User: Password:
|
|
Subscribe / Log in / New account

Some ado about zero

Some ado about zero

Posted Jul 9, 2009 3:45 UTC (Thu) by neilbrown (subscriber, #359)
Parent article: Some ado about zero

I'm curious as to why the "per-CPU zero pages" approach wasn't pursued. It should preserve most of the benefits of a single zero page, while significantly reducing the contention on the refcount.... or would it just give us the worst of both worlds?


(Log in to post comments)

Some ado about zero

Posted Jul 9, 2009 10:56 UTC (Thu) by johill (subscriber, #25196) [Link]

Well imagine having 4096 zero pages. I guess you could better get away with just a percpu refcounter. But then again, why do you need to refcount the zero page anyway? It's always there to start with, and never writable, isn't it?

Some ado about zero

Posted Jul 9, 2009 12:35 UTC (Thu) by Cyberax (✭ supporter ✭, #52523) [Link]

That's just 16MB. That's close to 'nothing' for a computer with 4096 CPUs.

Some ado about zero

Posted Jul 9, 2009 14:51 UTC (Thu) by ejr (subscriber, #51652) [Link]

There are single chips with dozens of CPUs from the OS's view. Soon these will be in the hundreds. And they'll end up being used in smaller devices relatively soon, simply because of economies of scale. The extra space can add up pretty quickly.


Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds