[Ffmpeg-devel] [PATCH] Little optimization to fft_sse.c
Sun Mar 5 21:05:43 CET 2006
On 3/5/06, Zuxy Meng <zuxy.meng at gmail.com> wrote:
> This patch uses xorps instead of mulps to toggle the sign of a float,
> as suggested by Software Optimization Guide for AMD64 Processors.
Did you benchmark the speed difference with the hardware counters?
Reinventing the wheel certainly is annoying, but as long as all other
wheels are square...
More information about the ffmpeg-devel