mma_tensor_op_tile_iterator_wmma.h

Go to the documentation of this file.

/***************************************************************************************************
 * Copyright (c) 2017-2019, NVIDIA CORPORATION.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without modification, are permitted
 * provided that the following conditions are met:
 *     * Redistributions of source code must retain the above copyright notice, this list of
 *       conditions and the following disclaimer.
 *     * Redistributions in binary form must reproduce the above copyright notice, this list of
 *       conditions and the following disclaimer in the documentation and/or other materials
 *       provided with the distribution.
 *     * Neither the name of the NVIDIA CORPORATION nor the names of its contributors may be used
 *       to endorse or promote products derived from this software without specific prior written
 *       permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
 * FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
 * STRICT LIABILITY, OR TOR (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 **************************************************************************************************/
#pragma once


#include "cutlass/cutlass.h"
#include "cutlass/arch/wmma.h"

#if defined(CUTLASS_ARCH_WMMA_ENABLED)

#include "cutlass/wmma_array.h"
#include "cutlass/numeric_types.h"
#include "cutlass/tensor_ref.h"
#include "cutlass/matrix_shape.h"

#include "cutlass/arch/memory_sm75.h"
#include "cutlass/gemm/gemm.h"

#include "cutlass/layout/matrix.h"
#include "cutlass/layout/tensor.h"
#include "cutlass/layout/pitch_linear.h"
#include "cutlass/layout/tensor_op_multiplicand_sm75.h"

#include "cutlass/platform/platform.h"
#include "cutlass/fast_math.h"


namespace cutlass {
namespace gemm {
namespace warp {

template <
    typename Shape_,
    Operand Operand,
    typename Element_,
    typename Layout_,
    int OpDelta_,
    int Threads,
    typename Policy_>
class MmaTensorOpWmmaMultiplicandTileIterator;


template <
    typename Shape_,
    typename Element_,
    typename Layout_,
    int OpDelta_,    
    typename Policy_>
class MmaTensorOpWmmaMultiplicandTileIterator<
    Shape_, Operand::kA, Element_, Layout_,
    OpDelta_, 32, Policy_> {
 public:

  using Shape = Shape_;

  static Operand const kOperand = Operand::kA;

  using Element = Element_;

  using Layout = Layout_;

  static int const kOpDelta = OpDelta_;

  using Policy = Policy_;


  //
  // Derived quantities
  //
  using TensorRef = TensorRef<Element, Layout>;

  using Index = typename TensorRef::Index;

  using LongIndex = typename TensorRef::LongIndex;

  using TensorCoord = typename TensorRef::TensorCoord;

  using WmmaShape = MatrixShape<
    Policy::Operator::Shape::kM, 
    Policy::Operator::Shape::kK
  >;

  using WmmaDataType = typename cutlass::arch::CutlassToWmmaDataType<Element>::Type;

  using Iterations = MatrixShape<
    Shape::kRow / WmmaShape::kRow,
    1 
  >;

  using Fragment = WmmaFragmentArray<typename Policy::Operator::FragmentA, Iterations::kCount>;


  static_assert(kOperand == Operand::kA,
    "MmaTensorOpWmmaMultiplicandTileIterator may only be instantiated for A operands to warp-level Mma.");

  static_assert(
    platform::is_same<cutlass::layout::RowMajor, Layout>::value ||
    platform::is_same<cutlass::layout::ColumnMajor, Layout>::value,
    "Supported list of memory layouts for WMMA are: RowMajor, ColumnMajor");

  static_assert(kOpDelta == 1,
    "Alternative arrangements not supported at present.");


private:

  char const *pointer_;
  
  Index byte_offset_;
  
  Index stride_;

  Layout layout_;

public:
  
  CUTLASS_HOST_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator() { }

  CUTLASS_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator(
    TensorRef const &ref, 
    int lane_id
  ): pointer_(reinterpret_cast<char const*>(ref.data())), byte_offset_(0), stride_(ref.stride(0)), layout_(ref.stride(0)) { 
  
  }

  CUTLASS_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator &add_pointer_offset(LongIndex offset) {
    byte_offset_ += (offset * sizeof_bits<Element>::value) / 8;
    return *this;
  }

  CUTLASS_HOST_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator &add_tile_offset(TensorCoord const &tile_offset) {

    Index elements_offset = layout_({tile_offset.row() * Shape::kRow, tile_offset.column() * WmmaShape::kColumn});
    
    byte_offset_ += (elements_offset * sizeof_bits<Element>::value) / 8;

    return *this;
  }

  CUTLASS_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator & operator++() {
    
