News for hackers. Stuff that matters.

A comment on Transcoding Unicode strings at crazy speeds with AVX-512. Parent comment

Re:Transcoding Unicode strings at crazy speeds with AVX-512 by RaisingSpear on Wednesday September 13, 2023 @09:07PM (#37503655)
> wasting up to half the memory

OSes generally implement overcommit - allocating a bunch of memory doesn't mean it's actually being used.

[ Parent | on HN ]
Re:Transcoding Unicode strings at crazy speeds with AVX-512 by amluto on Thursday September 14, 2023 @10:27AM (#37509505)
Sorry, but allocating approximately 2x the needed size for each string is not going to get covered up well by overcommit unless you are directly mmapping and munmapping (or DONTNEEDing or similar) each allocation and each allocation is more than 8kB. (Probably well over 8kB, and for actual good efficiency in the presence of huge pages, over 4MB so there’s a chance that half of it is unused.)

I doubt this workload exists. Overcommit is kind of nifty but not at all magical.

[ Parent | on HN ]
Re:Transcoding Unicode strings at crazy speeds with AVX-512 by RaisingSpear on Thursday September 14, 2023 @11:30PM (#37517981)
> unless you are directly mmapping and munmapping (or DONTNEEDing or similar) each allocation

You don't free up the memory you're not using? Sounds like you're doing it wrong then.

> I bet a really good AVX-512 transcoder with output size checking could be almost as fast as the OP.

Submit your pull request and we can evaluate your bet.

[ Parent | on HN ]