[Ffmpeg-cvslog] r6213 - in trunk: Changelog MAINTAINERS doc/ffmpeg-doc.texi libavcodec/Makefile libavcodec/allcodecs.c libavcodec/avcodec.h libavcodec/vp5.c libavcodec/vp56.c libavcodec/vp56.h libavcodec/vp56data.c libavcodec/vp56data.h libavcodec/vp5data.h libavcodec/vp6.c libavcodec/vp6data.h libavformat/flvdec.c libavformat/nsvdec.c libavformat/riff.c libavformat/swf.c

Diego Biurrun diego
Mon Sep 11 11:12:32 CEST 2006


On Mon, Sep 11, 2006 at 02:16:53AM +0200, Aurelien Jacobs wrote:
> On Mon, 11 Sep 2006 01:00:12 +0200
> Aurelien Jacobs <aurel at gnuage.org> wrote:
> 
> > On Mon, 11 Sep 2006 00:52:53 +0200
> > Diego Biurrun <diego at biurrun.de> wrote:
> > 
> > > On Mon, Sep 11, 2006 at 12:45:35AM +0200, Aurelien Jacobs wrote:
> > > > On Sun, 10 Sep 2006 18:42:13 +0200
> > > > 
> > > > As there is no altivec implementation of vp3 idct, it must be
> > > > put_pixels_tab[0][0].
> > > > Could you confirm it, testing with the following patch ?
> > > 
> > > Bingo, problem solved.
> > 
> > Ok, I will try to find a clean fix.
> 
> For now I can't find a fix to align memory, so here is a lame fix.
> It replaces the dsp call by a pure C implementation. The code is
> obviousily a bit slower (on pentium-m):
> 
> dsp version:
> 1706 dezicycles in vp_dsp, 2096933 runs, 219 skips
> C version:
> 2001 dezicycles in vp_dsp, 2096934 runs, 218 skips
> 
> But the overall speed drop is less than 0.5%, so it's not so bad.
> 
> I will still think about it and if I can't find anything better
> I will apply this tomorrow.
> 
> Can you confirm that it works ?

Yes, the artifacts are gone.

Diego




More information about the ffmpeg-cvslog mailing list