Re:Run CUDA, unmodified, on AMD GPUs
by almostgotcaught on Tuesday July 16, 2024 @08:37AM
(#40975960)
You're wrong. The people building the models don't write CUDA kernels. The people optimizing the models write CUDA kernels. And you don't need to know a bunch of ML bs to optimize kernels. Source: I optimize GPU kernels. I don't make 500k but I'm not that far from.
[ Parent
| on HN ]
Re:Run CUDA, unmodified, on AMD GPUs
by HarHarVeryFunny on Tuesday July 16, 2024 @10:58AM
(#40977230)
How much performance difference is there between writing a kernel in a high level language/framework like PyTorch (torch.compile) or Triton, and hand optimizing? Are you writing kernels in PTX?
What's your opinion on the future of writing optimized GPU code/kernels - how long before compilers are as good or better than (most) humans writing hand-optimized PTX?
Re:Run CUDA, unmodified, on AMD GPUs
by throwaway81523 on Tuesday July 16, 2024 @04:54PM
(#40980251)
The CUDA version of LCZero was around 2x or 3x faster than the Tensorflow(?) version iirc.
Re:Run CUDA, unmodified, on AMD GPUs
by throwaway81523 on Tuesday July 16, 2024 @08:56AM
(#40976079)
Heh I'm in the wrong business then. Interesting. Used to be that game programmers spent lots of time optimizing non-ML CUDA code. They didn't make anything like 500k at that time. I wonder what the ML industry has done to game development, or for that matter to scientific programming. Wow.