    Index elements_offset = layout_({0, WmmaShape::kColumn});

    byte_offset_ += (elements_offset * sizeof_bits<Element>::value) / 8;

    return *this;
  }

  CUTLASS_HOST_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator & operator--() {
    
    Index elements_offset = layout_({0, WmmaShape::kColumn});

    byte_offset_ -= (elements_offset * sizeof_bits<Element>::value) / 8;

    return *this;
  }

  CUTLASS_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator & operator+=(TensorCoord const &tile_offset) {
    add_tile_offset(tile_offset);
    return *this;
  }

  CUTLASS_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator & operator-=(TensorCoord const &tile_offset) {
    add_tile_offset(-tile_offset);
    return *this;
  }

  CUTLASS_HOST_DEVICE
  void load_with_byte_offset(Fragment &frag, Index byte_offset) const {

    CUTLASS_PRAGMA_UNROLL
    for (int k = 0; k < Iterations::kColumn; ++k) {
      CUTLASS_PRAGMA_UNROLL
      for (int m = 0; m < Iterations::kRow; ++m) {

        Index load_byte_offset = layout_({m * WmmaShape::kRow, k * WmmaShape::kColumn}) * sizeof_bits<Element>::value / 8;

        const WmmaDataType *ptr = reinterpret_cast<const WmmaDataType *>(pointer_ + byte_offset_ + load_byte_offset + byte_offset); 

        nvcuda::wmma::load_matrix_sync(frag[m], ptr, stride_); 
      
      }
    }
  }
  CUTLASS_HOST_DEVICE
  void load(Fragment &frag) const {
    load_with_byte_offset(frag, 0);
  }
    
  CUTLASS_HOST_DEVICE
  void store_with_byte_offset(Fragment const &frag, Index byte_offset) const {
    
    CUTLASS_PRAGMA_UNROLL
    for (int k = 0; k < Iterations::kColumn; ++k) {
      CUTLASS_PRAGMA_UNROLL
      for (int m = 0; m < Iterations::kRow; ++m) {

        Index store_byte_offset = layout_({m * WmmaShape::kRow, k * WmmaShape::kColumn}) * sizeof_bits<Element>::value / 8;

        WmmaDataType *ptr = reinterpret_cast<WmmaDataType *>(pointer_ + byte_offset_ + store_byte_offset + byte_offset);

        nvcuda::wmma::store_matrix_sync(ptr, frag[m], stride_); 
      
      }
    }
  }

  CUTLASS_HOST_DEVICE
  void store(Fragment const &frag) const {
    store_with_byte_offset(frag, 0);
  }

  CUTLASS_DEVICE
  void set_kgroup_index(int k_group) {
    // no operation here
  }
};



template <
    typename Shape_,
    typename Element_,
    typename Layout_,
    int OpDelta_,    
    typename Policy_>
class MmaTensorOpWmmaMultiplicandTileIterator<
    Shape_, Operand::kB, Element_, Layout_,
    OpDelta_, 32, Policy_> {
 public:

  using Shape = Shape_;

  static Operand const kOperand = Operand::kB;

  using Element = Element_;

  using Layout = Layout_;

  static int const kOpDelta = OpDelta_;

  using Policy = Policy_;


  //
  // Derived quantities
  //

  using TensorRef = TensorRef<Element, Layout>;

  using Index = typename TensorRef::Index;

  using LongIndex = typename TensorRef::LongIndex;

  using TensorCoord = typename TensorRef::TensorCoord;

  using WmmaShape = MatrixShape<
    Policy::Operator::Shape::kK, 
    Policy::Operator::Shape::kN
  >;

  using WmmaDataType = typename cutlass::arch::CutlassToWmmaDataType<Element>::Type;

  using Iterations = MatrixShape<
    1,
    Shape::kColumn / WmmaShape::kColumn
  >;

  using Fragment = WmmaFragmentArray<typename Policy::Operator::FragmentB, Iterations::kCount>;


  static_assert(kOperand == Operand::kB,
    "MmaTensorOpWmmaMultiplicandTileIterator may only be instantiated for B operands to warp-level Mma.");

  static_assert(
    platform::is_same<cutlass::layout::RowMajor, Layout>::value ||
    platform::is_same<cutlass::layout::ColumnMajor, Layout>::value,
    "Supported list of memory layouts for WMMA are: RowMajor, ColumnMajor");

  static_assert(kOpDelta == 1,
    "Alternative arrangements not supported at present.");


private:

  char const *pointer_;
  
  Index byte_offset_;
  
  Index stride_;

  Layout layout_;

public:
  
