Talk:FFmpeg Summer Of Code 2008: Difference between revisions

From MultimediaWiki
Jump to navigation Jump to search
Line 67: Line 67:
I couldnt find it with a quick search on the source. --[[User:Angustia]]
I couldnt find it with a quick search on the source. --[[User:Angustia]]
: No, it does not. And please remember to sign your discussion entries (it's the signature icon on top of the Wiki text box. --[[User:Multimedia Mike|Multimedia Mike]] 13:37, 28 March 2008 (EDT)
: No, it does not. And please remember to sign your discussion entries (it's the signature icon on top of the Wiki text box. --[[User:Multimedia Mike|Multimedia Mike]] 13:37, 28 March 2008 (EDT)
Wouldn't it be more natural to first write a decoder and only then an encoder? --[[User:Angustia|Angustia]] 21:14, 28 March 2008 (EDT)

Revision as of 18:14, 28 March 2008

WMV3 Encoder

And this to This patch?

Realistic time frame

A problem that has come up in past years is the number of projects that go uncompleted. I think the Matroska muxer was the only project that was actually completed by the end of the summer. If the student is willing to continue working on a project past the end of summer then that's great, but we should try to estimate a realistic time frame. In looking over the current project list, I think the following could reasonably be completed in three months time:

  • ALAC encoder
  • Speex decoder
  • MXF muxer

These are all very clearly defined tasks, as opposed to some things involving psychoacoustics and perceptual coding, which can be very open-ended. -Mkhodor 06:00, 26 February 2008 (EST)

I'm not sure about the ALAC task. Do we have enough complete information documented for a student to reasonably complete the task? --Multimedia Mike 11:59, 26 February 2008 (EST)
We have a working, not too obfuscated decoder. Also, I think it is similar enough to FLAC to put the student in the right track. Also, the original author of the alac decoder says [1]:
 Although an encoder is not provided, by using the decoder as a sort of specification it should be fairly 
 trivial to write an encoder. By exploiting other lossless audio encoders, such as FLAC, the task will be 
 much easier. Although one wouldn't be able to copy the compression algorithms verbatim, as adaptive 
 compression is used in ALAC and not in FLAC. There are, however, a bunch of academic papers on the issue.
--Vitor 12:13, 26 February 2008 (EST)

Cook Encoder

It could be done based on the G.722.1 specification as banan suggested once somewhere.

(I know, I haven't submitted the G.722.1 decoder yet, but it's almost cook with different huffman tables and a 320-point iMDCT) (posted by User:Angustia)

I want to see some more solid documentation before I will consider it for a project. I also need a mentor for such a project. --Multimedia Mike 13:03, 27 February 2008 (EST)


Port FFmpeg to Visual Studio

Are you sure this is welcome by most devs? Who would mentor it? If so, this would be my preferred project for the summer (winter here =).

IMHO a port is a bad idea because it will quickly become obsolete. Why not develop scripts and/or patches that convert the FFmpeg source tree into a MSVC-compatible source tree. This would reduce maintenance effort, and would faciliate automatic MSVC-regression testing. Task breakdown:

  • convert C99 struct and sparse array initialisation statements into MSVC-compatible (ala C89) syntax.
  • convert GAS inline assembly syntax into MSVC inline assembly syntax (and/or compiler intrinsics).
  • address other MSVC-specific issues: MSVC pedantic about certain typecasts, does not support inline definition of structures, raises error on deliberate attempts to divide-by-zero, and so on.

--Suxen drol 22:00, 15 March 2008 (EDT)

Yes, I've already developed some ideas on how to do this over the past years... The most important is to use the current build system (not .sln files), with some gcc->cl.exe wrapper, like wgcc.


Qualification Tasks wiki

since we have to come up with a list each year of tasks, can we just make a seperate page for them? and update it accordingly, instead of having it spread across 4-5 different pages... Compn 22:08, 18 March 2008 (EDT)

I had a similar idea, but calling it "bite-sized FFmpeg tasks" which would be good for anyone who wants to jump into FFmpeg development. There is no reason to restrict this to the GSoC. --Multimedia Mike 00:07, 19 March 2008 (EDT)
videolan has a similar wiki for small projects -- Compn 17:15, 26 March 2008 (EDT)
Done. Edit as desired but heed my warning about vague tasks. --Multimedia Mike 21:29, 26 March 2008 (EDT)

Cell stuff

The Cell work sounds fascinating, of course. Please provide a few more details so that a student can get a clear idea of what the tasks(s) entail. Also, I assume they will be operating on IBM's Cell simulator if they don't have an actual PS3? Or will the mentor provide SSH access to a PS3/Linux machine? --Multimedia Mike 12:50, 20 March 2008 (EDT)

Prerequisite: Filterbanks (Sigsys)

what the hell is this ? --Compn 11:49, 25 March 2008 (EDT)

If you knew you wouldn't ask :) Signals and systems. Regular signal processing knowledge.--Merbanan 05:17, 26 March 2008 (EDT)

MPEG-4 ALS decoder

Does FFmpeg have an MPEG-4 ALS decoder? I couldnt find it with a quick search on the source. --User:Angustia

No, it does not. And please remember to sign your discussion entries (it's the signature icon on top of the Wiki text box. --Multimedia Mike 13:37, 28 March 2008 (EDT)

Wouldn't it be more natural to first write a decoder and only then an encoder? --Angustia 21:14, 28 March 2008 (EDT)