On Sun, Oct 25, 2015 at 12:08 AM, James Almer <[email protected]> wrote:
> +    __asm__ ("bsfq %1,%0" : "=r" (c) : "rm" (v));

Shouldn't we be using tzcnt instead of bsf? Or rep bsf (which is the
same opcode as tzcnt) if we need to support ancient compilers. tzcnt
is generally faster than bsf on modern CPUs and backwards-compatible
with older CPUs.
_______________________________________________
ffmpeg-devel mailing list
[email protected]
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Reply via email to