Just wait until someone trains an ML model that can translate any CUDA code into...

		amelius on July 16, 2024 \| parent \| context \| favorite \| on: Run CUDA, unmodified, on AMD GPUs Just wait until someone trains an ML model that can translate any CUDA code into something more portable like HIP. GP says it is just some #ifdefs in most cases, so an LLM should be able to do it, right?

OpenAI Triton? Pytorch 2.0 already uses it.