![]() |
CUTLASS
CUDA Templates for Linear Algebra Subroutines and Solvers
|
#include <regular_tile_access_iterator_pitch_linear.h>
Public Types | |
| using | Shape = Shape_ |
| using | Element = Element_ |
| using | Layout = layout::ColumnMajor |
| using | Index = typename Layout::Index |
| using | LongIndex = typename Layout::LongIndex |
| using | TensorRef = TensorRef< Element, Layout > |
| using | TensorCoord = typename Layout::TensorCoord |
| using | ThreadMap = ThreadMap_ |
| using | UnderlyingIterator = RegularTileAccessIterator< layout::PitchLinearShape< Shape::kRow, Shape::kColumn >, Element, layout::PitchLinear,(kAdvanceRank==0?0:1), ThreadMap_ > |
| Underlying iterator type. More... | |
| using | AccessType = typename UnderlyingIterator::AccessType |
Public Member Functions | |
| CUTLASS_HOST_DEVICE | RegularTileAccessIterator (TensorRef ref, int thread_id) |
| Construct a TileIterator with zero threadblock offset. More... | |
| CUTLASS_HOST_DEVICE void | set_iteration_index (int index) |
| Overrides the internal iteration index. More... | |
| CUTLASS_HOST_DEVICE void | add_pointer_offset (LongIndex pointer_offset) |
| Adds a pointer offset in units of Element. More... | |
| CUTLASS_HOST_DEVICE AccessType * | get () const |
| Returns a pointer. More... | |
| CUTLASS_DEVICE void | add_tile_offset (TensorCoord const &coord) |
| Adds a tile offset. More... | |
| CUTLASS_HOST_DEVICE RegularTileAccessIterator & | operator++ () |
| Advances to the next tile in memory. More... | |
| CUTLASS_HOST_DEVICE RegularTileAccessIterator | operator++ (int) |
| Advances to the next tile in memory. More... | |
Static Public Attributes | |
| static int const | kAdvanceRank = AdvanceRank |
| static int const | kAlignment = Alignment |
Tile iterator specialized for column major layouts
Satisfies: ForwardTileIteratorConcept | ReadableContiguousTileIteratorConcept | WriteableContiguousTileIteratorConcept
| using cutlass::transform::threadblock::RegularTileAccessIterator< Shape_, Element_, layout::ColumnMajor, AdvanceRank, ThreadMap_, Alignment >::AccessType = typename UnderlyingIterator::AccessType |
| using cutlass::transform::threadblock::RegularTileAccessIterator< Shape_, Element_, layout::ColumnMajor, AdvanceRank, ThreadMap_, Alignment >::Element = Element_ |
| using cutlass::transform::threadblock::RegularTileAccessIterator< Shape_, Element_, layout::ColumnMajor, AdvanceRank, ThreadMap_, Alignment >::Index = typename Layout::Index |
| using cutlass::transform::threadblock::RegularTileAccessIterator< Shape_, Element_, layout::ColumnMajor, AdvanceRank, ThreadMap_, Alignment >::Layout = layout::ColumnMajor |
| using cutlass::transform::threadblock::RegularTileAccessIterator< Shape_, Element_, layout::ColumnMajor, AdvanceRank, ThreadMap_, Alignment >::LongIndex = typename Layout::LongIndex |
| using cutlass::transform::threadblock::RegularTileAccessIterator< Shape_, Element_, layout::ColumnMajor, AdvanceRank, ThreadMap_, Alignment >::Shape = Shape_ |
| using cutlass::transform::threadblock::RegularTileAccessIterator< Shape_, Element_, layout::ColumnMajor, AdvanceRank, ThreadMap_, Alignment >::TensorCoord = typename Layout::TensorCoord |
| using cutlass::transform::threadblock::RegularTileAccessIterator< Shape_, Element_, layout::ColumnMajor, AdvanceRank, ThreadMap_, Alignment >::TensorRef = TensorRef<Element, Layout> |
| using cutlass::transform::threadblock::RegularTileAccessIterator< Shape_, Element_, layout::ColumnMajor, AdvanceRank, ThreadMap_, Alignment >::ThreadMap = ThreadMap_ |
| using cutlass::transform::threadblock::RegularTileAccessIterator< Shape_, Element_, layout::ColumnMajor, AdvanceRank, ThreadMap_, Alignment >::UnderlyingIterator = RegularTileAccessIterator< layout::PitchLinearShape<Shape::kRow, Shape::kColumn>, Element, layout::PitchLinear, (kAdvanceRank == 0 ? 0 : 1), ThreadMap_> |
|
inline |
| ref | Pointer to start of tensor |
| thread_id | ID of each participating thread |
|
inline |
|
inline |
|
inline |
|
inline |
|
inline |
|
inline |
|
static |
|
static |
1.8.11