CUTLASS: cutlass::gemm::threadblock::Gemv< Core_ > Class Template 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

• gemm

• threadblock

• Gemv

Public Types | Public Member Functions | List of all members

cutlass::gemm::threadblock::Gemv< Core_ > Class Template Reference

Structure to compute the matrix-vector product using SIMT math instructions.

#include <gemv.h>

|

Public Types

| | using | Shape = typename Core_::Shape | | | | using | Operator = typename Core_::Operator | | | The MMA operator that computes GEMV. More...
| | | | using | IteratorA = typename Core_::IteratorA | | | Iterates over A in global memory. More...
| | | | using | IteratorB = typename Core_::IteratorB | | | Iterates over B in global memory. More...
| | | | using | IteratorC = typename Core_::IteratorC | | | Fragment of operand C loaded from global memory. More...
| | | | using | FragmentA = typename IteratorA::Fragment | | | Fragment of operand A loaded from global memory. More...
| | | | using | FragmentB = typename IteratorB::Fragment | | | Fragment of operand B loaded from global memory. More...
| | | | using | FragmentC = typename Operator::FragmentC | | | Fragment of operand accumulator loaded/stored to global memory. More...
| | | | using | ThreadShape = typename Core_::ThreadShape | | | Shape of the per-thread GEMV operation. More...
| | |

|

Public Member Functions

| | CUTLASS_DEVICE | Gemv () | | | | CUTLASS_DEVICE void | operator() (GemmCoord const &problem_size, FragmentC &accum, IteratorA iterator_A, IteratorB iterator_B, FragmentC const &src_accum) | | |

Member Typedef Documentation

template<class Core_ >

| using cutlass::gemm::threadblock::Gemv< Core_ >::FragmentA = typename IteratorA::Fragment |

template<class Core_ >

| using cutlass::gemm::threadblock::Gemv< Core_ >::FragmentB = typename IteratorB::Fragment |

template<class Core_ >

| using cutlass::gemm::threadblock::Gemv< Core_ >::FragmentC = typename Operator::FragmentC |

template<class Core_ >

| using cutlass::gemm::threadblock::Gemv< Core_ >::IteratorA = typename Core_::IteratorA |

template<class Core_ >

| using cutlass::gemm::threadblock::Gemv< Core_ >::IteratorB = typename Core_::IteratorB |

template<class Core_ >

| using cutlass::gemm::threadblock::Gemv< Core_ >::IteratorC = typename Core_::IteratorC |

template<class Core_ >

| using cutlass::gemm::threadblock::Gemv< Core_ >::Operator = typename Core_::Operator |

template<class Core_ >

| using cutlass::gemm::threadblock::Gemv< Core_ >::Shape = typename Core_::Shape |

template<class Core_ >

| using cutlass::gemm::threadblock::Gemv< Core_ >::ThreadShape = typename Core_::ThreadShape |

Constructor & Destructor Documentation

template<class Core_ >

|

| CUTLASS_DEVICE cutlass::gemm::threadblock::Gemv< Core_ >::Gemv | ( | | ) | |

| inline |

Member Function Documentation

template<class Core_ >

|

| CUTLASS_DEVICE void cutlass::gemm::threadblock::Gemv< Core_ >::operator() | ( | GemmCoord const & | problem_size, | | | | FragmentC & | accum, | | | | IteratorA | iterator_A, | | | | IteratorB | iterator_B, | | | | FragmentC const & | src_accum | | | ) | | |

| inline |

< source accumualtor tile

Parameters

| problem_size | problem size of batched GEMV | | accum | destination accumulator tile | | iterator_A | iterator over A operand in global memory | | iterator_B | iterator over B operand in global memory |

---

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

• gemv.h

---

Generated by 1.8.11