[MPlayer-dev-eng] [BUG] Mplayer 1.0 pre5 ppc/altivec issues

[Michael Niedermayer]

Hi

On Friday 16 July 2004 04:27, Luca Barbato wrote:
> Michael Niedermayer wrote:
> >>I almost finished a "kludge patch" it should at least let you build
> >>everything, I could bet that that code will break something somewhere,
> >>please audit the patch.
> >
> > including altivec.h at random places like vo_x11.c is IMHO unacceptable,
> > not to mention the #undef vector stuff which is needed then too
>
> The patch works and is just a quick solution, seems that I would have to
> change lots of vo_x11.c to make something cleaner.
>
> > u replace several constants with vec_splat*(constant), is this generating
> > slower code?
>
> I can just avoid them. That was the quicker way to address the issue
>
> > btw, another question as u seem to be fairly interrested in this code,
> > are u interrested in fixing it so beside compiling it actually outputs
> > correct data?
>
> It looks to output the correct data (tested just a bit, please test it
> more),
well i have no ppc, so i cant, but u can, simply compare the output of 
swscale-example in postproc between c and altivec, and also ensure that the 
altivec code is really used in the later case

btw, gcc-3.4.0 misscompiles part of the swscaler here so if swscale-example 
segfaults with 3.4.* try 3.3 or 2.95

btw2, see the cvs commit message of the patch and the disscussion on the 
mailinglist about the issues with the patch

[...]
-- 
Michael
level[i]= get_vlc(); i+=get_vlc();		(violates patent EP0266049)
median(mv[y-1][x], mv[y][x-1], mv[y+1][x+1]);	(violates patent #5,905,535)
buf[i]= qp - buf[i-1];				(violates patent #?)
for more examples, see http://mplayerhq.hu/~michael/patent.html
stop it, see http://petition.eurolinux.org & http://petition.ffii.org/eubsa/en