Radeon Driver Rewrite Only Has A Few Things Left

[smitty3268]

ok, so this doesn't have to do anything with a gallium3d-driver (aka drm-driver?), does it?

I don't think it does directly, no. But Gallium3D requires things like a memory manager, and this provides GEM support, so this is sort of a pre-requisite to finish up before the gallium work can really begin.

Can anyone confirm that Gallium3D work will begin soon after this is merged to master? Or is that still waiting on other work, like improvements to LLVM?

[Regenwald]

I don't think it does directly, no. But Gallium3D requires things like a memory manager, and this provides GEM support, so this is sort of a pre-requisite to finish up before the gallium work can really begin.

Can anyone confirm that Gallium3D work will begin soon after this is merged to master? Or is that still waiting on other work, like improvements to LLVM?

hm llvm itself is already optimized for shaders i thought. at least apple uses it for their shader optimizations, too.

[Adarion]

Good to read the driver's progressing. I hope it'll also inherit all sorts of power management (clock it down, switch of unneded parts that are unneeded for the time).

[monraaf]

Like others I find all this stuff about GEM/TTS/KMS/whatever very confusing and frankly I don't really want to know.

What I do want to know is when OpenGL support is available for r6xx/r7xx how do I get it on my machine. What other stuff besides some specific branch of mesa do you have to install. I sure hope someone writes a howto when the OpenGL r6xx/r7xx support is in the public repositories.

[bridgman]

The radeon-rewrite effort is for r1xx-r5xx, but we're going to be building on it for 6xx-7xx as well. The rollout of 3d code for 6xx-7xx will probably follow a similar sequence to what EXA/Xv required :

- first you'll need both drm and mesa from branches (for EXA/Xv it was drm and ddx)
- after a while mesa will get merged to master but you'll still need drm from a branch
- eventually the drm code will end up in the kernel tree and you won't need anything special

I'm sure there will be multiple "how to build it" guides

[FunkyRider]

To normal users:

Stick with nvidia binary driver for another year, and look back afterward to decide whether to switch to open source stack or not.

I change my graphics card in a 3-6 month pace so I will enjoy my nvidia 3D experience while keeps an eye on oss side. Just sold my beloved HD 4850 because I can't do any useful OGL work under Linux because the driver is too broken... I can't even create a workable x pixmap, render to it and write it to disk. The program I wrote immediately works after I switched to nvidia stack. Well, what can I say... better is better

Also the tear-free (v-synced) compiz desktop is such a joy to use...

[dungeon]

Woohoo! This is the best news i read at phoronix recently, after AMD specs releases. Thank you Michael for news and thank you Dave for doing this!

[hirolau]

The radeon-rewrite effort is for r1xx-r5xx, but we're going to be building on it for 6xx-7xx as well.

I have a Radeon Mobility FireGL v5200. And it say it is an M56 not rxxx. Is M56 the same as RV530?

[bridgman]

Yes. The drivers usually bring up the name of the consumer equivalent, which is X1600 or RV530, so your GPU is in the 5xx family. The rv530 designation describes the chip die, while M56 or V5200 describes a combination of die, tested clock speeds, memory configuration etc...

[hmmm]

This rewrite looks to be an advantage for older radeon users over windows now that the fgrlx/catalysts are abadoning support for older cards; with the exception of a future one-off driver (like the r100 7.4 catalysts iirc), windows users of older ati hardware have all the more reason to shift to linux.