Closed Matthies closed 2 years ago
Some rewrite on the NNUE code using some of the latest SF intrinsics code which gives a solid 1-3% speed gain for x86 (more for Clang, less for gcc) and even > 10% for ARM/Nano (tested an a Rpi2).
Some rewrite on the NNUE code using some of the latest SF intrinsics code which gives a solid 1-3% speed gain for x86 (more for Clang, less for gcc) and even > 10% for ARM/Nano (tested an a Rpi2).