I want metal support in qemu and am going to put 1000 dollars down as a bounty.
Ask HN: How do I do a feature request for GPU support in QEMU
1–8 of 8 posts
Re: Ask HN: How do I do a feature request for GPU support in QEMU
#2I Am Not A Graphics Programmer, but $1000 is not going to buy you Metal support.
Re: Ask HN: How do I do a feature request for GPU support in QEMU
#3I Am Not A Graphics Programmer, but $1000 is not going to buy you Metal support.
Not on its own, but it could be a start...
Re: Ask HN: How do I do a feature request for GPU support in QEMU
#4Re: Ask HN: How do I do a feature request for GPU support in QEMU
#5You can buy a Mac for $1000.
Re: Ask HN: How do I do a feature request for GPU support in QEMU
#6Make a feature request, there is even a feature request label in the QEMU issue tracker - https://gitlab.com/qemu-project/qemu/-/issues/?label_name=ki...
And pick a bounty platform like https://app.bountysource.com/ (top bounty there is also a QEMU issue) to post your bounty
Re: Ask HN: How do I do a feature request for GPU support in QEMU
#7Earlier quoted context omitted.
Not on its own, but it could be a start...
Would need an emulated MacOS GPU. The only way I could see this working is that there is a way through OS X hypervisor on Big Sur or > to passthrough metal.
That's like saying to emulate an ARM or Alpha processor, you need an ARM or Alpha processor to emulate... that's what emulation is about... emulating one CPU/architecture on a more different CPU/architecture...
Re: Ask HN: How do I do a feature request for GPU support in QEMU
#8Earlier quoted context omitted.
Would need an emulated MacOS GPU. The only way I could see this working is that there is a way through OS X hypervisor on Big Sur or > to passthrough metal.
That's like saying to emulate an ARM or Alpha processor, you need an ARM or Alpha processor to emulate... that's what emulation is about... emulating one CPU/architecture on a more different CPU/architecture...
I think you're mistaken, virtualization and emulation are two different things. Metal is not low-level enough to perform high-speed virtualization, what the parent is suggesting is that you'd need to use Metal to accelerate an emulated GPU, which would result in pretty substantial performance degradation. On top of that, most virtual machines wouldn't recognize the M1 APU, which means you'd probably also need to emulate interfaces to compliment it. When all is said-and-done, you'd have an incredibly fragile and slow stack that has a good chance of breaking with every Metal update and almost certainly breaks with each new GPU.
The struggle is not worth it, certainly not for $1,000. Probably not for $500,000.