CUTLASS
CUDA Templates for Linear Algebra Subroutines and Solvers
|
Internal structure of iterator - made public to enable introspection.
#include <mma_tensor_op_tile_iterator_sm70.h>
Public Types | |
using | InterleavedTile = MatrixShape< 32, 32 > |
Volta Tensor Op uses 32x32 interleaved tile. More... | |
using | TileIterations = MatrixShape< Shape::kRow/InterleavedTile::kRow, Shape::kColumn/InterleavedTile::kColumn > |
Number of mma operations performed. More... | |
using | MmaIterations = MatrixShape< InterleavedTile::kRow/InstructionShape::kM, InterleavedTile::kColumn/InstructionShape::kN > |
using cutlass::gemm::warp::MmaVoltaTensorOpAccumulatorTileIterator< Shape_, Element_, Layout_, InstructionShape_, OpDelta_ >::Policy::InterleavedTile = MatrixShape<32, 32> |
using cutlass::gemm::warp::MmaVoltaTensorOpAccumulatorTileIterator< Shape_, Element_, Layout_, InstructionShape_, OpDelta_ >::Policy::MmaIterations = MatrixShape<InterleavedTile::kRow / InstructionShape::kM, InterleavedTile::kColumn / InstructionShape::kN> |
using cutlass::gemm::warp::MmaVoltaTensorOpAccumulatorTileIterator< Shape_, Element_, Layout_, InstructionShape_, OpDelta_ >::Policy::TileIterations = MatrixShape< Shape::kRow / InterleavedTile::kRow, Shape::kColumn / InterleavedTile::kColumn > |