Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Wow, very cool! Two questions:

1) One of the things that used to be the case with NIFs is that they can interfere with the BEAM's famed pre-emptive scheduling. Any chance that's what's going on here? ie: the NIF solution was fast because it's getting an unfair amount of CPU time, potentially leading to latency issues elsewhere in the app? I believe recently that was changed, though, and the BEAM can pre-empt NIFs?

2) Did you consider/benchmark a port solution before NIF? I think of those as a little safer, but with not as good of performance.



Their SortedMap operation times are quite a lot less than the 1ms recommended yield time for NIFs http://erlang.org/doc/man/erl_nif.html#lengthy_work


Ah, good point! Thanks.




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

Search: