[XviD-devel] Changes to get_pmv2

[Radek Czyz]

Tom Jacobs wrote:

>>The prediction is also used to calculate how many bits a motion vector
>>will take in a bitstream. Therefore, even at full search, it's pretty
>>much neccessary to have it.
> 
> 
> but if i give it a wrong prediction andit uses this wrong prediction for its
> search then the size of the MV will be correct

Yes, such vector will still work: the differential can be re-calculated 
later, before it's written to the bitstream.
However, any ME must take into account the motion overhead: how many 
bits it takes to code motion. Any mistake here, even any approximation, 
will simply ruin the quality of the video.

>>Why won't you parallize b-frames first? Any row of any b-frame can be
>>done in any order, so you can easly divide arbitrary number of b-frames
>>into arbitrary number of threads.
> 
> 
> does that mean in a B-frame predictions are not made based on top and top
> right?

Yes. It's much simpler in b-frames, very similar to mpeg 1/2. Every row 
is independant, prediction is only from left (and reset at boundary).

Radek