[MPlayer-dev-eng] [PATCH] correct illegal instruction on < armv5tE

[Siarhei Siamashka]

On Tuesday 14 October 2008, GNUtoo at no-log.org wrote:
> > On the other hand, implementing full hardware video acceleration (the
> > Glamo can do MPEG4), would be awesome. vaapi has still a long way to
> > go, it seems.
>
> we have hardware acceleration for videos with a patched mplayer bug:
> *it's ugly(i don't mean the code...i mean the colours)

Regarding the colors, this must be some bug probably, too bad if it is in some
closed source component. Regarding "patched mplayer bug", is it the one with
PLD instruction or there is something else to fix?

> *it uses framebuffer and not X
> see http://unadventure.wordpress.com/2008/06/08/accelerating-in-my-pocket/
>
> and yes it's slow...

From what I can see, it shows sthe numbers that are not impressive at all,
maybe even just decoding MPEG4 video on the ARM core could be faster. One of
the comments there mentions that you have Xv supported there with Xglamo, so
this is a good thing for video playback (unless implementation sucks and it
has a heavy overhead itself).

> and between lastest trunk and the 1.0rc2-r7 the 
> lastest trunk seems a lot slower on youtube(i'll test with mpeg2)

Do you mean the latest trunk of that Glamo thing? Or trunk in MPlayer SVN?
If there are some noticeable performance regressions in MPlayer, you can try
to search for the revision where it got introduced and report it. As 
libavcodec is an svn:externals, this makes everything a bit more complicated,
but you can still get some older SVN snapshots using date:
'svn up -r {SOMEDATE}'. It might be easier to do performance regressions
investigation with FFmpeg itself.

-- 
Best regards,
Siarhei Siamashka