ARB_gpu_shader_int64 Patches For Intel's Mesa Driver

[phoronix]

Phoronix: ARB_gpu_shader_int64 Patches For Intel's Mesa Driver

Well known Intel open-source developer Ian Romanick has published a massive set of 59 patches for enabling GL_ARB_gpu_shader_int64 within core Mesa and their i965 DRI driver...

http://www.phoronix.com/scan.php?page=news_item&px=Int64-Mesa-Intel-Patches

[smitty3268]

RadeonSI support is already done as well. Just needs to be flipped on after this lands.

[Luke_Wolf]

So question, If we were to presume that no new extensions or versions of OpenGL were to be released between now and when Mesa finishes off the current list of "other extensions" on mesamatrix.net will the Mesa OpenGL stack be feature complete or will there be more extensions to implement after that?

[Serafean]

So question, If we were to presume that no new extensions or versions of OpenGL were to be released between now and when Mesa finishes off the current list of "other extensions" on mesamatrix.net will the Mesa OpenGL stack be feature complete or will there be more extensions to implement after that?

You can always create your own extensions
Feature complete, that depends on who you ask : regarding openGL extensions : yes. After that it depends on how you count features like on disk shader cache, multithreaded GLSL and other stuff.
Then there are drivers that support hardware that could support later GL extentsions (looking at nv50 and r600g + all ARM drivers)
Assuming of course that you were only talking about OpenGL. After this there is vaapi, vdpau, openmax, Vulkan...
And finally Performance optimizations.

Great milestone Mesa guys! But the race isn't over.

[ElectricPrism]

Question, after MESA achieves feature complete status how much room will there be for optimization? Will that become second priority or will the drive be not as much to optimize?

[OneTimeShot]

will the Mesa OpenGL stack be feature complete or will there be more extensions to implement after that?

As far as I can see, until there is an OpenGL 4.6, Mesa is "feature complete". There shouldn't be any developers writing code that mandates the optional extensions on top of OpenGL 4.5 (sure - they can test for them being there and use them if they are). I guess that implementing them means that Mesa would have "day0" support for the next version of OpenGL.

IMO it is more important to keep on top of Vulcan, though. I can already hear the whinging: "game devs should have tested their code with RADV" which is currently a bit tricky to do

[DMJC]

Or they could finish up support for the API already in use by all the major desktop environments/APIs and work on Vulkan when it's actually needed. Considering the lack of Vulkan ports/titles. It'd be nice to have fully completed OpenGL support before the developers go blow a ton of time working on an API that has minimal market share.

[-MacNuke-]

There are not many OpenGL 4.4/4.5 games out there. DOOM uses OpenGL 4.3 + some extensions if they are there. No Man's Sky asks for OpenGL 4.5 but can be run on the Radeon HD 6000 series which only has OpenGL 4.4.

Since macOS only supports OpenGL 4.1 most ports stick to that version. As far as I read Alien Isolation uses OpenGL 4.3 with a workaround for macOS.

So there are not many reasons to support OpenGL 4.5. Vulkan adoption might be faster than any OpenGL 4.5 adoption. Porting future DX12 games will be faster and less problematic when using Vulkan on Linux and Metal on macOS than using OpenGL at all.

[pal666]

Will that become second priority or will the drive be not as much to optimize?

what do you mean? that mesa will ban optimizations? there is a drive to optimize already

[pal666]

So there are not many reasons to support OpenGL 4.5.

even if game does not require some extension, it could use it optionally for some benefits