4 ms·

Didn't AMD just commit to a CUDA compatible api? Did I hallucinate that?

by HWR_14 3y ago

tomoyoirl 3y ago
If you mean the one posted here earlier today, I believe that article was more like “we paid a contractor to implement this, and then decided not to use it, so per our terms it’s open source now.”
lhl 3y ago
Well, you probably read a inaccurate headline about it. The project is called ZLUDA https://github.com/vosen/ZLUDA https://github.com/vosen/ZLUDA and it had a recent public update because of the opposite - AMD decide not to continue sponsoring work on it: > Shortly thereafter I got in contact with AMD and in early 2022 I have left Intel and signed a ZLUDA development contract with AMD. Once again I was asked for a far-reaching discretion: not to advertise the fact that AMD is evaluating ZLUDA and definitely not to make any commits to the public ZLUDA repo. After two years of development and some deliberation, AMD decided that there is no business case for running CUDA applications on AMD GPUs. > > One of the terms of my contract with AMD was that if AMD did not find it fit for further development, I could release it. Which brings us to today. It's worth noting that while ZLUDA is a very cool project, it's probably not so relevant for ML. Also from the README: > PyTorch received very little testing. ZLUDA's coverage of cuDNN APIs is very minimal (just enough to run ResNet-50) and realistically you won't get much running. > However if you are interested in trying it out you need to build it from sources with the settings below. Default PyTorch does not ship PTX and uses bundled NCCL which also builds without PTX: PyTorch has OOTB ROCm support btw and while there are some CUDA-only libraries I'd like (FA2 for RDNA, bitsandbytes, ctranslate2, FlashInfer among others), I think sponsoring direct porting/upstreaming compatibility of the libraries probably makes more sense. Also from the ZLUDA README: > ZLUDA offers limited support for performance libraries (cuDNN, cuBLAS, cuSPARSE, cuFFT, OptiX, NCCL).
zozbot234 3y ago
Different kinds of compatibility. HIP is source compatible and officially supported. Zluda is the newly released project for running CUDA-compiled binaries.

Didn't AMD just commit to a CUDA compatible api? Did I hallucinate that?

by HWR_14 3y ago

Didn't AMD just commit to a CUDA compatible api? Did I hallucinate that?

If you mean the one posted here earlier today, I believe that article was more like “we paid a contractor to implement this, and then decided not to use it, so per our terms it’s open source now.”

lhl 3y ago

Well, you probably read a inaccurate headline about it. The project is called ZLUDA https://github.com/vosen/ZLUDA https://github.com/vosen/ZLUDA and it had a recent public update because of the opposite - AMD decide not to continue sponsoring work on it: > Shortly thereafter I got in contact with AMD and in early 2022 I have left Intel and signed a ZLUDA development contract with AMD. Once again I was asked for a far-reaching discretion: not to advertise the fact that AMD is evaluating ZLUDA and definitely not to make any commits to the public ZLUDA repo. After two years of development and some deliberation, AMD decided that there is no business case for running CUDA applications on AMD GPUs. > > One of the terms of my contract with AMD was that if AMD did not find it fit for further development, I could release it. Which brings us to today. It's worth noting that while ZLUDA is a very cool project, it's probably not so relevant for ML. Also from the README: > PyTorch received very little testing. ZLUDA's coverage of cuDNN APIs is very minimal (just enough to run ResNet-50) and realistically you won't get much running. > However if you are interested in trying it out you need to build it from sources with the settings below. Default PyTorch does not ship PTX and uses bundled NCCL which also builds without PTX: PyTorch has OOTB ROCm support btw and while there are some CUDA-only libraries I'd like (FA2 for RDNA, bitsandbytes, ctranslate2, FlashInfer among others), I think sponsoring direct porting/upstreaming compatibility of the libraries probably makes more sense. Also from the ZLUDA README: > ZLUDA offers limited support for performance libraries (cuDNN, cuBLAS, cuSPARSE, cuFFT, OptiX, NCCL).

zozbot234 3y ago

Different kinds of compatibility. HIP is source compatible and officially supported. Zluda is the newly released project for running CUDA-compiled binaries.