Win32Shell
John M McIntosh
johnmci at smalltalkconsulting.com
Fri Aug 22 03:14:24 UTC 2003
On Thursday, August 21, 2003, at 10:25 AM, Andreas Raab wrote:
> The VM tries Really Hard (tm) to get millisecond accuracy but there is
> only
> so much it can do. The way to see the spread is something like:
>
> ((1 to: 1000)
> collect:[:i| [(Delay forMilliseconds: 1) wait] timeToRun])
> asBag sortedCounts.
>
> On XP using a 3.5.1 VM this yields:
>
> a SortedCollection(990->1 10->2)
>
> so we're pretty close to 1ms.
Mmm on the mac with 3.4.x which used
pthread_cond_timedwait_relative_np() under os-x we get
a SortedCollection(895->1 99->2 5->3 1->4)
Ah, but with 3.5.x which uses the unix logic for
ioRelinquishProcessorForMicroseconds
aka
aioPoll(realTimeToWait*1000);
that gets
a SortedCollection(872->10 95->11 25->1 4->12 1->2 1->13 1->25 1->32)
which is interesting...
I'm wondering if a new linux/bsd folks can run the benchmark and report
what they see with the latest unix vm.
Expressing surprise that noone noticed a 10ms error in delay times...
--
========================================================================
===
John M. McIntosh <johnmci at smalltalkconsulting.com> 1-800-477-2659
Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com
========================================================================
===
More information about the Squeak-dev
mailing list
|