Altivec IDCT

[Christophe Massiot]

À (At) 10:58 -0700 6/09/01, Michel LESPINASSE écrivait (wrote) :

>Have you looked at the precision of this routine ? I suppose its
>precise enough as its based on floats - but I'd love to see the ieee
>test results for it though.

As usual I didn't look at anything :-), I didn't even try it since I 
do not have a G4. I just read 
<http://lists.linuxppc.org/linuxppc-dev/200108/msg00267.html>, where 
Michel Lanners says he will have a look at it. Maybe you should 
contact him directly...

>I've been looking at motorola's integer implementation (16-bit), but
>its not precise enough (far from it) and I dont think there is any
>hope to bring it to compliance. Its blindingly fast though.
>
>I've been thinking about an integer implementation mixing 32-bit for
>rows and 16-bit for columns (similar to the mmx implementation) but
>I'm not done with it yet. I suppose I can bring it to a speed
>in-between motorola's routine and paulus's routine though.

That would be great...


-- 
Christophe Massiot.