On Mar 8, 2012, at 2:51 PM, Brian Lambert wrote:
> I should have said, "relatively". As in, good enough for many scenarios
> where it will be used infrequently and keeping the code simple makes sense.
>
> Of course, you're right, Charles.
>
> 1,000,000 integer increments synchronized by @synchronized, OSSpinLock, and
> NSLock:
> PerfTimer[95716:403] 1,000,000 @synchronized ++value: [118 ms] [117,813,038
> ns]
> PerfTimer[95716:403] 1,000,000 OSSpinLock ++value: [8 ms] [8,444,567 ns]
> PerfTimer[95716:403] 1,000,000 NSLock ++value: [63 ms] [63,049,168 ns]
>
> A million operations in ~100 ms is fast. A million operations in ~10 ms is
> WICKED fast.
>
> Brian
Another interesting option is mentioned in Apple’s GCD man pages; dispatch_sync
can actually be used as a locking mechanism, like so:
__block NSUInteger value = 0;
dispatch_sync(myLockQueue, ^{
value++;
});
This results in something almost as simple as @synchronized, and quite a lot
faster. Here are the results I get when running it on my MBP. All my results do
take about twice as long as yours; I suspect you have a rather nice machine. :-)
2012-03-08 15:14:56.408 Lock Benchmarks[57369:403] @synchronized value++:
259.156942 ms
2012-03-08 15:14:56.429 Lock Benchmarks[57369:403] OSSpinLock value++:
19.486010 ms
2012-03-08 15:14:56.517 Lock Benchmarks[57369:403] pthread_mutex_t value++:
86.329997 ms
2012-03-08 15:14:56.645 Lock Benchmarks[57369:403] NSLock value++: 127.897978 ms
2012-03-08 15:14:56.692 Lock Benchmarks[57369:403] dispatch_sync value++:
45.979023 ms
Not bad; OSSpinLock is the only thing that seems to be able to best it.
Charles
_______________________________________________
Cocoa-dev mailing list (Cocoa-dev@lists.apple.com)
Please do not post admin requests or moderator comments to the list.
Contact the moderators at cocoa-dev-admins(at)lists.apple.com
Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com
This email sent to arch...@mail-archive.com
