CUTLASS: cutlass::transform::PitchLinearWarpStripedThreadMap< Shape_, Threads, WarpThreadArrangement_, ElementsPerAccess >::Detail Struct Reference

| | CUTLASS

CUDA Templates for Linear Algebra Subroutines and Solvers |

• Main Page

• Modules

• Namespaces

• Classes

• Files

• Class List

• Class Index

• Class Hierarchy

• Class Members

• cutlass

• transform

• PitchLinearWarpStripedThreadMap

• Detail

Public Types | Static Public Attributes | List of all members

cutlass::transform::PitchLinearWarpStripedThreadMap< Shape_, Threads, WarpThreadArrangement_, ElementsPerAccess >::Detail Struct Reference

Internal details made public to facilitate introspection Iterations along each dimension (concept: PitchLinearShape)

#include <pitch_linear_thread_map.h>

|

Public Types

| | using | WarpThreadArrangement = WarpThreadArrangement_ | | | Fixed arrangement of threads within a warp (units of threads). More...
| | | | using | ShapeInAccesses = layout::PitchLinearShape< Shape::kContiguous/kElementsPerAccess, Shape::kStrided > | | | Compute the 'shape' of the overall tile in units of vectors. More...
| | | | using | WarpAccessIterations = layout::PitchLinearShape< ShapeInAccesses::kContiguous/WarpThreadArrangement::kContiguous, ShapeInAccesses::kStrided/WarpThreadArrangement::kStrided > | | | | using | WarpArrangement = layout::PitchLinearShape< kWarpsContiguous, kWarpsStrided > | | | Arrangement of warps within a threadblock-scoped tile. More...
| | |

|

Static Public Attributes

| | static int const | kWarpSize = WarpThreadArrangement::kCount | | | Number of threads per warp. More...
| | | | static int const | kWarpCount = kThreads / kWarpSize | | | Number of participating warps. More...
| | | | static int const | kWarpsStrided | | | | static int const | kWarpsContiguous | | |

Member Typedef Documentation

template<typename Shape_ , int Threads, typename WarpThreadArrangement_ , int ElementsPerAccess = 1>

| using cutlass::transform::PitchLinearWarpStripedThreadMap< Shape_, Threads, WarpThreadArrangement_, ElementsPerAccess >::Detail::ShapeInAccesses = layout::PitchLinearShape< Shape::kContiguous / kElementsPerAccess, Shape::kStrided > |

template<typename Shape_ , int Threads, typename WarpThreadArrangement_ , int ElementsPerAccess = 1>

| using cutlass::transform::PitchLinearWarpStripedThreadMap< Shape_, Threads, WarpThreadArrangement_, ElementsPerAccess >::Detail::WarpAccessIterations = layout::PitchLinearShape< ShapeInAccesses::kContiguous / WarpThreadArrangement::kContiguous, ShapeInAccesses::kStrided / WarpThreadArrangement::kStrided > |

template<typename Shape_ , int Threads, typename WarpThreadArrangement_ , int ElementsPerAccess = 1>

| using cutlass::transform::PitchLinearWarpStripedThreadMap< Shape_, Threads, WarpThreadArrangement_, ElementsPerAccess >::Detail::WarpArrangement = layout::PitchLinearShape< kWarpsContiguous, kWarpsStrided > |

template<typename Shape_ , int Threads, typename WarpThreadArrangement_ , int ElementsPerAccess = 1>

| using cutlass::transform::PitchLinearWarpStripedThreadMap< Shape_, Threads, WarpThreadArrangement_, ElementsPerAccess >::Detail::WarpThreadArrangement = WarpThreadArrangement_ |

Member Data Documentation

template<typename Shape_ , int Threads, typename WarpThreadArrangement_ , int ElementsPerAccess = 1>

|

| int const cutlass::transform::PitchLinearWarpStripedThreadMap< Shape_, Threads, WarpThreadArrangement_, ElementsPerAccess >::Detail::kWarpCount = kThreads / kWarpSize |

| static |

template<typename Shape_ , int Threads, typename WarpThreadArrangement_ , int ElementsPerAccess = 1>

|

| int const cutlass::transform::PitchLinearWarpStripedThreadMap< Shape_, Threads, WarpThreadArrangement_, ElementsPerAccess >::Detail::kWarpsContiguous |

| static |

Initial value:

=

(kWarpCount > WarpAccessIterations::kStrided ?

WarpAccessIterations::kContiguous / kWarpsStrided : 1)

template<typename Shape_ , int Threads, typename WarpThreadArrangement_ , int ElementsPerAccess = 1>

|

| int const cutlass::transform::PitchLinearWarpStripedThreadMap< Shape_, Threads, WarpThreadArrangement_, ElementsPerAccess >::Detail::kWarpSize = WarpThreadArrangement::kCount |

| static |

template<typename Shape_ , int Threads, typename WarpThreadArrangement_ , int ElementsPerAccess = 1>

|

| int const cutlass::transform::PitchLinearWarpStripedThreadMap< Shape_, Threads, WarpThreadArrangement_, ElementsPerAccess >::Detail::kWarpsStrided |

| static |

Initial value:

=

(WarpAccessIterations::kStrided >= kWarpCount

? kWarpCount : (kWarpCount / WarpAccessIterations::kStrided))

---

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

• [pitch_linear_thread_map.h](pitch linear thread__map_8h_source.html)

---

Generated by 1.8.11