[XviD-devel] Status of GMC

[Christoph Lampert]

On Sun, 3 Nov 2002, Dirk Knop wrote:

> Hi,
> 
> wow, GMC, very nice :) In my first tests it doesn't help much (at least 
> in terms of filesize), the min framesize isn't reached/used, but raises 
> it by the factor of 6(!) somehow (without gmc: framesize 96-107 bytes, 
> with GMC ~609 bytes...).


Yes, I believe that. Current GME is far from perfect, so much fewer blocks
are _SKIPped_ and this leads to larger files. (there are others
reasons, too...) 

> Another note, in bitstream/mbcoding.c, the first function
> 
> void inline bs_put_spritetrajectory(Bitstream * bs,
>                           const int val)
> 
> throws up 51 compiling errors in windows environment.
> 
> Removing the "inline" or adding a __:
> 
> void __inline bs_put_spritetrajectory(Bitstream * bs, const int val)
> 
> changes this behaviour, it compiles flawless.

Okay, I changed it... I guess this was gcc syntex or simply plain stupid. 


> I don't know if removing the inline changes the wanted behaviour, but 
> that's the way my current build is working.

Oh dear, please _don't_ use GMC in any of your released builds. I just
uploaded to CVS in order for people to have a look at the code. I can't
think of any situation where it leads to smaller files, yet, and for users
there isn't anything to test or so. After all, this baby is less than 48
hours old... 

Give me one more week to change GME and mode decision, then it might show
better results and user testing might become interesting. 

gruel