CUTLASS
CUDA Templates for Linear Algebra Subroutines and Solvers
Public Types | Public Member Functions | Static Public Member Functions | Static Public Attributes | List of all members
cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize > Struct Template Reference

Template based on element size (in bits) - defined in terms of pitch-linear memory.

#include <tensor_op_multiplicand_sm70.h>

Public Types

using Index = int32_t
 Index type used for coordinates. More...
 
using LongIndex = int64_t
 Long index type used for offsets. More...
 
using TensorCoord = PitchLinearCoord
 Logical coordinate. More...
 
using Stride = Coord< kStrideRank, Index, LongIndex >
 Stride vector. More...
 
using TileShape = PitchLinearShape< 8, 4 >
 Fundamental tile shape in units of vectors. More...
 
using PartitionShape = PitchLinearShape< 8, 2 >
 Fundamental partition shape in units of vectors. More...
 
using PartitionCount = PitchLinearShape< TileShape::kContiguous/PartitionShape::kContiguous, TileShape::kStrided/PartitionShape::kStrided >
 
using AccessCount = PitchLinearShape< PartitionShape::kContiguous, PartitionShape::kStrided >
 

Public Member Functions

CUTLASS_HOST_DEVICE VoltaTensorOpMultiplicandCongruous (Index ldm=0)
 Ctor. More...
 
CUTLASS_HOST_DEVICE VoltaTensorOpMultiplicandCongruous (Stride stride)
 Ctor. More...
 
CUTLASS_HOST_DEVICE LongIndex operator() (TensorCoord const &coord) const
 
CUTLASS_HOST_DEVICE Stride stride () const
 Returns the stride of the layout. More...
 
CUTLASS_HOST_DEVICE Stridestride ()
 Returns the stride of the layout. More...
 
CUTLASS_HOST_DEVICE LongIndex capacity (TensorCoord const &extent) const
 Compute the number of contiguous elements needed to store a tensor with the given size. More...
 

Static Public Member Functions

static CUTLASS_HOST_DEVICE VoltaTensorOpMultiplicandCongruous packed (TensorCoord const &extent)
 Helper returns a layout to a tightly packed tensor. More...
 

Static Public Attributes

static int const kRank = 2
 Logical rank of tensor. More...
 
static int const kStrideRank = 1
 Rank of stride vector. More...
 
static int const kAccessSize = 128
 This layout is optimized for 128b accesses. More...
 
static int const kElementSize = ElementSize
 
static int const kElementsPerAccess = kAccessSize / kElementSize
 

Member Typedef Documentation

template<int ElementSize>
using cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::Index = int32_t
template<int ElementSize>
using cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::LongIndex = int64_t
template<int ElementSize>
using cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::PartitionShape = PitchLinearShape<8, 2>
template<int ElementSize>
using cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::Stride = Coord<kStrideRank, Index, LongIndex>
template<int ElementSize>
using cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::TensorCoord = PitchLinearCoord
template<int ElementSize>
using cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::TileShape = PitchLinearShape<8, 4>

Constructor & Destructor Documentation

Member Function Documentation

template<int ElementSize>
CUTLASS_HOST_DEVICE LongIndex cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::capacity ( TensorCoord const &  extent) const
inline
template<int ElementSize>
CUTLASS_HOST_DEVICE LongIndex cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::operator() ( TensorCoord const &  coord) const
inline

Returns the offset of a coordinate in linear memory. Assumes coordinate has convention (contiguous, strided)

template<int ElementSize>
static CUTLASS_HOST_DEVICE VoltaTensorOpMultiplicandCongruous cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::packed ( TensorCoord const &  extent)
inlinestatic
template<int ElementSize>
CUTLASS_HOST_DEVICE Stride cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::stride ( ) const
inline
template<int ElementSize>
CUTLASS_HOST_DEVICE Stride& cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::stride ( )
inline

Member Data Documentation

template<int ElementSize>
int const cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::kAccessSize = 128
static
template<int ElementSize>
int const cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::kElementSize = ElementSize
static
template<int ElementSize>
int const cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::kElementsPerAccess = kAccessSize / kElementSize
static
template<int ElementSize>
int const cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::kRank = 2
static
template<int ElementSize>
int const cutlass::layout::VoltaTensorOpMultiplicandCongruous< ElementSize >::kStrideRank = 1
static

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