[x264-devel] Re: [Tom Jacobs <T.R.Jacobs at lboro.ac.uk>] Re: Re: Rif: Re: JM 9.6 vs. x264 rev 228: results

[Loren Merritt]

On Mon, 23 May 2005, Tom Jacobs wrote:

> thanks for all your help. i have managed to make a few graphs :)
>
> i have attached 2 of the graphs (not sure if attachements work in the
> mailing list or not) showing how the multisliced implementation works.
> these 2 graphs are taken with 25frame click with 1 ref frame. the major
> difference between the sliced version seam to be at lower psnr the relative
> bitrate of the multislices are bigger. this is understandable since the rate
> control for multislice isnt that good.
>
> having said this what do people think? is it aceptable (assuming im getting
> good benifits from using my multi-prossor system)?

Ratecontrol should no longer be affected by slicing.
The only sources of bitrate increase should be:
- the slice header itself (~5 bytes each)
- MV and intra prediction can't cross slice boundaries (makes more 
difference at low bitrates since it reduces the usability of P_SKIP 
macroblocks.)

--Loren Merritt

-- 
This is the x264-devel mailing-list
To unsubscribe, go to: http://developers.videolan.org/lists.html