Live data from Hacker News

VirGL – A virtual 3D GPU for use inside QEMU virtual machines

docs.mesa3d.org

11–20 of 84 posts

Re: VirGL – A virtual 3D GPU for use inside QEMU virtual machines

#12
post #10

Its virtual... but passthrough, I would like to see something thats virtual but runs on pure cpu x86 or arm, do not care how slow.

See LLVMpipe a little ways up the list in the left hand pane. It should do exactly what you're looking for.

It's worth noting the terminology "passthrough" means something else in this particular context and VirGL is not considered a passthrough solution (as noted in the "out of scope" section). VirGL is a form of paravirtualization often called "API remoting".

Re: VirGL – A virtual 3D GPU for use inside QEMU virtual machines

#13
I've always found it interesting, and a bit unfortunate, that it seems like things like apitrace are generally pretty reliable, but tools like VirGL seem to often have trouble with typical programs. It seems like the problem of accurately serializing what is being done to the GPU using an API is possible, but in practice there is clearly a lot more to it than that. (Of course, in case of remoting, you do not have the benefit of observing the interactions with a known-good driver and with direct memory mapping/etc. So, that probably answers the "why" right there.)

It's been a while, though. Maybe this works really well... I'm curious to hear if anyone has more recent experience messing with these things.

Re: VirGL – A virtual 3D GPU for use inside QEMU virtual machines

#14
There is a project I am following to embed a Linux machine into mobile VR applications using virtio-gpu & his latest effort gets 200fps.

"VR Linux is now testing with working GPU acceleration via the kernel over virtio-gpu. This is a major step forward following on from the Shadertoy device that we made a few months ago. Now we can make OpenGLES Linux programs that are accelerated by the Quest's Adreno GPU."

[1] Thread: https://twitter.com/anjin_games/status/1371870094490537987?t...

Re: VirGL – A virtual 3D GPU for use inside QEMU virtual machines

#15
post #10

Its virtual... but passthrough, I would like to see something thats virtual but runs on pure cpu x86 or arm, do not care how slow.

On Windows, there's a pure software implementation of Direct3D 11 called WARP https://learn.microsoft.com/en-us/windows/win32/direct3darti...

Re: VirGL – A virtual 3D GPU for use inside QEMU virtual machines

#19
post #8

uh, this doesn't look like what those gpu passthrough things like vfio do, is it a different thing? Or I am more interested in, can this do what vfio do, run games in almost native performance in windows guest vm?

Virgl just sends opengl commands to the host. And it also has to copy some buffers back and forth. So it won't be as fast as vfio. But, there is a new feature called VirtGPU DRM native contexts which eliminates some of this overhead and run at near native speeds. Unfortunately it's only implemented for adreno/msm GPUs for now.I think there is work on the way for Intel and AMD GPUs.

Re: VirGL – A virtual 3D GPU for use inside QEMU virtual machines

#20
I've never been able to get this to work under UTM on an ARM Mac; if anyone knows the magic qemu and VM settings, please drop me an email (email in profile).

It would be nice to have fullscreen, full res 3d accelerated VM guests finally. I've never been able to get this to work as smoothly as I think it should.

Post reply on HN