Mesa 10.4 Has Been Branched - No OpenGL 4 This Year

[gutigen]

Wasn't there supposed to be on disk shader cache for Mesa in this branch?

[Marc Driftmeyer]

Them being able to do what they have done in a short amount of time makes up for not reaching OpenGL 4 compliance. Everybody involved keep up the excellent work.

Without all the heavy lifting from corporations this stack is still below OpenGL 2.1 status.

[gutigen]

Without all the heavy lifting from corporations this stack is still below OpenGL 2.1 status.

That's vendors job after all.

Everyone benefits in one way or another from corporate involvement in Linux projects. Obviously as long as open source is not hindered by some ridiculous licenses like CLA (though Canonical brings users to Linux, which is good thing in itself).

[MistaED]

Those features are very important (tessellation, compute, SSBO's). Although Most engines can fall back to GL 3.3 for now. I'd also like to see some of the sparse texture/buffer type stuff to land in and bindless but these are extensions.

Currently Unreal Engine 4 targets ES2.x, SM4 (GL3.3/DX10) and SM5 (GL4.x/DX11) which is probably most all AAA engines right now. There are some just supporting SM3 (GL2.1/DX9) right now but I assume they will be dropped soon apart from the engines which want to keep consistency over mobile and desktop, however supporting ES2.x+extensions is a similar pipeline so maintaining it might not be too bad. What I'm saying is if Mesa is stable and fast enough for GL3.3 then games should play fine for Steam or whatever, it is in good shape for developers.

I think targeting GL 5 should be faster once that lands an official spec, same for DX12/Mantle, unless the Gallium3D interface needs adjusting to accommodate the refactor, the driver should be far thinner as it will pass most of the responsibility to the client app.

I can't wait for Mesa to support GL4.x spec so that OpenSubdiv can be used, I have a toy renderer project which uses it. Same for the plans to use sparse textures and eventually bindless, but no major engine that I know of are using these special things as DX11 doesn't do them so there isn't a huge need yet.

[mmstick]

Is this race for GL4 compliance so important after all?? I mean things that the average user uses everyday need some attention also.

What else should they be working on? Those 'things that the average user uses every day' are already completed. The whole point of Mesa is that it is an open source OpenGL library -- it would be a contradictory if they did not care to implement the full GL spectrum. If Mesa supported GL4 then there'd no longer be an argument for anyone to keep using Catalyst.

[DanL]

Without all the heavy lifting from corporations this stack is still below OpenGL 2.1 status.

What???

[bridgman]

What???

I think he means "would be still below OpenGL 2.1", not "is below OpenGL 2.1"...

[Mat2]

How much work is needed to complete the missing features?

BUMP!