[XviD-devel] question about the decoder+slice rendering

[elcabesa]

i don0t know what happened to my pc, today it work with 54% of cputime, 
ffmpeg lavcodec can decode it with 9% of cpu time. i tryied some optimization 
but it seems that gcc already uses these optimization=)

see you
On Sunday 14 December 2003 00:56, Michael Militzer wrote:
> Hi,
>
> yes, I know it's slow. However it's already much faster than the original
> version I wrote in May. So I think you won't get the c-code much faster
> anymore - meantime I also have a faster PC now than in May and together
> with the code optimizations, I'm quite happy personally ;-))
>
> Also I think that the quality of the postprocessing is pretty nice and
> that is because we treat each single edge pixel individually. This however
> makes it very hard (if not impossible) to create a MMX version that runs
> faster than the current c implementation. I have some other ideas how
> to reduce the complexity of the deblocking process, however I need to
> check first if I can retain the same quality level then.
>
> You can expect some more new postprocessing stuff soon that should
> further enhance the quality (and maybe speed as well). BTW: What kind of
> resolution did you try to decode on your Athlon? 263% cpu usage sounds
> very high, because I was able to decode a 640xXXX sequence with post-
> processing in real-time on a PIII 1.1 GHz.
>
> bye,
> Michael
>
> Quoting elcabesa <elcabesa at inwind.it>:
> > wow
> > i tryed your new postprocessing filter and it work very well.
> > do you think will be possible to meke it faster? or should we wait for a
> > asm
> >
> > rewrite?
> > i tested it with a athlon 900 and mplayer told me that video decoding
> > need 263% of cpu time, so i'll never will be able to use it. i looked
> > inside code
> >
> > and found not a way to have it fast, maybe some small enanchement but i
> > think
> >
> > it oculd not make big difference.
> > without deblocking cpu time goes down to 10 15 %
> >
> > On Thursday 11 December 2003 18:20, Michael Militzer wrote:
> > > Quoting Edouard Gomez <ed.gomez at free.fr>:
> > > > elcabesa (elcabesa at inwind.it) wrote:
> > > > > hi i was looking new code and i saw that postprocessing is not
> >
> > compiled
> >
> > > > when i
> > > >
> > > > > run make
> > > > > i tryed a make clean a
> > > > > autogen.sh
> > > > > configure.sh
> > > > > but postprocessing files arent compiled
> > > > > i think that makefile shoudl be updated
> > > >
> > > > sources.inc  is already up  to date  since yesterday,  "cvs up"  and
> > > > try again. (a cvs  diff on my working  dir doesn't show any diff  to
> > > > the cvs and i'm sure it compiles this file: C:
> > > > image/postprocessing.c)
> > >
> > > and while you're at updating your sources, just do it once again ;-) I
> > > forgot to add some code to xvid.c in order to initialize the postproc
> > > tables - fixed now...
> > >
> > > bye,
> > > Michael
> > > _______________________________________________
> > > XviD-devel mailing list
> > > XviD-devel at xvid.org
> > > http://list.xvid.org/mailman/listinfo/xvid-devel
> >
> > _______________________________________________
> > XviD-devel mailing list
> > XviD-devel at xvid.org
> > http://list.xvid.org/mailman/listinfo/xvid-devel
>
> _______________________________________________
> XviD-devel mailing list
> XviD-devel at xvid.org
> http://list.xvid.org/mailman/listinfo/xvid-devel