[Ffmpeg-devel] Threading in libavcodec

Mathieu Monnier manao
Mon Jan 23 07:13:07 CET 2006

>>Can you refine what you want to compare ? Quality, or speed ?
> PSNR at a given bitrate or bitrate at a given PSNR.


In sheer efficiency - without rate control, bframes, and without 
considering speed - I seem to recall that lavc gives slightly better 
results. That comes mainly from the fact it can be slowed down more than 
XviD ( rd choice of the quantizer for example ). There's no magic 
involved : when both codec uses RD cost to do all their decision, 
they'll often be quite close efficiency wise - provided than none is 
buggy - and often the one doing the most tests will be the better.

But of course, as soon as rate control is considered, it's another 
story. And psnr is definitely not the tool to use in order to judge the 
efficiency of a rate control algorithm.

In any case, how XviD and lavc compares are in no way related to the 
original issue : XviD implements multithreading without losing any 
efficiency, and scales well, while lavc, from what you're saying, doesn't.



More information about the ffmpeg-devel mailing list