[XviD-devel] pre0 milestone

James Hauxwell xvid-devel@xvid.org
Mon, 16 Dec 2002 08:18:29 -0000


Hi,

Could someone please explain what is planned for dynamic enable/disable
of qpel?  If I have read this correctly then it's planned to switch qpel
on and off on a per frame basis, to save encoding a longer motion
vector.

Please would you think long and hard before providing an encoder which
generates non standard streams, as it's likely that any hardware
decoders (dvd type) will not be able to support it, or in fact be likely
to support it.  Hardware manufacturers usually stick to predefined
standards, and any deviation from that is likely to kill the codec for
embedded devices.

Please, I really don't want to be stuck with real/windows media or divx
as an encoder choice.

Jim

-----Original Message-----
From: xvid-devel-admin@xvid.org [mailto:xvid-devel-admin@xvid.org] On
Behalf Of suxen_drol@hotmail.com
Sent: 14 December 2002 09:41
To: xvid-devel@xvid.org
Subject: [XviD-devel] pre0 milestone

i propse the following stuff for our pre/alpha/new-year-ish milestone.


xvidcore (orded in priority)
* fix existing bugs
    - possible qpel/gmc iso problems
    - chroma interpolation bugs?

* intergrate internal ratecontrol
    - gom, ive looked at your code and it seems okay.
	do you need help getting this into core? because its 

* resolve the whole VOL/VOP issues (ie. frame-base qpel, etc..)
    - i reckon, perform the qpel tests/analysis in the first pass, but
      only enable qpel for ippp..ppp sequences in the second pass.
    - koepi and/or syskin have suggested a nonspec xvid_enc_frame field,
      which instructs the encoder to generate non standard streams.
    - any other ideas

* minor clean up api:
    - add preprocess field to xvid_enc_frame; mf's chromaopt is ready
    - add versioning fields to api structures
    - add num_threads to xvid_enc_param.
      make xvid_init return the "num_cpus" available (such that the
      encoing application doesnt have to thouch libpthreads)
    - xvid init should also return a groovy ident string. 
      ie. somthing similar to uname -a.

frontends
* cleanup vfw config box
* fix directshow-related colorspace bugs
* remove the opendivx/divx4 wrapper from xvidcore
* consider moving the fronends inside the xvidcore project??

any comments?

-- pete; life is like a box of ammo


_______________________________________________
XviD-devel mailing list
XviD-devel@xvid.org
http://list.xvid.org/mailman/listinfo/xvid-devel