[XviD-devel] [CVS commit] devapi4 -- Speed optims

[Edouard Gomez]

Christoph Lampert (chl at math.uni-bonn.de) wrote:

> Sound good. 
> Was  this with  B-frames,  or  how did  you  get it  below  1 call  to
> setedge() per frame?

Yep,  that's with bframes,  we used  to edge  reference frames  for each
bframe, now  a flag keep track  of that so  we do it only  once, similar
trick is used for interpolation so  if rounding is the same as requested
for  the current  coded frame,  it's  just used  as is.   There is  also
another optim that swaps backward and reference frame interpolations.

I  forgot  to  precise  the  overall  speedup is  3%  when  bframes  are
activated.   I did  not measure  the  quarterpel speedup  caused by  the
other change pointed in the ChangeLog entry.

-- 
Edouard Gomez