That's probably gonna run in gpu anyway. There is a lot of data work that involves double (or single) precision arithmetic, outside of ML, LLMs etc. It is defin

That's probably gonna run in gpu anyway. There is a lot of data work that involves double (or single) precision arithmetic, outside of ML, LLMs etc. It is definitely also worth testing 32 bit floats too, though.