So how long will it take until GL2/improved 3D stack on OSS drivers?
Like the topic title says. Dont get me wrong, I am not complaining or anything, I LOVE the OSS drivers, and I in turn love the devs for making such good work in a short time (and AMD for releasing documents).
my system has never been so stable before
But also, I'm waiting for the work on 3D to start, since I'd really love if I dont need to switch to FGLRX or Windows to play a game.
So my questions are:
1) Will it be after the R600/R700 support is added to the drivers? After DRI2?
2) and what can we expect? Can we expect OpenGL 2, GLSL, and better performance? and by how much?
3) Oh, and, will compatibility of Wine be like that of nVIDIA, or will it be like fglrx?
Thanks a lot, and keep up the good work OSS ATI devs
I'm essentially just going to rehash what bridgman has already said, but I think it may be a useful starter to this topic:
-r600-700 initial support (EXA and texturedvideo, with a "sample" 3D implementation) is already pretty much done and just needs to make it through final IP review (along with the documentation). After that, there should be enough out that with the requisite work, a feature-complete driver should be able to be made.
-Basic 3d support (1.3) comes first to understand how the hardware works, then a rebased driver against Gallium that should provide OpenGL 2.0 and probably 3.0 in short order.
-Wine's compatibility is affected not only by bugs in the driver, but also by the fact that Wine devs have built Wine around the Nvidia drivers, so they rely on assumptions that things work the way Nvidia's drivers make them work. As a result some of the Wine compatibilty fixes actually have to be on the Wine end, not always the driver end. Working with software Mesa would probably be a good start for now.
Moving to my own guesses, I would say that after r600-700 docs are released, a lot depends on how many developers work on it. I have the interest, but no experience in writing drivers (although I can write user-level programs fairly well) and little time. If I had the ability and time, though, I could write an OpenGL 2.0 Gallium3D driver for my card (r500), because all the docs for it are already out (except for video decode).
I'm guessing it will be about a month for initial r600-700 support, and about a year for OpenGL 2.0 and 3.0 with decent performance, using Gallium3D.
Last edited by TechMage89; 12-22-2008 at 09:52 PM.
MrCooper (TG) mentioned today that the Mesa changes to support GL2.0 over Gallium3D are in place today and running OK over softpipe, which is further along than I thought. After the holidays I'm going to suggest that we take a look at Gallium3D and consider whether we should switch over sooner than planned. My guess is that it will still make sense to go for 5xx parity with classic Mesa, assuming that's as far as we can go by just plugging in 6xx/7xx programming sequences, but if it turns out we need to change a lot more then jumping to Gallium early might make sense.
I need to check whether Gallium3D has merged to mesa master yet. I think I may have been subconsciously waiting until that happened before committing to Gallium3D
I'm following this as closely as I can. Hey, I might be a computer engineer, so this stuff is interesting to me! I only wish I had the time and knowledge to actually be involved in driver development. I can't wait for a working OpenGL 2.0 implementation for my x1600.
Time is a challenge for everyone. I think most people get involved in driver development when a specific bug is getting on their nerves. They learn just enough of the driver to figure out how that specific function works but there's one thing that doesn't seem to make sense -- they ask on IRC or mail, one of the devs says "that doesn't sound right", and presto they just found their first bug.
Repeat, and after a while you will have knowledge but no time