File size: 1,400 Bytes
9375c9a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 |
// Copyright (C) 2015 Davis E. King ([email protected])
// License: Boost Software License See LICENSE.txt for the full license.
#ifndef DLIB_DNN_CuBLAS_H_
#define DLIB_DNN_CuBLAS_H_
#ifdef DLIB_USE_CUDA
#include "tensor.h"
#include "cuda_errors.h"
namespace dlib
{
namespace cuda
{
// -----------------------------------------------------------------------------------
void gemm (
float beta,
tensor& dest,
float alpha,
const tensor& lhs,
bool trans_lhs,
const tensor& rhs,
bool trans_rhs
);
/*!
requires
- The dimensions of lhs and rhs must be compatible for matrix
multiplication. In particular:
- Let L == trans_lhs ? trans(mat(lhs)) : mat(lhs)
- Let R == trans_rhs ? trans(mat(rhs)) : mat(rhs)
- Let D == mat(dest)
- D.nr() == L.nr() && D.nc() == R.nc()
(i.e. dest must be preallocated and have the correct output dimensions)
- L.nc() == R.nr()
ensures
- performs: dest = alpha*L*R + beta*mat(dest)
!*/
// ------------------------------------------------------------------------------------
}
}
#endif // DLIB_USE_CUDA
#endif // DLIB_DNN_CuBLAS_H_
|