Intel Aims For Open-Source OpenGL 3.0 Driver By Year's End

[pingufunkybeat]

Yoohoo!!! w00t w00t!

That's very good news. I hope that they continue working on GLSL after that, since that's the major problem with OpenGL 3.0 - 3.3.

[Kayden]

Will this cover the G45 (4500MHD) chips?

Hopefully! So far, I've been enabling features on all Gen4+ chips, including G45. I believe it should be possible.

I still don't have OpenGL2 I think...

You should have OpenGL 2.1 and GLSL 1.20 support, plus a bunch of extra extensions. I'm pretty sure Gen4 has supported GL 2 since the initial 2006 release...a really long time in any case.

btw: what's the code name of the G45? It's not Sandy Bridge is it?

Eaglelake for desktops, Cantiga if you have a GM45 (mobile). We usually just call them G45/GM45 though.

Here's a handy codename translator:
Gen 4 Desktop - Broadwater (BRW) - G965 - GMA X3000
Gen 4 Mobile - Crestline (CL) - GM965 - GMA X3100
Gen 4.5 Desktop - Eaglelake (ELK) - G45 - GMA X4500HD
Gen 4.5 Mobile - Cantiga (CTG) - GM45 - GMA 4500MHD
Gen 5 - Ironlake (ILK) - HD Graphics (desktop sometimes "Clarkdale", mobile "Arrandale")
Gen 6 - Sandybridge (SNB) - HD Graphics 2000 (codename GT1) and HD Graphics 3000 (codename GT2).

[smitty3268]

The floating point texture/framebuffer patent is the issue, though. You can't have GL 3.0 without licensing, invalidating, or circumventing that patent. Unless Intel thinks that patent is no longer a problem, Mesa cannot and will not support OpenGL 3.0+ for another 10 years (or however long it will be until the patent expires; I forget the exact date).

Well, Mesa itself would support OpenGL 3. It's just the version that your distro ships might not.

I expect that once Mesa 8.0 arrives the unofficial 3rd party repos might create Mesa packages with the patented float code more often. Right now, they don't seem to place a high priority on it but the difference between an OpenGL 2 and 3 driver might be enough to make it more noticeable.

[cl333r]

The floating point texture/framebuffer patent is the issue, though. You can't have GL 3.0 without licensing, invalidating, or circumventing that patent. Unless Intel thinks that patent is no longer a problem, Mesa cannot and will not support OpenGL 3.0+ for another 10 years (or however long it will be until the patent expires; I forget the exact date).

And don't forget that the corporate lobby might extend the life of software patents from 20 years to say 40. It's not uncommon. So just waiting for it to expire might be a BIG mistake. Better try to invalidate it or plant a big chunk of dynamite into the building of the company who owns the patent and BOOM!! No patent to fear any longer, no company to sue you over the patent. Problem solved. Next one please.

[elanthis]

IIRC, GLSL 1.3 is the only major roadblock left for 3.0, so there might not be a significant difference between those two in terms of work left, at least from a software acceleration standpoint. Look at the list: http://cgit.freedesktop.org/mesa/mesa/tree/docs/GL3.txt

The floating point texture/framebuffer patent is the issue, though. You can't have GL 3.0 without licensing, invalidating, or circumventing that patent. Unless Intel thinks that patent is no longer a problem, Mesa cannot and will not support OpenGL 3.0+ for another 10 years (or however long it will be until the patent expires; I forget the exact date).

[DanL]

That is, he might have meant the driver supports GLSL 1.30 rather than the whole GL 3.0 API itself.

IIRC, GLSL 1.3 is the only major roadblock left for 3.0, so there might not be a significant difference between those two in terms of work left, at least from a software acceleration standpoint. Look at the list: http://cgit.freedesktop.org/mesa/mesa/tree/docs/GL3.txt

[elanthis]

I just want to spread love all over right now! Michael, you make me so happy! I work in a med-tech-company with visualization using OpenGL (which I prefer since I'm a Linux nerd), and we can't use anything more modern than OpenGL2.1 since our code is often run on virtualized machines (Xen). The virtualization means that there is only a software driver, so we depend heavily on Mesa (I think at least!).
When these changes are released, we can finally start looking at porting everything to OpenGL 3, at least bit by bit.

For what it's worth, it's not until OpenGL 3.1 through 3.3 that you start seeing any real utility out of "porting" to GL3. The idea of the core/compatibility profiles doesn't even exist until 3.1 iirc, and all 3.0 did was make a few extensions core. It was pretty worthless, hence all the immense outrage from GL developers when Khronos released it in 2008. The really cool features of the GL 3.x series didn't show up until GL 3.2, and the performance of the API isn't even capable of achieving parity with D3D 10 until GL 3.3. Same goes with GL 4.x, where D3D 11 still has features that GL doesn't (despite all the nonsense of various articles and announcements claiming that 4.0 had feature parity with D3D 11).

We need GL 3.0 to get to GL 3.3, but don't expect the world to become a magical place when Mesa supports GL 3.0. It'll basically just mean that floating point textures are supported and integers in shaders are working properly. The first already works but is disabled due to legal issues, and the latter should be done soon, which may be all that Ian meant when he said that the Intel driver would support GL 3. That is, he might have meant the driver supports GLSL 1.30 rather than the whole GL 3.0 API itself.

[darkbasic]

It works for me!

Faster than a slideshow? Can you please post your experience there: http://phoronix.com/forums/showthrea...vailable/page4

[elmariachi]

Probably like the H264 VAAPI G45 support

It works for me!

[Azpegath]

I just want to spread love all over right now! Michael, you make me so happy! I work in a med-tech-company with visualization using OpenGL (which I prefer since I'm a Linux nerd), and we can't use anything more modern than OpenGL2.1 since our code is often run on virtualized machines (Xen). The virtualization means that there is only a software driver, so we depend heavily on Mesa (I think at least!).
When these changes are released, we can finally start looking at porting everything to OpenGL 3, at least bit by bit.