Type Stability in Julia: Avoiding Performance Pathologies in JIT Compilation (Extended Version) · HackerPapers discussions

Type Stability in Julia: Avoiding Performance Pathologies in JIT Compilation (Extended Version) — mentioned by ChrisRackauckas

Wed, 27 Apr 2022 01:02:08 GMT

Or more directly, the property of type groundedness + multiple dispatch allows for devirtualization of dispatches to get static-like performance in a dynamic language. That was the main design principle, which was a conjecture from Jeff's thesis later proven in a type-theoretic way by Jan Vitek's group. https://arxiv.org/pdf/2109.01950.pdf

Paper and discussion history. A mention is not an endorsement.

Type Stability in Julia: Avoiding Performance Pathologies in JIT Compilation (Extended Version) — mentioned by ChrisRackauckas

Sat, 25 Dec 2021 23:16:42 GMT

That is not a very charitable characterization of the position of multiple dispatch in Julia. It's not something that's optional: multiple dispatch is essential for the performance that Julia is looking to achieve. If you notice where acceleration DSLs tend to have trouble, you'll notice that it's always at the point where you get beyond built-in float primitives and onto object support. For example, Numba's object mode has the caveat that "code compiled in object mode will often run no faster than Python interpreted code, unless the Numba compiler can take advantage of loop-jitting", where…

Paper and discussion history. A mention is not an endorsement.

Type Stability in Julia: Avoiding Performance Pathologies in JIT Compilation (Extended Version) — HN submission

Wed, 08 Dec 2021 19:57:18 GMT

Type Stability in Julia: Avoiding Performance Pathologies in JIT Compilation

123 points; 39 discussion replies on HN. Paper and source history. A submission is not an endorsement.