_dev_

◆ in_elementwise_op_

◆ inLengths_

◆ inStrides_

◆ invariant_lowest_length

◆ invariant_total_length

◆ numBlockTileIteration

◆ out_dev_

◆ out_index_dev_

◆ outLengths_

◆ outStrides_

◆ reduce_lowest_length

◆ reduce_total_length