HMLP: High-performance Machine Learning Primitives
Public Member Functions | Public Attributes | List of all members
hmlp::pack2D_pbxib< NB, T, TPACK > Struct Template Reference

#include <packing.hpp>

Inheritance diagram for hmlp::pack2D_pbxib< NB, T, TPACK >:
hmlp::pack_pbxib< NB, T, TPACK >

Public Member Functions

virtual void operator() (size_t k, size_t pc, size_t pb, size_t m, size_t ic, size_t ib, T *X, TPACK *packX)
 

Public Attributes

bool trans = false
 
size_t ldx = 0
 

Detailed Description

template<int NB, typename T, typename TPACK>
struct hmlp::pack2D_pbxib< NB, T, TPACK >

end struct unpack_ibxjb column-major matrix packing routine

Member Function Documentation

template<int NB, typename T , typename TPACK >
virtual void hmlp::pack2D_pbxib< NB, T, TPACK >::operator() ( size_t  k,
size_t  pc,
size_t  pb,
size_t  m,
size_t  ic,
size_t  ib,
T *  X,
TPACK *  packX 
)
inlinevirtual

Loop over ib * pb of X to fill packX. Type cast from T to TPACK if necessary.

( pc, ic ) offset

( ic, pc ) offset

Parameters
kk is the number cols, pc is the col offset, pb is the packed size
mm is the number rows, ic is the row offset, ib is the packed size
Xinput data in type T
packXpacked data in type TPACK

Implements hmlp::pack_pbxib< NB, T, TPACK >.

Member Data Documentation

template<int NB, typename T , typename TPACK >
bool hmlp::pack2D_pbxib< NB, T, TPACK >::trans = false

structure closure, e.g. ldx

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