[XviD-devel] recent mmx patch (xvid_bench results)

[jmtest at voila.fr]

Sounds like I'm a bad developper.... :-( I will do more test with an exhaustive loop. Anyway, for the mem_transfer the gain is small, may be it doesn't worth the risk of change.


> Message du 13/04/04 à 12h11
> De : Edouard Gomez <ed.gomez at free.fr>
> A : xvid-devel at xvid.org
> Copie à : 
> Objet : Re: [XviD-devel] recent mmx patch (xvid_bench results)
> 
> pete (p_ross at iprimus.com.au) wrote:
> > it appears mmx 8to16sub is buggy.
> 
> fine you found which one is buggy.
> 
> > PLAINC - 8to16sub  1.178 usec       crc32(1)=0xa1e07163 crc32(2)=0xd86c5d23
> > MMX    - 8to16sub  0.052 usec       crc32(1)=0xa1e07163 crc32(2)=0xd86c5d23
> > 3DNOW  - 8to16sub  0.975 usec       crc32(1)=0xa1e07163 crc32(2)=0xd86c5d23
> 
> Hmmm how the hell does it pass the xvid_bench test ?!
> /me angry. A crc should spot that kind of things.
> 
> > PLAINC - dequant_h263_inter 98.984 usec       crc32=0xe7df8fba
> > <crash>
> 
> Hu? dunno what  can be the cause of that, no  changes happened there.  I
> still don't  understand why my 5  sequence test did not  show so serious
> troubles, moreover the  sequence has been tested 3  times, and each time
> the output was bit to bit matching...
> 
> I'll have a look this evening.
> 
> -- 
> Edouard Gomez
> _______________________________________________
> XviD-devel mailing list
> XviD-devel at xvid.org
> http://list.xvid.org/mailman/listinfo/xvid-devel
> 
------------------------------------------

Faites un voeu et puis Voila ! www.voila.fr