Zink OpenGL-on-Vulkan Now Offers Broader OpenGL Coverage Than RadeonSI, Intel

[phoronix]

Phoronix: Zink OpenGL-on-Vulkan Now Offers As Broad OpenGL Coverage As RadeonSI, Intel

When it comes to OpenGL extension support, the Zink generic OpenGL-on-Vulkan implementation now has as robust coverage as core Mesa offers and what is implemented by the LLVMpipe software driver, RadeonSI Gallium3D, and the Intel i965 drivers...

https://www.phoronix.com/scan.php?page=news_item&px=Zink-OpenGL-Exts-160

[FireBurn]

I think you missed the main point of his blog, zink now offers *more* GL extensions that aren't apart of any particular release. Zink implements 39 vs radeonsi (the next nearest) with 37

Extensions that are not part of any OpenGL or OpenGL ES version
There is a total of 54 extensions to implement. The ranking is based on the number of extensions done by driver.

# Driver Extensions
1 mesa (85.2%) 46
2 zink (72.2%) 39
3 radeonsi (68.5%) 37
4 i965 (64.8%) 35
5 nvc0 (50.0%) 27
6 softpipe (37.0%) 20
7 r600 (35.2%) 19
7 llvmpipe (35.2%) 19
9 freedreno (31.5%) 17
10 panfrost (29.6%) 16
11 nv50 (27.8%) 15
12 virgl (20.4%) 11
13 lima (18.5%) 10
14 etnaviv (14.8%) 8
14 v3d (14.8%) 8
16 vc4 (13.0%) 7
16 d3d12 (13.0%) 7

[Michael]

I think you missed the main point of his blog, zink now offers *more* GL extensions that aren't apart of any particular release. Zink implements 39 vs radeonsi (the next nearest) with 37

Extensions that are not part of any OpenGL or OpenGL ES version
There is a total of 54 extensions to implement. The ranking is based on the number of extensions done by driver.

# Driver Extensions
1 mesa (85.2%) 46
2 zink (72.2%) 39
3 radeonsi (68.5%) 37
4 i965 (64.8%) 35
5 nvc0 (50.0%) 27
6 softpipe (37.0%) 20
7 r600 (35.2%) 19
7 llvmpipe (35.2%) 19
9 freedreno (31.5%) 17
10 panfrost (29.6%) 16
11 nv50 (27.8%) 15
12 virgl (20.4%) 11
13 lima (18.5%) 10
14 etnaviv (14.8%) 8
14 v3d (14.8%) 8
16 vc4 (13.0%) 7
16 d3d12 (13.0%) 7

Yeah Mike mentioned it, updated it now to reflect that table, not used to MesaMatrix tables when looking at extension counts.

[oiaohm]

I think you missed the main point of his blog, zink now offers *more* GL extensions that aren't apart of any particular release. Zink implements 39 vs radeonsi (the next nearest) with 37

Extensions that are not part of any OpenGL or OpenGL ES version
There is a total of 54 extensions to implement. The ranking is based on the number of extensions done by driver.

# Driver Extensions
1 mesa (85.2%) 46
2 zink (72.2%) 39
3 radeonsi (68.5%) 37
4 i965 (64.8%) 35
5 nvc0 (50.0%) 27
6 softpipe (37.0%) 20
7 r600 (35.2%) 19
7 llvmpipe (35.2%) 19
9 freedreno (31.5%) 17
10 panfrost (29.6%) 16
11 nv50 (27.8%) 15
12 virgl (20.4%) 11
13 lima (18.5%) 10
14 etnaviv (14.8%) 8
14 v3d (14.8%) 8
16 vc4 (13.0%) 7
16 d3d12 (13.0%) 7

That not the complete story.

Yeah Mike mentioned it, updated it now to reflect that table, not used to MesaMatrix tables when looking at extension counts.

There is a total of 4 extensions that zink does not implement but at least 1 intel or amd driver do implement. Those percentages and raw count are not the full story.

The bad news is one that not done yet could be a performance problem with some games.

https://www.khronos.org/registry/OpenGL/extensions/ARB/ARB_parallel_shader_compile.txt

The ablity to tell the opengl system how many threads to consume when building shaders zink is still missing.

Fun optimisation limit shader compiler to 1 thread and give up a stack of shaders that will be need in a future section of game when player gets to the loading area increase the thread count to the shader compiler.

Zink developer had done great work so far. But its a little too soon to say done with extensions in my eyes.

[andre30correia]

nouveau driver need a vulkan driver and left opengl driver. with the things like this many dont need native opengl driver

[caligula]

Virgl should be at 4.6. This is needed for vm uses.

[rmfx]

nouveau driver need a vulkan driver and left opengl driver. with the things like this many dont need native opengl driver

That's the goal. One native Vulkan driver by vendor, and a universal Zink driver for OpenGL support.
When Zink achieves top conformance and perf, it's going to be the only gl driver to focus on basically.

[risho]

is the goal for opengl to get absorbed into vulkan using zink so that it doesn't need to be maintained by each driver and product individually and can all be maintained there? or is this just a fun side project? what is the ultimate goal of this project?

[ZuseZ4]

is the goal for opengl to get absorbed into vulkan using zink?

I'm curious what the opengl group (and others related) think about that.
Usually I'd vote for having more competition. Here however, it seems that zink as a global solution seems to outperform the individual solutions,
so as an end-user in this case I'd rather have zink.

[billyswong]

is the goal for opengl to get absorbed into vulkan using zink so that it doesn't need to be maintained by each driver and product individually and can all be maintained there? or is this just a fun side project? what is the ultimate goal of this project?

My guess is some future GPUs will skip opengl hardware support and leave it all to zink. Percentage will rise to 100% in 10 to 15 years.