Live data from Hacker News

3D support for X11 guests

virtualbox.org

21–30 of 74 posts

Re: 3D support for X11 guests

#21
post #9

Without 3d support, basically no-one can use virtualbox, so is the dilemma whether to kill the project or not? (This is a serious question--what can possibly be more important to average virtualbox users than having a working desktop environment? Virtualbox has always occupied the desktop virtualization niche, so I'm trying to figure out what has changed...)

I've used a default Ubuntu desktop install in VirtualBox and it seems to work (though a little laggy). I haven't tried playing video games in it.

Re: 3D support for X11 guests

#22
post #11

Earlier quoted context omitted.

I use Virtualbox daily without ever using a desktop environment in the VM. It's very useful in headless form for development machines when working locally.

I find this baffling when KVM is much more amenable to being scripted from the command line.

KVM isn't an option on OS X or Windows - in those cases, VirtualBox is the most mature free VM software available.

Re: 3D support for X11 guests

#23
post #11

Earlier quoted context omitted.

I find this baffling when KVM is much more amenable to being scripted from the command line.

While I do love KVM, it takes a lot more to get running properly than installing Virtualbox and Vagrant and being up and running with a whole ecosystem of every type of prebuilt box you could imagine.

Yep, the last time I tried to get a KVM machine going I had to hack XML files to get the correct SCSI controller just so I could mount an ISO. It's 3 clicks in VirtualBox.

Re: 3D support for X11 guests

#24
post #13

Obvious question, why not just run an X11 server on the host? Perhaps Oracle and Microsoft (WSL) can collaborate on supporting an X11 server (ones already exist) for Windows 10.

Can't you already do that? With DISPLAY variable. Although there could be less overhead in running local (guest) X server and stream only GL commands.

Re: 3D support for X11 guests

#26

Earlier quoted context omitted.

While I do love KVM, it takes a lot more to get running properly than installing Virtualbox and Vagrant and being up and running with a whole ecosystem of every type of prebuilt box you could imagine.

Yep, the last time I tried to get a KVM machine going I had to hack XML files to get the correct SCSI controller just so I could mount an ISO. It's 3 clicks in VirtualBox.

virt-manager makes it much easier

https://virt-manager.org

Re: 3D support for X11 guests

#28
post #13

Obvious question, why not just run an X11 server on the host? Perhaps Oracle and Microsoft (WSL) can collaborate on supporting an X11 server (ones already exist) for Windows 10.

That would work, but perform like caro. There are a large number of effects that need shared memory tricks to work efficiently in X. So, any "modern" desktop is going to be iffy running that way.

Re: 3D support for X11 guests

#29

I don't understand how this is a "dilemma." It's like saying, "on one hand, I really need to mow my lawn. On the other hand, I really don't want to . It's a real dilemma!"

They mow the grass on a park that's open to us all to use, but they don't have time to mow that bit over there. They're just asking if anyone else has time to do it instead.

Re: 3D support for X11 guests

#30
Virtualbox 3D progress has been unfortunately slow for a long time. It was kind of anticipated with Oracle's acquisition, as their laser focus on legal engagements and consulting pushes a lot of technical projects to the back burner.

I'd love to see a bit more alignment between VirtualBox and Qemu. Even though KVM and Qemu's virtualized graphics acceleration is still a WIP, having a shared code base could accelerate the project on things like SPICE (https://www.spice-space.org/download.html) and Virgl (https://virgil3d.github.io/). Unfortunately, I think it uses a lot of Linux-specific technologies (e.g. KVM, Gallium), so the likelihood of sharing at that level seems pretty low. Although, supposedly Gallium isn't locked into Windows.

Theoretically, if we could agree on a common host-guest interface for a virtual graphics adapter, it could share the guest implementations and host implementations could be added as needed. And it could be reused in multiple projects. But it always seems that this is the tech that never gets sufficient cross-project collaboration. And given the many differences between vendor hardware, a portable interface has been difficult. Maybe Vulkan will provide enough low-level functionality to ease the abstraction?

Post reply on HN