en/latest/bld__mult__cuda__funcs_8hpp_source.html

 #pragma once

 #include "cuda/memory.hpp"

 namespace bnmf_algs {
 namespace details {
 namespace bld_mult {
 template <typename Real>
 void update_grad_plus_cuda(const cuda::DeviceMemory3D<Real>& S,
                            const cuda::DeviceMemory2D<Real>& beta_eph,
                            cuda::DeviceMemory3D<Real>& grad_plus);

 template <typename Real>
 void update_nom_cuda(const cuda::DeviceMemory2D<Real>& X_reciprocal,
                      const cuda::DeviceMemory2D<Real>& grad_minus,
                      const cuda::DeviceMemory3D<Real>& S,
                      cuda::DeviceMemory2D<Real>& nom);

 template <typename Real>
 void update_denom_cuda(const cuda::DeviceMemory2D<Real>& X_reciprocal,
                        const cuda::DeviceMemory3D<Real>& grad_plus,
                        const cuda::DeviceMemory3D<Real>& S,
                        cuda::DeviceMemory2D<Real>& denom);

 template <typename Real>
 void update_S_cuda(const cuda::DeviceMemory2D<Real>& X,
                    const cuda::DeviceMemory2D<Real>& nom,
                    const cuda::DeviceMemory2D<Real>& denom,
                    const cuda::DeviceMemory2D<Real>& grad_minus,
                    const cuda::DeviceMemory3D<Real>& grad_plus,
                    const cuda::DeviceMemory2D<Real>& S_ijp,
                    cuda::DeviceMemory3D<Real>& S);

 } // namespace bld_mult
 } // namespace details
 } // namespace bnmf_algs
bnmf_algs::cuda::DeviceMemory3D
A wrapper template class around 3D row-major pitched memory stored in device memory (GPU memory)...
Definition: device_memory_3d.hpp:30

bnmf_algs::details::bld_mult::update_nom_cuda
void update_nom_cuda(const cuda::DeviceMemory2D< Real > &X_reciprocal, const cuda::DeviceMemory2D< Real > &grad_minus, const cuda::DeviceMemory3D< Real > &S, cuda::DeviceMemory2D< Real > &nom)
Perform nom_mult update employed in bld_mult algorithm using CUDA.

bnmf_algs::details::bld_mult::X_reciprocal
matrix_t< T > X_reciprocal(const matrix_t< T > &X, double eps)
Compute the reciprocal  of the input matrix .
Definition: bld_mult_funcs.hpp:87

bnmf_algs::bld::bld_mult
tensor_t< T, 3 > bld_mult(const matrix_t< T > &X, const size_t z, const alloc_model::Params< Scalar > &model_params, size_t max_iter=1000, bool use_psi_appr=false, double eps=1e-50)
Compute tensor , the solution of BLD problem , from matrix  using multiplicative update rules...
Definition: bld_mult.hpp:83

bnmf_algs::details::bld_mult::update_denom_cuda
void update_denom_cuda(const cuda::DeviceMemory2D< Real > &X_reciprocal, const cuda::DeviceMemory3D< Real > &grad_plus, const cuda::DeviceMemory3D< Real > &S, cuda::DeviceMemory2D< Real > &denom)
Perform denom update employed in bld_mult algorithm using CUDA.

bnmf_algs::details::bld_mult::update_grad_plus_cuda
void update_grad_plus_cuda(const cuda::DeviceMemory3D< Real > &S, const cuda::DeviceMemory2D< Real > &beta_eph, cuda::DeviceMemory3D< Real > &grad_plus)
Perform grad_plus update employed in bld_mult algorithm using CUDA.

bnmf_algs::cuda::DeviceMemory2D
A wrapper template class around 2D row-major pitched memory stored in device memory (GPU memory)...
Definition: device_memory_2d.hpp:30

bnmf_algs::details::bld_mult::update_S_cuda
void update_S_cuda(const cuda::DeviceMemory2D< Real > &X, const cuda::DeviceMemory2D< Real > &nom, const cuda::DeviceMemory2D< Real > &denom, const cuda::DeviceMemory2D< Real > &grad_minus, const cuda::DeviceMemory3D< Real > &grad_plus, const cuda::DeviceMemory2D< Real > &S_ijp, cuda::DeviceMemory3D< Real > &S)
Perform S update employed in bld_mult algorithm using CUDA.

bnmf_algs
Main namespace for bnmf-algs library.
Definition: alloc_model_funcs.hpp:12

memory.hpp