  CUTLASS_HOST_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator() { }

  CUTLASS_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator(
    TensorRef const &ref, 
    int lane_id
  ): pointer_(reinterpret_cast<char const*>(ref.data())), byte_offset_(0), stride_(ref.stride(0)), layout_(ref.stride(0)) {
  }

  CUTLASS_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator &add_pointer_offset(LongIndex offset) {
    
    byte_offset_ += (offset * sizeof_bits<Element>::value) / 8;

    return *this;
  }

  CUTLASS_HOST_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator &add_tile_offset(TensorCoord const &tile_offset) {
    
    Index elements_offset = layout_({tile_offset.row() * WmmaShape::kRow, tile_offset.column() * Shape::kColumn});
    
    byte_offset_ += (elements_offset * sizeof_bits<Element>::value) / 8;

    return *this;
  }

  CUTLASS_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator & operator++() {
    
    Index elements_offset = layout_({WmmaShape::kRow, 0});

    byte_offset_ += (elements_offset * sizeof_bits<Element>::value) / 8;
    
    return *this;
  }

  CUTLASS_HOST_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator & operator--() {

    Index elements_offset = layout_({WmmaShape::kRow, 0});

    byte_offset_ -= (elements_offset + sizeof_bits<Element>::value) / 8;
    return *this;
  }

  CUTLASS_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator & operator+=(TensorCoord const &tile_offset) {
    add_tile_offset(tile_offset);
    return *this;
  }

  CUTLASS_DEVICE
  MmaTensorOpWmmaMultiplicandTileIterator & operator-=(TensorCoord const &tile_offset) {
    add_tile_offset(-tile_offset);
    return *this;
  }

  CUTLASS_HOST_DEVICE
  void load_with_byte_offset(Fragment &frag, Index byte_offset) const {

    CUTLASS_PRAGMA_UNROLL
    for (int k = 0; k < Iterations::kRow; ++k) {
      CUTLASS_PRAGMA_UNROLL
      for (int n = 0; n < Iterations::kColumn; ++n) {
        
        Index load_byte_offset = layout_({k * WmmaShape::kRow, n * WmmaShape::kColumn}) * sizeof_bits<Element>::value / 8;

        const WmmaDataType *ptr = reinterpret_cast<const WmmaDataType *>(pointer_ + byte_offset_ + load_byte_offset + byte_offset);

        nvcuda::wmma::load_matrix_sync(frag[n], ptr, stride_);        
      }
    }
  }
  CUTLASS_HOST_DEVICE
  void load(Fragment &frag) const {
    load_with_byte_offset(frag, 0);
  }
    
  CUTLASS_HOST_DEVICE
  void store_with_byte_offset(Fragment const &frag, Index byte_offset) const {
    
    CUTLASS_PRAGMA_UNROLL
    for (int k = 0; k < Iterations::kRow; ++k) {
      CUTLASS_PRAGMA_UNROLL
      for (int n = 0; n < Iterations::kColumn; ++n) {

        Index store_byte_offset = layout_({k * WmmaShape::kRow, n * WmmaShape::kColumn}) * sizeof_bits<Element>::value / 8;

        WmmaDataType *ptr = reinterpret_cast<WmmaDataType *>(pointer_ + byte_offset_ + store_byte_offset + byte_offset);
        
        nvcuda::wmma::store_matrix_sync(ptr, frag[n], stride_);        
      }
    }
  }

  CUTLASS_HOST_DEVICE
  void store(Fragment const &frag) const {
    store_with_byte_offset(frag, 0);
  }

  CUTLASS_DEVICE
  void set_kgroup_index(int k_group) {
    // no operation here
  }
};

template <
    typename Shape_,
    typename Element_,
    typename Layout_,
    typename OpDelta_,
    typename Policy_>
class MmaTensorOpWmmaAccumulatorTileIterator;


template <
    typename Shape_,
    typename Element_,
    typename Layout_,
    typename OpDelta_,    
    typename Policy_>
class MmaTensorOpWmmaAccumulatorTileIterator
{
 public:

  using Shape = Shape_;

  using Element = Element_;

  using Layout = Layout_;

  using OpDelta = OpDelta_;

  static int const kThreads = 32;

  using Policy = Policy_;


