[XviD-devel] want to give a hand

Alban Bedel xvid-devel@xvid.org
Wed, 24 Jul 2002 17:03:20 +0200


Hi Edouard Gomez,

on Wed, 24 Jul 2002 16:09:03 +0200 you wrote:

> Marc FD (marcfd@free.fr) wrote:
> 
> > Hi.
> >
> > I'm playing around with the mmx/xmm/isse/3dnow i.sets since 3 days and
> > i think  it's time to make  something "usefull" s.o.  said there where
> > always things to be "MMXed" so if  there is something like it to do, i
> > would help if i can :) (it would be kind if you can lemme do something
> > "easy" to begin, but i think all easy opt. are already done...)
> >
> > i ask because  i'm not used enough  with XviD to say :  "let's work on
> > xx_xx.c and then i will patch xx.c, ect..."
> >
> > i need to "aim" on something
> >
> > tell me if i can help (i hope i could :)
> 
> I was  thinking about an  easy function to  code. It's not asm  but it
> could be  interesting for  you because you'll  have to learn  how XviD
> internals work.
> 
> Short explaination :
> 
> XviD uses lot of flags to  make it very flexible. The problem is flags
> are not  even checked and  there're many combinations that  are simply
> wrong (eg:  8x8 search when inter4v  is not set...  many more nonsense
> combinations are possible).
> 
> What about coding a function :
> 
> unsigned long xvid_valid_flag(unsigned long flag, int flagtype);
> 
> that would return a valid flag (doing its best to match library client wishes)
> 
> (the flagtype is just to indicate what sort of flag has to be checked:
> general, global, motion ...)
> 
> This would be great to have such a flag validator in the core.

Imho the flags must simply use well choosen values. So when you look at
the flag values you know wich are exclusive and wich aren't. Having a 'validator'
func isn't a bad idea but i think we must first take the problem at the root.
	Albeu