|
Cutlass
CUDA Templates for Linear Algebra Subroutines and Solvers
|
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... | |
| typedef Tile_ cutlass::TileTraitsStandard< Tile_, Threads >::Tile |
|
static |
|
static |
|
static |
|
static |
1.8.14