  //
  // Derived quantities
  //
  using TensorRef = TensorRef<Element, Layout>;

  using Index = typename TensorRef::Index;

  using LongIndex = typename TensorRef::LongIndex;

  using TensorCoord = typename TensorRef::TensorCoord;

  using WmmaShape = MatrixShape<
    Policy::Operator::Shape::kM, 
    Policy::Operator::Shape::kN
  >;
  
  using WmmaDataType = typename cutlass::arch::CutlassToWmmaDataType<Element>::Type;

  static nvcuda::wmma::layout_t const WmmaLayout = cutlass::arch::CutlassToWmmaLayout<Layout>::value;

  using Iterations = MatrixShape<
    Shape::kRow / WmmaShape::kRow,
    Shape::kColumn / WmmaShape::kColumn
  >;

  using Fragment = WmmaFragmentArray<typename Policy::Operator::FragmentC, Iterations::kCount>;

  static_assert(
    platform::is_same<cutlass::layout::RowMajor, Layout>::value ||
    platform::is_same<cutlass::layout::ColumnMajor, Layout>::value,
    "Supported list of memory layouts for WMMA are: RowMajor, ColumnMajor");

private:
  
  cutlass::TensorRef<Element, Layout> ref_;

public:
  
  CUTLASS_HOST_DEVICE
  MmaTensorOpWmmaAccumulatorTileIterator() { }

  CUTLASS_DEVICE
  MmaTensorOpWmmaAccumulatorTileIterator(
    TensorRef const &ref, 
    int lane_id
  ): ref_(ref) { }

  CUTLASS_DEVICE
  MmaTensorOpWmmaAccumulatorTileIterator &add_pointer_offset(LongIndex offset) {
    ref_.add_pointer_offset(offset);
    return *this;
  }

  CUTLASS_HOST_DEVICE
  MmaTensorOpWmmaAccumulatorTileIterator &add_tile_offset(TensorCoord const &tile_offset) {
    ref_.add_coord_offset({tile_offset.row() * Shape::kRow, tile_offset.column() * Shape::kColumn});
    return *this;
  }

  CUTLASS_DEVICE
  MmaTensorOpWmmaAccumulatorTileIterator & operator++() {
    ref_.add_coord_offset({Shape::kRow, 0});
    return *this;
  }

  CUTLASS_HOST_DEVICE
  MmaTensorOpWmmaAccumulatorTileIterator & operator--() {
    ref_.add_coord_offset({-Shape::kRow, 0});
    return *this;
  }

  CUTLASS_DEVICE
  MmaTensorOpWmmaAccumulatorTileIterator & operator+=(TensorCoord const &tile_offset) {
    add_tile_offset(tile_offset);
    return *this;
  }

  CUTLASS_DEVICE
  MmaTensorOpWmmaAccumulatorTileIterator & operator-=(TensorCoord const &tile_offset) {
    add_tile_offset(-tile_offset);
    return *this;
  }

  CUTLASS_HOST_DEVICE
  void load_with_pointer_offset(Fragment &frag, Index pointer_offset) const {
    
    CUTLASS_PRAGMA_UNROLL
    for (int m = 0; m < Iterations::kRow; ++m) {
      CUTLASS_PRAGMA_UNROLL
      for (int n = 0; n < Iterations::kColumn; ++n) {

        const WmmaDataType * ptr = reinterpret_cast<const WmmaDataType*> (ref_.data() + ref_.offset({m * WmmaShape::kRow, n * WmmaShape::kColumn}) + pointer_offset);
        
        nvcuda::wmma::load_matrix_sync(frag[m * Iterations::kColumn + n], ptr, ref_.stride()[0], WmmaLayout); 

      }
    }
  }
  CUTLASS_HOST_DEVICE
  void load(Fragment &frag) const {
    load_with_pointer_offset(frag, 0);
  }
    
  CUTLASS_HOST_DEVICE
  void store_with_pointer_offset(Fragment const &frag, Index pointer_offset) const {
    
    CUTLASS_PRAGMA_UNROLL
    for (int m = 0; m < Iterations::kRow; ++m) {
      CUTLASS_PRAGMA_UNROLL
      for (int n = 0; n < Iterations::kColumn; ++n) {

        WmmaDataType * ptr = reinterpret_cast<WmmaDataType*> (ref_.data() + ref_.offset({m * WmmaShape::kRow, n * WmmaShape::kColumn}) + pointer_offset);

        nvcuda::wmma::store_matrix_sync(ptr, frag[m * Iterations::kColumn + n], ref_.stride()[0], WmmaLayout); 
      }
    }
  }

  CUTLASS_HOST_DEVICE
  void store(Fragment const &frag) const {
    store_with_pointer_offset(frag, 0);
  }

  CUTLASS_DEVICE
  void set_kgroup_index(int k_group) {
    // no operation here
  }
};



} // namespace warp
} // namespace gemm
} // namespace cutlass


#endif // if defined(CUTLASS_ARCH_WMMA_ENABLED)