Google Working On Making VirtIO-GPU More Extensible

[phoronix]

Phoronix: Google Working On Making VirtIO-GPU More Extensible

Google Chrome OS engineers are working on making Linux's VirtIO-GPU driver more extensible. The VirtIO-GPU driver has been modeled around the Virgl protocol for handling 3D within guest virtual machines but with a new context type addition they aim to support additional protocols...

https://www.phoronix.com/scan.php?page=news_item&px=VirtIO-Context-Type

[uid313]

Could it use Zink?

[Quackdoc]

Could it use Zink?

mmmmm. why? I mean, I guess it could, but why? currently Opengl is handled like this.
OGL -> gallium virgl driver -> pipe to host -> Gallium to opengl -> gallium (or whatever opengl driver is used).

Using zink would be
OGL -> Gallium Virgl ->(Either Pipe to host/zink) -> (whatever it didn't do) -> Vulkan

OR the BETTER way of doing it, by letting guest use zink
OGL -> zink -> Venus -> vulkan pipe -> vulkan.

Personally I am far more excited for potential GFXstream capabilities. I would love to compare it with venus. But I would also like to see mesa get GFXstream...

[schmidtbag]

Could it use Zink?

Are you asking whether Google could integrate Zink or are you asking whether Zink would work with this? If the former, I'd agree with what Quackdoc said. Otherwise:

mmmmm. why? I mean, I guess it could, but why? currently Opengl is handled like this.

A program as ubiquitous as Chrome could greatly benefit from having both OGL and Vulkan compatibility. Zink on a lot of GLES systems could give proper hardware acceleration to them, without Google having to go through a lot of heavy development.
In general, I don't think Chrome really needs a Vulkan stack, though I wouldn't complain if it got one.

[uid313]

Are you asking whether Google could integrate Zink or are you asking whether Zink would work with this? If the former, I'd agree with what Quackdoc said.

I was wondering if emulators could make use of Zink.

[schmidtbag]

I was wondering if emulators could make use of Zink.

Oooh lol. Yeah, I can definitely see the benefit in that.

[Quackdoc]

I was wondering if emulators could make use of Zink.

A program as ubiquitous as Chrome could greatly benefit from having both OGL and Vulkan compatibility. Zink on a lot of GLES systems could give proper hardware acceleration to them, without Google having to go through a lot of heavy development.
In general, I don't think Chrome really needs a Vulkan stack, though I wouldn't complain if it got one.

Zink is a gallium driver, it would be best to let the OS handle zink if needed but I suppose it could be baked into a program, but in the end, it would be optimal for it to be done at OS level. the only time I could really see this being useful is for example, android, that has no way to run opengl apps out of box.

[schmidtbag]

Zink is a gallium driver, it would be best to let the OS handle zink if needed but I suppose it could be baked into a program, but in the end, it would be optimal for it to be done at OS level. the only time I could really see this being useful is for example, android, that has no way to run opengl apps out of box.

I agree; I don't think Zink should be baked into a program.

[artivision]

Any Vulkan support for Windows Guests on Linux Hosts?

[Quackdoc]

Any Vulkan support for Windows Guests on Linux Hosts?

No. Qemu doesn't even have support for it yet, so there is no real way to test it regardless so I doubt anyone is working on it at the moment. I think it is a safe assumption that someone may pickup the project in the future. but as of now, none that I am aware of. Viogpudo needs to get modified to work for vulkan passthrough (the virtio-gpu driver for windows), and then you need a userspace vulkan driver for venus.