Category "neon"

Optimizing FMA sequences on different arm64 micro-architectures

In order to optimize a heavily used inner loop (3x3xN tensor convolution in winograd domain), I had some good results by using the maximum amount of neon regist