zlacker

[parent] [thread] 3 comments
1. ameliu+(OP)[view] [source] 2023-10-14 13:55:13
Isn't this already implemented in QEMU?
replies(3): >>mikepu+v81 >>bonzin+w52 >>imtrin+C52
◧
2. mikepu+v81[view] [source] 2023-10-14 22:33:42
>>ameliu+(OP)
I don't think so. I was quite interested in getting emulated GPUs to work for CI workers to be able to test CUDA stuff without needing actual GPU passthrough, but all I could find in this space were abandoned, early stage hobby projects on GitHub.
◧
3. bonzin+w52[view] [source] 2023-10-15 10:15:05
>>ameliu+(OP)
No, QEMU only supports up to AVX2. Based on the time it took to implement AVX and AVX2, I expect it would be about a month of full time work to implement AVX512F, AVX512BW and AVX512VL (it would not be much simpler to implement only one of them).
◧
4. imtrin+C52[view] [source] 2023-10-15 10:16:45
>>ameliu+(OP)
You mean llvmpipe. You can run OpenCL programs with it.
[go to top]