Hacker News new | past | comments | ask | show | jobs | submit login

> Even software that is directly ported has trouble because M1 tile memory cache is designed for minimal tasks, not heavy computing, and computationally heavy software has to be re-written from scratch to optimize it for the platform.

Huh? I have no idea what this is supposed to mean. Tile based memory refers to the GPU, and you don't have to rewrite your GPU software from scratch, that's insane? That's what Metal is for. Whatever this is supposed to mean I'm pretty sure it's not even wrong




I think they mean that GPU applications that are not optimized for TBDR may experience reduced performance.


Even then, “M1 tile memory cache is designed for minimal tasks, not heavy computing” is such a stupid thing to write…




Join us for AI Startup School this June 16-17 in San Francisco!

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

Search: