DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize > Struct Template Reference

DeviceReduceMultiBlock&lt; InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize &gt; Struct Template Reference#

Composable Kernel: ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize > Struct Template Reference
Classes | Public Types | Public Member Functions | Static Public Member Functions | Static Public Attributes | List of all members
ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize > Struct Template Reference

#include <device_reduce_multiblock.hpp>

Inheritance diagram for ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >:
ck::tensor_operation::device::DeviceReduce< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex > ck::tensor_operation::device::BaseOperator

Classes

struct  Argument
 
struct  Invoker
 

Public Types

using IndexDataType = int32_t
 

Public Member Functions

bool IsSupportedArgument (const BaseArgument *p_arg) override
 
std::unique_ptr< BaseArgument > MakeArgumentPointer (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 void *in_dev, const void *in_index_dev, void *out_dev, void *out_index_dev, const InElementwiseOperation in_elementwise_op, const AccElementwiseOperation acc_elementwise_op) override
 
std::unique_ptr< BaseInvoker > MakeInvokerPointer () override
 
std::string GetTypeString () const override
 
- Public Member Functions inherited from ck::tensor_operation::device::DeviceReduce< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex >
virtual std::unique_ptr< BaseArgument > MakeArgumentPointer (const std::array< index_t, Rank > inLengths, const std::array< index_t, Rank > inStrides, const std::array< index_t, NumOutDim > outLengths, const std::array< index_t, NumOutDim > outStrides, const std::array< int, NumReduceDim > reduceDims, double alpha, double beta, const void *in_dev, const void *in_index_dev, void *out_dev, void *out_index_dev, const InElementwiseOperation in_elementwise_op, const AccElementwiseOperation acc_elementwise_op)=0
 
- 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 auto MakeSrc2dDescriptor (const std::array< index_t, Rank > &inLengths, const std::array< index_t, Rank > &inStrides, int blkGroupSize, int numBlockTileIteration)
 
static auto MakeDst1dDescriptor (const std::array< index_t, NumDstDim > &outLengths, const std::array< index_t, NumDstDim > &outStrides)
 
static auto MakeDst1dDescriptorForBufferSet (const std::array< index_t, NumDstDim > &outLengths, const std::array< index_t, NumDstDim > &outStrides)
 
static bool IsSupportedArgument (const Argument *pArg)
 

Static Public Attributes

static constexpr bool HaveIndexInput = OutputIndex && HaveIndexInputIfOutputIndex
 
static constexpr index_t NumInvariantDim = Rank - NumReduceDim
 
static constexpr index_t NumSrcDim = Rank
 
static constexpr index_t NumDstDim = (NumInvariantDim == 0) ? 1 : NumInvariantDim
 
static constexpr bool reduceAllDim = (NumInvariantDim == 0)
 
static constexpr bool use_multiblock
 
static constexpr index_t M_BlockTileSize = MThreadClusterSize * MThreadSliceSize
 
static constexpr index_t K_BlockTileSize = KThreadClusterSize * KThreadSliceSize
 
- Static Public Attributes inherited from ck::tensor_operation::device::DeviceReduce< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex >
static constexpr index_t NumOutDim = (Rank - NumReduceDim == 0) ? 1 : Rank - NumReduceDim
 

Member Typedef Documentation

◆ IndexDataType

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
using ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::IndexDataType = int32_t

Member Function Documentation

◆ GetTypeString()

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
std::string ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::GetTypeString ( ) const
inlineoverridevirtual

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

◆ IsSupportedArgument() [1/2]

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
static bool ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::IsSupportedArgument ( const Argument *  pArg)
inlinestatic

◆ IsSupportedArgument() [2/2]

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
bool ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::IsSupportedArgument ( const BaseArgument *  p_arg)
inlineoverridevirtual

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

◆ MakeArgumentPointer()

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
std::unique_ptr<BaseArgument> ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::MakeArgumentPointer ( 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 void *  in_dev,
const void *  in_index_dev,
void *  out_dev,
void *  out_index_dev,
const InElementwiseOperation  in_elementwise_op,
const AccElementwiseOperation  acc_elementwise_op 
)
inlineoverride

◆ MakeDst1dDescriptor()

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
static auto ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::MakeDst1dDescriptor ( const std::array< index_t, NumDstDim > &  outLengths,
const std::array< index_t, NumDstDim > &  outStrides 
)
inlinestatic

◆ MakeDst1dDescriptorForBufferSet()

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
static auto ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::MakeDst1dDescriptorForBufferSet ( const std::array< index_t, NumDstDim > &  outLengths,
const std::array< index_t, NumDstDim > &  outStrides 
)
inlinestatic

◆ MakeInvokerPointer()

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
std::unique_ptr<BaseInvoker> ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::MakeInvokerPointer ( )
inlineoverridevirtual

Implements ck::tensor_operation::device::DeviceReduce< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, PropagateNan, OutputIndex >.

◆ MakeSrc2dDescriptor()

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
static auto ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::MakeSrc2dDescriptor ( const std::array< index_t, Rank > &  inLengths,
const std::array< index_t, Rank > &  inStrides,
int  blkGroupSize,
int  numBlockTileIteration 
)
inlinestatic

Member Data Documentation

◆ HaveIndexInput

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
constexpr bool ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::HaveIndexInput = OutputIndex && HaveIndexInputIfOutputIndex
staticconstexpr

◆ K_BlockTileSize

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
constexpr index_t ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::K_BlockTileSize = KThreadClusterSize * KThreadSliceSize
staticconstexpr

◆ M_BlockTileSize

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
constexpr index_t ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::M_BlockTileSize = MThreadClusterSize * MThreadSliceSize
staticconstexpr

◆ NumDstDim

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
constexpr index_t ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::NumDstDim = (NumInvariantDim == 0) ? 1 : NumInvariantDim
staticconstexpr

◆ NumInvariantDim

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
constexpr index_t ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::NumInvariantDim = Rank - NumReduceDim
staticconstexpr

◆ NumSrcDim

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
constexpr index_t ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::NumSrcDim = Rank
staticconstexpr

◆ reduceAllDim

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
constexpr bool ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::reduceAllDim = (NumInvariantDim == 0)
staticconstexpr

◆ use_multiblock

template<typename InDataType , typename AccDataType , typename OutDataType , index_t Rank, index_t NumReduceDim, typename ReduceOperation , typename InElementwiseOperation , typename AccElementwiseOperation , InMemoryDataOperationEnum OutMemoryDataOperation, bool PropagateNan, bool OutputIndex, bool HaveIndexInputIfOutputIndex, index_t BlockSize, index_t MThreadClusterSize, index_t KThreadClusterSize, index_t MThreadSliceSize, index_t KThreadSliceSize, index_t InSrcVectorDim, index_t InSrcVectorSize, index_t OutDstVectorSize>
constexpr bool ck::tensor_operation::device::DeviceReduceMultiBlock< InDataType, AccDataType, OutDataType, Rank, NumReduceDim, ReduceOperation, InElementwiseOperation, AccElementwiseOperation, OutMemoryDataOperation, PropagateNan, OutputIndex, HaveIndexInputIfOutputIndex, BlockSize, MThreadClusterSize, KThreadClusterSize, MThreadSliceSize, KThreadSliceSize, InSrcVectorDim, InSrcVectorSize, OutDstVectorSize >::use_multiblock
staticconstexpr
Initial value:
=
(OutMemoryDataOperation == InMemoryDataOperationEnum::AtomicAdd)
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_reduce_multiblock.hpp