CUTLASS
CUDA Templates for Linear Algebra Subroutines and Solvers
Public Types | Static Public Attributes | List of all members
cutlass::gemm::warp::MmaVoltaTensorOpMultiplicandTileIterator< Shape_, Operand_, Element_, cutlass::layout::VoltaTensorOpMultiplicandCrosswise< sizeof_bits< Element_ >::value, KBlock >, InstructionShape_, OpDelta_, 32 >::Policy Struct Reference

Internal structure of iterator - made public to enable introspection.

#include <mma_tensor_op_tile_iterator_sm70.h>

Public Types

using LdsShape = layout::PitchLinearShape< 1, 32 >
 Shape of one individual LDS instruction. More...
 
using LdsIterations = layout::PitchLinearShape< 1, Shape::kStrided/32 >
 Number and arrangement of LDSM instructions. More...
 

Static Public Attributes

static int const kElementsPerAccess = 8
 Using LDS.128. More...
 
static int const kContiguousElementsPerLine = 4
 Contiguous elements per line. More...
 

Member Typedef Documentation

template<typename Shape_ , Operand Operand_, typename Element_ , typename InstructionShape_ , int OpDelta_, int KBlock>
using cutlass::gemm::warp::MmaVoltaTensorOpMultiplicandTileIterator< Shape_, Operand_, Element_, cutlass::layout::VoltaTensorOpMultiplicandCrosswise< sizeof_bits< Element_ >::value, KBlock >, InstructionShape_, OpDelta_, 32 >::Policy::LdsIterations = layout::PitchLinearShape<1, Shape::kStrided / 32>
template<typename Shape_ , Operand Operand_, typename Element_ , typename InstructionShape_ , int OpDelta_, int KBlock>
using cutlass::gemm::warp::MmaVoltaTensorOpMultiplicandTileIterator< Shape_, Operand_, Element_, cutlass::layout::VoltaTensorOpMultiplicandCrosswise< sizeof_bits< Element_ >::value, KBlock >, InstructionShape_, OpDelta_, 32 >::Policy::LdsShape = layout::PitchLinearShape<1, 32>

Member Data Documentation

template<typename Shape_ , Operand Operand_, typename Element_ , typename InstructionShape_ , int OpDelta_, int KBlock>
int const cutlass::gemm::warp::MmaVoltaTensorOpMultiplicandTileIterator< Shape_, Operand_, Element_, cutlass::layout::VoltaTensorOpMultiplicandCrosswise< sizeof_bits< Element_ >::value, KBlock >, InstructionShape_, OpDelta_, 32 >::Policy::kContiguousElementsPerLine = 4
static
template<typename Shape_ , Operand Operand_, typename Element_ , typename InstructionShape_ , int OpDelta_, int KBlock>
int const cutlass::gemm::warp::MmaVoltaTensorOpMultiplicandTileIterator< Shape_, Operand_, Element_, cutlass::layout::VoltaTensorOpMultiplicandCrosswise< sizeof_bits< Element_ >::value, KBlock >, InstructionShape_, OpDelta_, 32 >::Policy::kElementsPerAccess = 8
static

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