Hacker News Reader

In graphics, we also often try to avoid any kind of rejection sampling because it wrecks importance sampling.

Hmm, interesting. Is this still the case if the rejection factor is known explicitly (as it would be here)?