Structure to compute the matrix product targeting CUDA cores and SIMT math instructions.
#include <mma_tensor_op_sm70.h>
|
| using | Shape = Shape_ |
| | Shape of warp-level matrix operation (concept: GemmShape) More...
|
| |
| using | ElementA = ElementA_ |
| | Data type of multiplicand A. More...
|
| |
| using | LayoutA = LayoutA_ |
| | Layout of multiplicand A. More...
|
| |
| using | ElementB = ElementB_ |
| | Data type of multiplicand B. More...
|
| |
| using | LayoutB = LayoutB_ |
| | Layout of multiplicand B. More...
|
| |
| using | ElementC = ElementC_ |
| | Data type of accumulator matrix C. More...
|
| |
| using | LayoutC = LayoutC_ |
| | Layout of accumulator matrix C. More...
|
| |
| using | Policy = Policy_ |
| | Shape of the warp in units of thread (concept: MmaLanePolicySimt) More...
|
| |
| using | OperatorClass = arch::OpClassTensorOp |
| | Indicates class of matrix operator. More...
|
| |
| using | InterleavedTileShape = GemmShape< 32, 32, 4 > |
| | interleaved 32x32 tiles More...
|
| |
| using | IteratorA = MmaVoltaTensorOpMultiplicandTileIterator< MatrixShape< Shape::kM, Shape::kK >, Operand::kA, ElementA, LayoutA, MatrixShape< Policy::Operator::Shape::kM, Policy::Operator::Shape::kK >, Policy::OpDelta::kRow, kThreadCount > |
| | Iterates over the A operand in memory. More...
|
| |
| using | FragmentA = typename IteratorA::Fragment |
| | Storage for A tile. More...
|
| |
| using | IteratorB = MmaVoltaTensorOpMultiplicandTileIterator< MatrixShape< Shape::kK, Shape::kN >, Operand::kB, ElementB, LayoutB, MatrixShape< Policy::Operator::Shape::kK, Policy::Operator::Shape::kN >, Policy::OpDelta::kRow, kThreadCount > |
| | Iterates over the B operand in memory. More...
|
| |
| using | FragmentB = typename IteratorB::Fragment |
| | Storage for B tile. More...
|
| |
| using | IteratorC = MmaVoltaTensorOpAccumulatorTileIterator< MatrixShape< Shape::kM, Shape::kN >, ElementC, LayoutC, typename Policy::Operator::Shape, typename Policy::OpDelta > |
| | Iterates over the C operand in memory. More...
|
| |
| using | FragmentC = typename IteratorC::Fragment |
| | Storage for C tile. More...
|
| |
|
| static int const | kThreadCount = 32 |
| | Number of threads participating in warp-level matrix product. More...
|
| |
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
| using cutlass::gemm::warp::MmaVoltaTensorOp< Shape_, ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, LayoutC_, Policy_, Enable >::InterleavedTileShape = GemmShape<32, 32, 4> |
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
| using cutlass::gemm::warp::MmaVoltaTensorOp< Shape_, ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, LayoutC_, Policy_, Enable >::IteratorA = MmaVoltaTensorOpMultiplicandTileIterator< MatrixShape<Shape::kM, Shape::kK>, Operand::kA, ElementA, LayoutA, MatrixShape< Policy::Operator::Shape::kM, Policy::Operator::Shape::kK >, Policy::OpDelta::kRow, kThreadCount > |
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
| using cutlass::gemm::warp::MmaVoltaTensorOp< Shape_, ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, LayoutC_, Policy_, Enable >::IteratorB = MmaVoltaTensorOpMultiplicandTileIterator< MatrixShape<Shape::kK, Shape::kN>, Operand::kB, ElementB, LayoutB, MatrixShape< Policy::Operator::Shape::kK, Policy::Operator::Shape::kN >, Policy::OpDelta::kRow, kThreadCount > |
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
| using cutlass::gemm::warp::MmaVoltaTensorOp< Shape_, ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, LayoutC_, Policy_, Enable >::IteratorC = MmaVoltaTensorOpAccumulatorTileIterator< MatrixShape<Shape::kM, Shape::kN>, ElementC, LayoutC, typename Policy::Operator::Shape, typename Policy::OpDelta > |
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
| CUTLASS_DEVICE void cutlass::gemm::warp::MmaVoltaTensorOp< Shape_, ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, LayoutC_, Policy_, Enable >::operator() |
( |
FragmentC & |
D, |
|
|
FragmentA const & |
A, |
|
|
FragmentB const & |
B, |
|
|
FragmentC const & |
C, |
|
|
int const & |
partitionN_idx = 0 |
|
) |
| |
|
inline |
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
template<typename Shape_ , typename ElementA_ , typename LayoutA_ , typename ElementB_ , typename LayoutB_ , typename ElementC_ , typename LayoutC_ , typename Policy_ , typename Enable = bool>
The documentation for this class was generated from the following file: