![]() |
CUTLASS
CUDA Templates for Linear Algebra Subroutines and Solvers
|
Template mapping a row-major view of pitch-linear memory to VoltaTensorOpMultiplicandCongruous.
#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 = MatrixCoord |
| Logical coordinate. More... | |
| using | Stride = Coord< kStrideRank, Index, LongIndex > |
| Stride vector. More... | |
| using | Base = VoltaTensorOpMultiplicandBCongruous< ElementSize > |
| using | TileShape = typename Base::TileShape |
| using | PartitionShape = typename Base::PartitionShape |
| using | PartitionCount = typename Base::PartitionCount |
| using | AccessCount = typename Base::AccessCount |
Public Member Functions | |
| CUTLASS_HOST_DEVICE | RowMajorVoltaTensorOpMultiplicandBCongruous (Index ldm=0) |
| Ctor. More... | |
| CUTLASS_HOST_DEVICE | RowMajorVoltaTensorOpMultiplicandBCongruous (Stride stride) |
| Ctor. More... | |
| CUTLASS_HOST_DEVICE LongIndex | operator() (TensorCoord const &coord) const |
| CUTLASS_HOST_DEVICE TensorCoord | inverse (LongIndex offset) const |
| Inverse of layout function, mapping linear offset to logical coordinate. More... | |
| CUTLASS_HOST_DEVICE Stride | stride () const |
| Returns the stride of the layout. More... | |
| CUTLASS_HOST_DEVICE Stride & | stride () |
| 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 RowMajorVoltaTensorOpMultiplicandBCongruous | 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 = Base::kAccessSize |
| This layout is optimized for 128b accesses. More... | |
| static int const | kElementSize = Base::kElementSize |
| static int const | kElementsPerAccess = Base::kElementsPerAccess |
| using cutlass::layout::RowMajorVoltaTensorOpMultiplicandBCongruous< ElementSize >::AccessCount = typename Base::AccessCount |
| using cutlass::layout::RowMajorVoltaTensorOpMultiplicandBCongruous< ElementSize >::Base = VoltaTensorOpMultiplicandBCongruous<ElementSize> |
| using cutlass::layout::RowMajorVoltaTensorOpMultiplicandBCongruous< ElementSize >::Index = int32_t |
| using cutlass::layout::RowMajorVoltaTensorOpMultiplicandBCongruous< ElementSize >::LongIndex = int64_t |
| using cutlass::layout::RowMajorVoltaTensorOpMultiplicandBCongruous< ElementSize >::PartitionCount = typename Base::PartitionCount |
| using cutlass::layout::RowMajorVoltaTensorOpMultiplicandBCongruous< ElementSize >::PartitionShape = typename Base::PartitionShape |
| using cutlass::layout::RowMajorVoltaTensorOpMultiplicandBCongruous< ElementSize >::Stride = Coord<kStrideRank, Index, LongIndex> |
| using cutlass::layout::RowMajorVoltaTensorOpMultiplicandBCongruous< ElementSize >::TensorCoord = MatrixCoord |
| using cutlass::layout::RowMajorVoltaTensorOpMultiplicandBCongruous< ElementSize >::TileShape = typename Base::TileShape |
|
inline |
|
inline |
|
inline |
|
inline |
|
inline |
Returns the offset of a coordinate in linear memory. Assumes coordinate has convention (contiguous, strided)
|
inlinestatic |
|
inline |
|
inline |
|
static |
|
static |
|
static |
|
static |
|
static |
1.8.11