Which reminds me of another thing I noticed--MPEnter/
ExitCriticalRegion
seemed to be a lot slower than the raw pthread solution, according to
Shark. Why is this? Seems like one should be just as good as the
other.
Interesting. I would have thought they'd be close too.
MPEnterCriticalRegion() et al handle some situations which I don't
believe the underlying raw pthreads stuff doesn't -- deletion of a
critical region on thread A while thread B is attempting to enter or
exit it.
The system also attempts to assign opaque ids in a way which makes it
very unlikely that any given id will get re-used shortly later, in an
attempt to prevent errors if a particular id or memory block gets
reused shortly after being released.
But, there are things which could be done to improve the performance
of the MP utilities. They didn't make it into the next expected
release, and we'd like to get to them, but they need to be weighed
against many other improvements in the system as well. Among others,
these are written up in <rdar://problem/3465075> and
<rdar://problem/3465078>.
-Keith
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Carbon-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/carbon-dev/email@hidden
