[Ffmpeg-devel] [PATCH] SSE counterpart of ff_imdct_calc_3dn2
Fri Aug 25 04:30:35 CEST 2006
On Fri, 2006-08-25 at 09:02 +0800, Zuxy Meng wrote:
> 2006/8/25, Uoti Urpala <uoti.urpala at pp1.inet.fi>:
> > Recent vorbis commit messages mention a 0.5% overall vorbis speedup from
> > converting intrinsics to inline asm, and a 0.5% overall vorbis slowdown
> > from supporting older gcc versions. So that's an order of magnitude more
> > slowdown than you said they caused "at worst", and a counterexample to
> > your claim that using intrinsics instead of asm would cause over 1%
> > slowdown.
> 0.5% difference is too small. Switch to another CPU or another version
> of gcc and the result may change.
Which of the 0.5% differences is "too small" for what? Such a difference
is small enough that other "random" changes might have bigger effects,
but he claimed that "at worst" the slowdown from supporting older gccs
would be 0.05% and that converting any of the inline asm to intrinsics
would cause over 1% slowdown. If the claim is about 0.05% then a 0.5%
difference shouldn't be too small, and certainly a 0.5% slowdown is not
"too small" to say it's below 1%.
More information about the ffmpeg-devel