What's Cooking For Mesa & X.Org This Summer?

[phoronix]

Phoronix: What's Cooking For Mesa & X.Org This Summer?

Summer is quickly approaching in the northern hemisphere so that means it's time for yet another year of Google's Summer of Code. Once again, X.Org / Mesa should be participating, so it's now time to submit ideas for areas where potential student developers could focus their summer work. Here's a few of the possibilities...

http://www.phoronix.com/vr.php?view=OTIwMQ

[deanjo]

I used to look forward to GSoC but nowdays I can't be bothered with it. Too many GSoC projects get started, reach about 50-60% completion and then die a slow agonizing death never to be looked at again. It's a great idea but it has a very poor execute to completion record.

[pvtcupcakes]

I used to look forward to GSoC but nowdays I can't be bothered with it. Too many GSoC projects get started, reach about 50-60% completion and then die a slow agonizing death never to be looked at again. It's a great idea but it has a very poor execute to completion record.

A lot of the GSOC projects are really ambitious. Just from this article, porting DRM to Hurd? It's hard enough finding college students who are familiar with Linux let alone Hurd. I don't think a college student would be capable of undertaking a project like that.

The H.264 Gallium3D decoding sounds like another one thats too big for a student.
I haven't taken a look at what other projects are doing yet, but maybe they're not as ambitious with their ideas.

[TechMage89]

I would like to work on a GSoC project, but I can't find one that interests me and for which I have the requisite skill set.

These projects all look quite hard. I would love to dive into X.org development, but a lot of these tasks seem to require highly specialized knowledge.

Also, why are we jumping straight to full H.264 decoding? Wouldn't a motion-compensation implementation using vaapi and Gallium3D be the right place to start? I feel like that would be more doable (esp. for someone like me who would have to learn the Gallium3D instruction set and the particular details of h264 mo-comp and in-loop deblocking).

[Veerappan]

Also, why are we jumping straight to full H.264 decoding? Wouldn't a motion-compensation implementation using vaapi and Gallium3D be the right place to start? I feel like that would be more doable (esp. for someone like me who would have to learn the Gallium3D instruction set and the particular details of h264 mo-comp and in-loop deblocking).

I'd have to agree with something like this. I know from experience that picking up a new infrastructure (Gallium3D, or in my case OpenCL), and then trying to implement something like a full video decoder (or at least subpixel prediction, IDCT/Dequant, and loop filtering in the case of VP8) is a very ambitious project. You can get something basic running in a 2-3 month time-frame, but it doesn't leave ANY room for real optimization of the algorithms. You'd have to have all of that analysis already done before the project actually started.

Just learning the Gallium/VA-API APIs and the actual logic of the video decode algorithms will take a while to do properly.

I'd say it'd be better for someone to attempt just the motion compensation (sub-pixel prediction) or loop filtering in a project like this, and then expand the scope a bit if you get done early, or just spend time optimizing the crap out of it.

[deanjo]

A lot of the GSOC projects are really ambitious.

I realize that, it would be nice however if they would continue on with some of these projects in the following GSoC instead of starting an new project all the time and leaving it unfinished.

[smitty3268]

I used to look forward to GSoC but nowdays I can't be bothered with it. Too many GSoC projects get started, reach about 50-60% completion and then die a slow agonizing death never to be looked at again. It's a great idea but it has a very poor execute to completion record.

Other projects actually have good records. I think a lot has to do with the ease of contributing. It's tough to get into X or Mesa in only 3 months, unless you're already pretty familiar with the codebase and have contributed to it before. Projects like KDE are much easier to quickly contribute to.

[deanjo]

Other projects actually have good records. I think a lot has to do with the ease of contributing. It's tough to get into X or Mesa in only 3 months, unless you're already pretty familiar with the codebase and have contributed to it before. Projects like KDE are much easier to quickly contribute to.

I haven't seen much from GSoC commited to KDE.

[smitty3268]

I haven't seen much from GSoC commited to KDE.

Something like 90% of the projects were successful, i believe. Now whether or not those projects were useful or not, is another question.

[mattst88]

I used to look forward to GSoC but nowdays I can't be bothered with it. Too many GSoC projects get started, reach about 50-60% completion and then die a slow agonizing death never to be looked at again. It's a great idea but it has a very poor execute to completion record.

Absolute bullshit.

http://code.google.com/p/google-summ...gramStatistics

The lowest completion rate ever was 80%. Last year it was 89%.

Could you really not be bothered to do a 15 second search?