cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail Struct Reference

#include <output_tile_thread_map.h>

Public Types
using	RowArrangement = detail::RowArrangement< Shape, kWarpsRemainingForRows, kElementsPerAccess, kElementSize,(Shape::kRow > kWarpsRemainingForRows) >
using	WarpPartitions = OutputTileShape< RowArrangement::kWarpPartitionsColumn, RowArrangement::kWarpPartitionsRow, kWarpPartitionsGroup, kWarpPartitionsCluster, 1 >

Static Public Attributes
static int const	kIterationsCluster
static int const	kDeltaCluster
static int const	kCompactedDeltaCluster
static int const	kWarpPartitionsCluster
static int const	kWarpsRemainingForGroups
static int const	kIterationsGroup
static int const	kDeltaGroup
static int const	kCompactedDeltaGroup
static int const	kWarpPartitionsGroup
static int const	kWarpsRemainingForRows
static int const	kAccessWidth = RowArrangement::kAccessWidth
static int const	kAccessRows = RowArrangement::kAccessRows

Member Typedef Documentation

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

using cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::RowArrangement = detail::RowArrangement< Shape, kWarpsRemainingForRows, kElementsPerAccess, kElementSize, (Shape::kRow > kWarpsRemainingForRows) >

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

using cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::WarpPartitions = OutputTileShape< RowArrangement::kWarpPartitionsColumn, RowArrangement::kWarpPartitionsRow, kWarpPartitionsGroup, kWarpPartitionsCluster, 1>

Member Data Documentation

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

int const cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::kAccessRows = RowArrangement::kAccessRows

static

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

int const cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::kAccessWidth = RowArrangement::kAccessWidth

static

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

int const cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::kCompactedDeltaCluster

static

Initial value:

=
      ((Shape::kCluster > kWarpCount) ?
        Shape::kRow * Shape::kGroup * Shape::kCluster / kIterationsCluster
        : 1)

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

int const cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::kCompactedDeltaGroup

static

Initial value:

=
      ((Shape::kGroup > kWarpsRemainingForGroups) ?
        Shape::kRow * Shape::kGroup / kIterationsGroup
        : 1)

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

int const cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::kDeltaCluster

static

Initial value:

=
      ((Shape::kCluster > kWarpCount) ?
        Shape::kRow * Count::kRow * Shape::kGroup * Count::kGroup * Shape::kCluster / kIterationsCluster
        : 1)

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

int const cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::kDeltaGroup

static

Initial value:

=
      ((Shape::kGroup > kWarpsRemainingForGroups) ?
        Shape::kRow * Count::kRow * Shape::kGroup / kIterationsGroup
        : 1)

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

int const cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::kIterationsCluster

static

Initial value:

= 
      ((Shape::kCluster > kWarpCount) ?
        Shape::kCluster / kWarpCount
        : 1)

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

int const cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::kIterationsGroup

static

Initial value:

=
      ((Shape::kGroup > kWarpsRemainingForGroups) ?
        Shape::kGroup / kWarpsRemainingForGroups
        : 1)

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

int const cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::kWarpPartitionsCluster

static

Initial value:

=
      ((Shape::kCluster > kWarpCount) ?
        kWarpCount
        : kWarpCount / Shape::kCluster)

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

int const cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::kWarpPartitionsGroup

static

Initial value:

=
      ((Shape::kGroup > kWarpsRemainingForGroups) ?
        1
        : kWarpsRemainingForGroups / Shape::kGroup)

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

int const cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::kWarpsRemainingForGroups

static

Initial value:

=
      ((Shape::kCluster > kWarpCount) ? 1 : kWarpCount / Shape::kCluster)

template<typename Shape_ , typename Count_ , int Threads, int ElementsPerAccess, int ElementSize>

int const cutlass::epilogue::threadblock::OutputTileOptimalThreadMap< Shape_, Count_, Threads, ElementsPerAccess, ElementSize >::Detail::kWarpsRemainingForRows

static

Initial value:

=
      ((Shape::kGroup > kWarpsRemainingForGroups) ?
        1
        : kWarpsRemainingForGroups / Shape::kGroup)

---

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

• output_tile_thread_map.h