[Mplayer-cvslog] CVS: main/libmpcodecs vf_tfields.c,1.2,1.3

[Michael Niedermayer]

Hi

On Wednesday 30 April 2003 21:33, D Richard Felker III wrote:
> On Wed, Apr 30, 2003 at 08:22:10PM +0200, Michael Niedermayer wrote:
> > Hi
> >
> > On Wednesday 30 April 2003 19:55, D Richard Felker III wrote:
> > > On Tue, Apr 29, 2003 at 11:56:33AM +0200, Michael Niedermayer wrote:
> > > > Hi
> > > >
> > > > On Monday 28 April 2003 04:54, Richard Felker CVS wrote:
> > > > > Update of /cvsroot/mplayer/main/libmpcodecs
> > > > > In directory mail:/var/tmp.root/cvs-serv5280/libmpcodecs
> > > > >
> > > > > Modified Files:
> > > > > 	vf_tfields.c
> > > > > Log Message:
> > > > > new mode for tfields filter -- shifts fields by a quarter-pixel so
> > > > > the output picture doesn't bob up and down :)
> > > >
> > > > hmm linear interpolation sucks IMHO, why dont u use some filter like
> > > > (-9a + 111b + 29c -3d + 64)>>7 or (-a + 14b + 3c + 8)>>4
> > >
> > > Because I'm lazy and don't know how to make this perform decent in
> > > mmx...
> >
> > i cant imagine any tricks beyond a simple unpack, multiply, add, ...,
> > shift, pack mmx implementation either ATM, but i guess its still more
> > then fast enough
>
> Hmm, ok, I've never done mmx multiply. Do you unpack first, or does
> the multiply automatically give 16bit outputs for 8bit inputs?
u unpack first (then use pmullw), theres no 8bit multiply, and no 8bit shift 
either in mmx allthough theres a 16bit->32bit multiply-add but thats of no 
use here i guess

btw, see http://www.intel.com/design/pentium4/manuals/245471.htm

[...]

-- 
Michael
screen[y][x] ^= cursor[cy][cx];			(violates patent #4,197,590)
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