ytcheng-lab/Softmax-on-Ara-RVV

This project evaluates the acceleration of softmax using RVV vector intrinsics. The result shows up to 99% execution time can be reduced by replacing glibc’s scalar expf() with a Taylor-approximated SIMD version.

语言：Assembly · GitHub 星标：5 · Fork：1

在 GitHub 上查看原仓库