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

[Quackdoc]

https://www.khronos.org/openglsc/

OpenGL SC would be validating what is already implemented in Zink and making a subset and validating the vulkan drivers Zink is using. OpenGL SC Lot of work validating.

OpenCL is a far harder problem. Mesa clover implementation of opencl needs major work and its only really clover 1.2. Very broken opencl 1.2. So clover getting working on zink would still leave lots of work ahead.

Maybe with Mr.Ekstrand's new job at collabora, we might actually see this happen, he did hint towards it being a potential point of work.

[rmfx]

Maybe Zink could support OpenCL and OpenGL SC?

I would prefer OpenCL support to be included in an other driver.
No need to bloat Zink with more features than full OpenGL.

[Quackdoc]

I would prefer OpenCL support to be included in an other driver.
No need to bloat Zink with more features than full OpenGL.

It wouldn't be bloat. Zink is a gallium driver. NOT just an OGL to VK layer like dxvk. something like OCL is well within the scope of zink. just like gallium nine is. Zink becoming a better more accurate gallium driver should be the goal. I even think Mr. Blumenkrantz had done some preliminary work on it.

The larger issue has already been stated. Clover is down right terrible implementation.

[oiaohm]

I would prefer OpenCL support to be included in an other driver.
No need to bloat Zink with more features than full OpenGL.

As the developer behind Zink has made a few statements.

https://linuxreviews.org/The_State_O..._In_April_2021

The major missing bit from Zink being WSI support is that was noted in April 2021 that prevent clover from working yes the developer notes this.

https://www.supergoodcode.com/NV-envy/

Clover is meant to be mesa opencl for galluim drivers.

The reality the features Zink in the galluim driver needs to implement so clover works Zink need to implement so opengl works and passes the complete test suite.

Of course this still has the problem that mesa clover really needs major amounts of development itself to be useful. Clover works at the moment is like 95 percent of opencl applications don't work. Yes that is most likely being kind saying ~5 percent work. Clover is in a horrible broken state without a decanted developer to pull it out of hell.

Yes some point in the future Zink will absolute support Opencl just it will not be a useful implementation of opencl if nothing changes on the development side of clover.

[Quackdoc]

As the developer behind Zink has made a few statements.

https://linuxreviews.org/The_State_O..._In_April_2021

The major missing bit from Zink being WSI support is that was noted in April 2021 that prevent clover from working yes the developer notes this.

https://www.supergoodcode.com/NV-envy/

Clover is meant to be mesa opencl for galluim drivers.

The reality the features Zink in the galluim driver needs to implement so clover works Zink need to implement so opengl works and passes the complete test suite.

Of course this still has the problem that mesa clover really needs major amounts of development itself to be useful. Clover works at the moment is like 95 percent of opencl applications don't work. Yes that is most likely being kind saying ~5 percent work. Clover is in a horrible broken state without a decanted developer to pull it out of hell.

Yes some point in the future Zink will absolute support Opencl just it will not be a useful implementation of opencl if nothing changes on the development side of clover.

I think I recently saw some zink WSI bits get merged into mainline not too long ago

[oiaohm]

I think I recently saw some zink WSI bits get merged into mainline not too long ago

https://www.supergoodcode.com/NV-envy/ I did quote this because yes the WSI stuff is merged in the zink code base and it is noted there. But we need what was called Copper now called Kopper to hook all the vulkan WSI stuff up and then it will be fully testing the WSI stuff in zink.

So hopefully sometime this year if all the stars align the zink WSI stuff be functional. Of course that still will leave clover that is a huge mess of fairly non functional opencl implementation.

[Quackdoc]

https://www.supergoodcode.com/NV-envy/ I did quote this because yes the WSI stuff is merged in the zink code base and it is noted there. But we need what was called Copper now called Kopper to hook all the vulkan WSI stuff up and then it will be fully testing the WSI stuff in zink.

So hopefully sometime this year if all the stars align the zink WSI stuff be functional. Of course that still will leave clover that is a huge mess of fairly non functional opencl implementation.

yeah, I was talking about wsi bits merged just the other week . here https://gitlab.freedesktop.org/mesa/...requests/14426 maybe eventually we will get good clover. I'm hoping Mr. Ekstrand does some work on it, he did as I said, alude to it in his blog post. it would be really nice.

[oiaohm]

yeah, I was talking about wsi bits merged just the other week . here https://gitlab.freedesktop.org/mesa/...requests/14426 maybe eventually we will get good clover. I'm hoping Mr. Ekstrand does some work on it, he did as I said, alude to it in his blog post. it would be really nice.

That is still missing the Copper/Kopper stuff. So part done.