[XviD-devel] Re: [Fwd: Data partitioning for encoder]

[sigdrak]

Hello SuperMMX,

SuperMMX wrote:
> yes, this is that why my DP implementation is choosen.
> In my implementation, the positions of different parts
> is stored in every MB, and written to a temporary bitstream,
> and when the VP threshold is met, these data are 
> written to the result bitstream in the order of DP.
> So this will not modify the interface a lot.

I'm not sure on how you do so: given that you don't know beforehand any 
of the partitions' sizes, you can't just dump data unless one partition 
is ready, or you store in a way or another the symbols that are to be 
VLC'ed.

> my implementation is to put the different part of every MB
> one by one, and sigdrak's is to concatentate the 3 bitstreams.

ffmpeg's for the brain part.

> Maybe there will be some advantages and disadvantages of
> both, which need detailed benchmark...

And it may also depend on the CPU, memory other parameters. A DSP, an 
athlon and a pentium4 may behave very differently.

> These days, i am busy doing my master thesis, after that,

Good luck with this. If it's not under NDA or something, and is MPEG-4 
Visual related, I'm sure this list would be interested. :)

> i will provide my patch, (encoding/decoding of RM, DP and RVLC).

Looking forward to it. BTW, does this mean the licence matters (if any) 
with your university are sorted out (as for your thesis) ?

Regards,
sigdrak