x264 development especially, interface, features, purposes, testing

andy

This is a partial continuation of the former x264 development thread. This thread is meant purely for development purposes, which does not include making any feature requests, especially not for features to be added to the VfW interface!. Any VfW relates issues have to go here instead. Any bugreports also do not belong to this thread but either in the VfW thread, or the x264 cli bugreport thread.

I am new to x264 code
The decoder testing part in x264.c was removed some months ago and it has not returned since then. The exact release was a snapshot of 24-08-2005.  
I want to work on the decoder for a project of my own. In that old snapshot the only thing in the coments was something like quot;The decoder test part is not working and we would appreciate some help in this directionquot;. I would like to know what eactly were the problems in this regard and how much work i would have to put in to get it working again ?

The decoder supposedly supports baseline profile only. I have no idea how close to working it was in 2004, since I have never compiled it. In addition to just making it work, much will need to be updated to match changes in the functions shared with the encoder, which have been modified since then.

from the compilation that i have i cannot get a clue as to which profiles it is supporting because i didnt see any comments about it(may b i ddint look hard enough). but i think it should be a bit more advanced.

No B-frames, no cabac, no wpred, no interlacing, and it was written before high profile existed. That makes it baseline.

and nobody ever mad eany changes to it ?

Fenrir converted from cvs to svn in 2004-06-03, at which point earlier revision history was lost. The decoder has not been modified since then.

is there any other open source decoder which supports the main profile or at least compatible with common output formats?

showthread.php?t=108570

thanks

I have implemented the x264 decoder supported baseline profile several weeks ago. Unexpected, the current x264 program structure is not suitable for decoder. Especially of the interpolation module and the macroblock context cache load/save module. The experimental results show that it's half lower than ffmpeg libavcodec.
However, I won't stop improving my decoder. For I think the structure of h.264 decoder of ffmpeg libavcodec is so bad. And it's decoding speed is not fast enough.


Originally Posted by lspbeyondI think the structure of h.264 decoder of ffmpeg libavcodec is so bad. And it's decoding speed is not fast enough.

That's right. Kick some CoreAVC's asses out there too  

And now my question to the developers. What is to be done to the X264 encoder in order to improve visual quality and compressibility (i.e. what features are yet to be implemented) ?
Is there a roadmap for improving or anything else ?
It is always interesting to know in advance about what we may expect in the future, isn't it ?

Thanks

There is no plan, do not ask about future features, I will not discuss anything I'm writing until it's ready to test.

Well, Leo 69, try to read TODO list from lorenm/src/x264/todo.txt

The RD subpel motion estimation (--subme 7) added in rev.476 is the old patch finally committed to the svn, or is it a completely new approach?

I don't know for sure, but I think that RD subpel motion estimation(--subme 7) is a old patch, with necessary changes.

it's the old patch updated to work with the new svn. I think he may have made some updates to it in other ways as well.

Is the in-loop filter adaptive??? I made a high bitrate encode (720x576 2500 kbps) once with in-loop and once without! I can't see a difference between them!!!!!!

yes it is.

Thanks for making it clear to me!!!

In the Sticky: quot;MPEG-4 AVC/H.264 Informationquot; bond doesn't say anything about adaptive loop in x264!!!