CUTLASS
CUDA Templates for Linear Algebra Subroutines and Solvers
Public Types | Static Public Attributes | List of all members
cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ > Struct Template Reference

#include <default_mma_core_sm75.h>

Public Types

using Shape = Shape_
 
using WarpShape = WarpShape_
 
using InstructionShape = InstructionShape_
 
using ElementA = ElementA_
 
using LayoutA = layout::ColumnMajor
 
using ElementB = ElementB_
 
using LayoutB = layout::ColumnMajor
 
using ElementC = ElementC_
 
using LayoutC = LayoutC_
 
using OperatorClass = arch::OpClassTensorOp
 
using WarpCount = GemmShape< Shape::kM/WarpShape::kM, Shape::kN/WarpShape::kN, Shape::kK/WarpShape::kK >
 Number of warps present. More...
 
using Operator = Operator_
 Default Operator. More...
 
using SmemLayoutA = layout::ColumnMajorTensorOpMultiplicandCongruous< sizeof_bits< ElementA >::value, int(128/sizeof(ElementA))>
 
using SmemLayoutB = layout::ColumnMajorTensorOpMultiplicandCrosswise< sizeof_bits< ElementB >::value, Shape::kK >
 
using IteratorThreadMapA = transform::PitchLinearWarpRakedThreadMap< layout::PitchLinearShape< Shape::kM, Shape::kK >, kThreads, layout::PitchLinearShape< 8, 4 >, kAccessSizeInBits/sizeof_bits< ElementA >::value >
 ThreadMap of iterator A. More...
 
using SmemIteratorA = transform::threadblock::RegularTileIterator< MatrixShape< Shape::kM, Shape::kK >, ElementA, SmemLayoutA, 1, IteratorThreadMapA >
 Shared memory iterator to A operand. More...
 
using IteratorThreadMapB = transform::PitchLinearWarpRakedThreadMap< layout::PitchLinearShape< Shape::kK, Shape::kN >, kThreads, layout::PitchLinearShape< kWarpThreadArrangementContiguousB, kWarpThreadArrangementStridedB >, kAccessSizeInBits/sizeof_bits< ElementB >::value >
 ThreadMap of iterator B. More...
 
using SmemIteratorB = transform::threadblock::RegularTileIterator< MatrixShape< Shape::kK, Shape::kN >, ElementB, SmemLayoutB, 1, IteratorThreadMapB >
 Shared memory iterator to B operand. More...
 
using MmaTensorOp = typename cutlass::gemm::warp::DefaultMmaTensorOp< WarpShape, InstructionShape, ElementA, SmemLayoutA, ElementB, SmemLayoutB, ElementC, LayoutC, Operator, WarpCount::kK >::Type
 
using MmaPolicy = MmaPolicy< MmaTensorOp, MatrixShape< 0, 0 >, MatrixShape< 0, 0 >, WarpCount::kK >
 Policy used to define MmaPipelined. More...
 

Static Public Attributes

static int const kWarpSize = warp::WarpSize<arch::OpClassTensorOp>::value
 Number of threads per warp. More...
 
static int const kThreads = WarpCount::kCount * kWarpSize
 Number of threads total. More...
 
static int const kAccessSizeInBits = 128
 Size of a threadblock-scoped access. More...
 
static int const kWarpThreadArrangementContiguousB
 
static int const kWarpThreadArrangementStridedB
 

Detailed Description

template<typename Shape_, typename WarpShape_, typename InstructionShape_, typename ElementA_, typename ElementB_, typename ElementC_, typename LayoutC_, typename Operator_>
struct cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >

Partial specialization:

A: column-major B: column-major Operator: tensor op class

This uses the default warp-level operator given tile sizes

Member Typedef Documentation

template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::ElementA = ElementA_
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::ElementB = ElementB_
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::ElementC = ElementC_
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::InstructionShape = InstructionShape_
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::IteratorThreadMapA = transform::PitchLinearWarpRakedThreadMap< layout::PitchLinearShape<Shape::kM, Shape::kK>, kThreads, layout::PitchLinearShape<8, 4>, kAccessSizeInBits / sizeof_bits<ElementA>::value>
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::IteratorThreadMapB = transform::PitchLinearWarpRakedThreadMap< layout::PitchLinearShape<Shape::kK, Shape::kN>, kThreads, layout::PitchLinearShape<kWarpThreadArrangementContiguousB, kWarpThreadArrangementStridedB>, kAccessSizeInBits / sizeof_bits<ElementB>::value>
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::LayoutA = layout::ColumnMajor
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::LayoutB = layout::ColumnMajor
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::LayoutC = LayoutC_
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::MmaPolicy = MmaPolicy<MmaTensorOp, MatrixShape<0, 0>, MatrixShape<0, 0>, WarpCount::kK>
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::MmaTensorOp = typename cutlass::gemm::warp::DefaultMmaTensorOp< WarpShape, InstructionShape, ElementA, SmemLayoutA, ElementB, SmemLayoutB, ElementC, LayoutC, Operator, WarpCount::kK>::Type
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::Operator = Operator_
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::OperatorClass = arch::OpClassTensorOp
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::Shape = Shape_
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::SmemIteratorA = transform::threadblock::RegularTileIterator< MatrixShape<Shape::kM, Shape::kK>, ElementA, SmemLayoutA, 1, IteratorThreadMapA>
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::SmemIteratorB = transform::threadblock::RegularTileIterator< MatrixShape<Shape::kK, Shape::kN>, ElementB, SmemLayoutB, 1, IteratorThreadMapB>
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::SmemLayoutA = layout::ColumnMajorTensorOpMultiplicandCongruous< sizeof_bits<ElementA>::value, int(128 / sizeof(ElementA))>
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::SmemLayoutB = layout::ColumnMajorTensorOpMultiplicandCrosswise< sizeof_bits<ElementB>::value, Shape::kK>
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::WarpCount = GemmShape<Shape::kM / WarpShape::kM, Shape::kN / WarpShape::kN, Shape::kK / WarpShape::kK>
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
using cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::WarpShape = WarpShape_

Member Data Documentation

template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
int const cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::kAccessSizeInBits = 128
static
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
int const cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::kThreads = WarpCount::kCount * kWarpSize
static
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
int const cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::kWarpSize = warp::WarpSize<arch::OpClassTensorOp>::value
static
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
int const cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::kWarpThreadArrangementContiguousB
static
Initial value:
template<typename Shape_ , typename WarpShape_ , typename InstructionShape_ , typename ElementA_ , typename ElementB_ , typename ElementC_ , typename LayoutC_ , typename Operator_ >
int const cutlass::gemm::threadblock::DefaultMmaCore< Shape_, WarpShape_, InstructionShape_, ElementA_, layout::ColumnMajor, ElementB_, layout::ColumnMajor, ElementC_, LayoutC_, arch::OpClassTensorOp, 2, Operator_ >::kWarpThreadArrangementStridedB
static

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