Comment by jaen

What? All of this has been solved for a long time. How do you think hyperscalers do this?

Search keyword: "Adaptive sampling"