Well, yeah, but…
That is getting a fix for free, literally the definition of leaving money on the table IMO. But oh well.
I have also seen several AMD engineers over at Phoronix, so might be a good idea to raise the issue there as well.
Go for it.
But you can never underestimate a company’s ability to overlook things.
No way am I an AMD apologist but I just think it’s not a high priority for them. The amount of people who are inconvenienced by this bug is probably in the hundreds. Perhaps they’re putting this off for the next large software release or RDNA 2.
1 Like
I don’t doubt it. They’ve got bigger fish to fry. When they’re competitive with nvidia, it would make sense to hammer out the kinks that are an edge case.
Also, RE: companies looking for a fix, that’s what the instinct cards are for. That company sees a cost benefit in using consumer cards, so they’re trying to cheap out.
Hello, Mac user here. I have been trying to virtualize Mac OS on ESXi (KVM passthrough doesn’t work on x58) and it worked just fine with Mac OS Mojave and Catalina. Since I want to be able to access Mac OS on my linux vm, I have been interested in using looking glass to be able to achieve this. I’m not sure how hard it will be, but I suppose it is worth a try…
You have basically two of the major three pieces missing.
- IVSHMEM support for ESXI
- Looking Glass (host) support for MacOS
- Looking Glass (guest) support for Linux (this exists).
#1 is going to be really REALLY hard if not impossible without VMWare doing it themselves. You won’t be able to cut and paste code for IVSHMEM from QEMU, because it is licensed under GPLv2, so you would have to completely re-implement from the specification. Good luck doing that with ESXI. Your time would probably be better served fixing MacOS on KVM.
#2 is more doable, although it would still be a huge amount of work. First, you would have to develop a IVSHMEM driver for MacOS, and secondarily you would have add MacOS support to Looking Glass (host). For the port, you would have to fine an equivalent API in MacOS to DXGI or NVFBC, then it should be relatively simple to port the rest of the code for the host application. For both of these, it may be complicated by apple signing requirements, although I don’t know enough to say if that will end up being a road block.