Subj : Re: questions re: atomic<>
To   : comp.programming.threads
From : Joe Seigh
Date : Mon Jun 27 2005 12:29 pm


> Those two memory barriers really slow things down.  On a ppc it's
> 102 nsec with them and 4 nsec without them.

That's raw speed.  In a testcase application (lock-free linked list traversal)
I saw up to a 10x slowdown.  That's with a bunch of scheduler artifacts
muddying things up but even so.


-- 
Joe Seigh

When you get lemons, you make lemonade.
When you get hardware, you make software.

.