[vlc-devel] [PATCH] Disable faad, lower priority of decoder below avcodec

[Rafaël Carré]

On Sun, 27 Jun 2010 04:42:10 +0300
"Rémi Denis-Courmont" <rem at videolan.org> wrote:

> On Friday 25 June 2010 22:42:46 Rafaël Carré, you wrote:
> > On Fri, 25 Jun 2010 21:29:01 +0200
> > 
> > Rémi Denis-Courmont <remi at remlab.net> wrote:
> > > To me, it sounds like we should remove the plugin completely
> > > instead
> > 
> > I was removing faad from extras/contrib/src/Distributions/* and I
> > noticed maemo.mak says:
> > "#.faad -> way slower than libavcodec"
> > 
> > Did you try a fixed point build for faad?
> 
> No.
> 
> > afaics ffmpeg decoder is float only so perhaps it makes sense to
> > keep faad for targets without FPU
> 
> N900 has a hardware FPU, as far as I know, though it is not as fast
> as the ALU(s).

I tested on ARM926ej-s (no FPU), playing one random aac file:

time LD_LIBRARY_PATH=src/.libs ./vlc -I dummy --verbose=-1 --no-random
--no-media-library --plugin-path=modules /path/to/my.aac vlc://quit
--codec XXX (avcodec/faad)
--sout "#transcode{acodec=s16l}:std{access=file,mux=raw,dst=/dev/null}"

faad -DFIXED_POINT  33.27s user 3.36s system 99% cpu 36.693 total
avcodec r23888      76.79s user 7.89s system 99% cpu 1:24.81 total
faad               105.78s user 7.74s system 99% cpu 1:53.69 total

If you build faad from contribs, Patches/faad-arm-fixed.patch will
define FIXED_POINT if gcc defines __arm__

New patch proposed:
	- Don't build faad in contribs when not targetting arm
	- Lower faad priority only if HAVE_FPU is 1

-- 
✍ Rafaël Carré ☺
-------------- next part --------------
A non-text attachment was scrubbed...
Name: faad.diff
Type: text/x-patch
Size: 6623 bytes
Desc: not available
URL: <http://mailman.videolan.org/pipermail/vlc-devel/attachments/20100629/5e6f620d/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
URL: <http://mailman.videolan.org/pipermail/vlc-devel/attachments/20100629/5e6f620d/attachment.sig>