Question: Why aren't we using LLMs to translate programs to use ROCm? Isn't tran... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		2OEH8eoCRo0 on Feb 12, 2024 \| parent \| context \| favorite \| on: AMD funded a drop-in CUDA implementation built on ... Question: Why aren't we using LLMs to translate programs to use ROCm? Isn't translation one of the strengths of LLMs?

JonChesterfield on Feb 12, 2024 [–]

You can translate cuda to hip using a regex. LLM is rather overkill.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact