Cutlass
CUDA Templates for Linear Algebra Subroutines and Solvers
Public Types | Static Public Attributes | List of all members
cutlass::TileTraitsStandard< Tile_, Threads > Struct Template Reference

Chooses 'best' shape to enable warp raking along contiguous dimension if possible.

#include <tile_traits_standard.h>

Public Types

typedef Tile_ Tile
 Shape of tile. More...
 

Static Public Attributes

static int const kThreads = Threads
 Number of participating threads. More...
 
static int const kWarpSize = 32
 Hard-coded warp size. More...
 
static int const kWarpCount = kThreads / kWarpSize
 Number of participating warps. More...
 
static int const kAccessSize = 1
 By default, do not do scalar loads. More...
 

Member Typedef Documentation

◆ Tile

template<typename Tile_ , int Threads>
typedef Tile_ cutlass::TileTraitsStandard< Tile_, Threads >::Tile

Member Data Documentation

◆ kAccessSize

template<typename Tile_ , int Threads>
int const cutlass::TileTraitsStandard< Tile_, Threads >::kAccessSize = 1
static

◆ kThreads

template<typename Tile_ , int Threads>
int const cutlass::TileTraitsStandard< Tile_, Threads >::kThreads = Threads
static

◆ kWarpCount

template<typename Tile_ , int Threads>
int const cutlass::TileTraitsStandard< Tile_, Threads >::kWarpCount = kThreads / kWarpSize
static

◆ kWarpSize

template<typename Tile_ , int Threads>
int const cutlass::TileTraitsStandard< Tile_, Threads >::kWarpSize = 32
static

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