Post #3810976
2026-04-30 19:41 UTC
@cfallin@mastodon.world Thanks for sharing! How much of a performance gap do you expect between weval at its ceiling vs. optimized JS running in its "native environment" at higher optimization tiers — and do you think the remaining gap is mostly the residual type guards, or something deeper in the approach?
Replies (1)
-
@cfallin@mastodon.world 2026-05-01 20:17
@ttraenkler@mastodon.online It's still about 10x, and is pretty fundamental to the fact that we don't have types, still, so we have a bunch of IC callsites. With full type info (somehow) there's no reason that this approach couldn't get all the same opts that e.g. IonMonkey gets -- it "just" takes a lot of engineering!