HMLP: High-performance Machine Learning Primitives
|
Public Member Functions | |
GSKS_OPERATOR (double) const | |
Public Attributes | |
const size_t | mr = 8 |
const size_t | nr = 4 |
const size_t | pack_mr = 8 |
const size_t | pack_nr = 4 |
const size_t | align_size = 32 |
const bool | row_major = false |
|
inline |
rank-k update segment
preload u03, u47
prefetch u and w
c = exp( c )
multiple rhs kernel summation