[FFmpeg-cvslog] random thoughts about refactoring (was: Re: r20938 - trunk/libavcodec/h263.c)
Wed Jan 6 06:48:59 CET 2010
On Wed, Jan 06, 2010 at 02:11:48AM +0000, M?ns Rullg?rd wrote:
> "Ronald S. Bultje" <rsbultje at gmail.com> writes:
> > Hi,
> > On Jan 5, 2010, at 6:18 PM, M?ns Rullg?rd <mans at mansr.com> wrote:
> >> Carl Eugen Hoyos <cehoyos at ag.or.at> writes:
> >>> M?ns Rullg?rd <mans <at> mansr.com> writes:
> >>>>>> Also, our H.264 decoder is not a speed demon.
> >>>>> I've read this claim before, but was never able to reproduce it
> >>>>> (on a single core cpu).
> >>>> CoreAVC is ~50% faster than lavc on ARMv7 (NEON).
> >>> Sorry, I meant (and tested) 32bit x86 cpu.
> >> Well, whatever makes it faster on ARM should be applicable anywhere.
> >> The NEON asm is almost exactly the same.
> > If x86 speed is comparable but arm isn't, and it's not the asm, then
> > likely they disable some steps (postprocessing / filters?) on arm.
> No, the output is bit for bit identical. It wouldn't be a conforming
> decoder otherwise. Cache sizes may have something to do with it
> though, the ARM with it's smaller caches being more sensitive to abuse
> of that kind.
maybe its possible to confirm this theory with cachegrind
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
> ... defining _GNU_SOURCE...
For the love of all that is holy, and some that is not, don't do that.
-- Luca & Mans
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: Digital signature
More information about the ffmpeg-cvslog