Core Mesa Is Now Just One Step Away From OpenGL 4.3 Compliance

[DMJC]

The important thing is OpenGL 4.5. It's the last major release of OpenGL. Once the drivers have supported that, then there is optimisation and bugfixing to be done before Mesa OpenGL is complete. My guess is that a lot of the OpenGL development work will be dropped once Vulkan is released and a large number of contributors try to move onto adding Vulkan. I'd like to see OpenGL get to a nice optimised point sooner than later, but feature completeness is a must. Looking at Mesa Matrix suggests that a lot of 4.4 and 4.5 is going to get completed very quickly. Most extensions have already had work done, and looking at what remains in 4.3 it appears that 4.4 and 4.5 rely on that work.

[RussianNeuroMancer]

if there is much effort we dont know, maybe from the 5 people that are on it, its really much effort, compared to proprietary drivers its maybe 1/100 effort. Maybe the factor is only 1/20 or so dont know the exact numbers but "much effort" is something relativ...

I dont want to bitch but valve could have interest in good free intel and amd drivers because they dont want to make their customers have to use fglrx, so they could pay 20 fulltime people, I know they do SOME stuff, thats always better than nothing, but maybe I am wrong what I read so far sounded more like 1-5 full-time developers on that front if not less.

You forgot about Vulkan. This is what Valve is care about, not OpenGL.

[geearf]

The important thing is OpenGL 4.5. It's the last major release of OpenGL. Once the drivers have supported that, then there is optimisation and bugfixing to be done before Mesa OpenGL is complete. My guess is that a lot of the OpenGL development work will be dropped once Vulkan is released and a large number of contributors try to move onto adding Vulkan. I'd like to see OpenGL get to a nice optimised point sooner than later, but feature completeness is a must. Looking at Mesa Matrix suggests that a lot of 4.4 and 4.5 is going to get completed very quickly. Most extensions have already had work done, and looking at what remains in 4.3 it appears that 4.4 and 4.5 rely on that work.

I don't believe OGL is end of life, we might still get new OGL in the future, so once mesa caught up to 4.5, it might still have to catch up to 4.6 or more... Of course I do hope I'm wrong about that.

[smitty3268]

I don't believe OGL is end of life, we might still get new OGL in the future, so once mesa caught up to 4.5, it might still have to catch up to 4.6 or more... Of course I do hope I'm wrong about that.

There will almost certainly be a GL 4.6 out that includes some Vulkan integration. At the very least the ability to run SPIR-V shaders.

[Chewi]

I'm curious about how much these drivers have in common. I know that Gallium unified things a lot but clearly there are still a lot of hardware-specific aspects to this. Does implementing this stuff in one driver at least give you a good clue about how to implement it in the others?

[bug77]

They might want a complete working implementation that behaves according to the specification as much as possible before they start looking into serious optimization.

A complete working implementation of what? As I understand Mesa has complete working implementation for everything up to OpenGL 4.2. Has work started into optimizing the performance for OpenGL 3.x?

[eydee]

But why bother with implementing it for Intel before they come up with a GPU that can run anything 3D faster than 2 fps?

[Michael]

The important thing is OpenGL 4.5. It's the last major release of OpenGL. Once the drivers have supported that, then there is optimisation and bugfixing to be done before Mesa OpenGL is complete. My guess is that a lot of the OpenGL development work will be dropped once Vulkan is released and a large number of contributors try to move onto adding Vulkan. I'd like to see OpenGL get to a nice optimised point sooner than later, but feature completeness is a must. Looking at Mesa Matrix suggests that a lot of 4.4 and 4.5 is going to get completed very quickly. Most extensions have already had work done, and looking at what remains in 4.3 it appears that 4.4 and 4.5 rely on that work.

NO, there will still be more revisions of OpenGL coming... OpenGL is not dead. I've heard it quite clear from Khronos they intend to continue to maintain and evolve it.

[Michael]

But why bother with implementing it for Intel before they come up with a GPU that can run anything 3D faster than 2 fps?

Their Iris Graphics today are really good... But they're the ones funding many developers to work on it plus also funding Collabora and Igalia, which is why it gets implemented for them.

[log0]

I'm curious about how much these drivers have in common. I know that Gallium unified things a lot but clearly there are still a lot of hardware-specific aspects to this. Does implementing this stuff in one driver at least give you a good clue about how to implement it in the others?

I think it depends on how similar or different the hardware is. The differences between r600 and radeonsi will be smaller than between r600 and intel e.g.