Based on something Hans mentioned, something approaching:
cputrack -c PAPI_l1_icm,PAPI_l2_ich,PAPI_l2_icm <your application>
Could be useful in comparison to a system on which this doesn't occur
(a synthetic test case would be better, I'm not sure these counters
would be so useful under the real workload).
Comparison with a different machine (a lowly 4-core Opteron with
Solaris 10) will need to wait until this evening.
I am astonished at how long it takes to spin-up 64 lwps. With
cputrack reporting a total run-time of 2.500, lwp #64 is created at
0.708, there is a short flurry of activity, a long pause, and then lwp
#65 is created at 1.433 after which there is a short flurry of
activity followed by a long pause before the program quits.
Presumably all of the program's actual threaded work is done in this
last short flurry of activity (until 2.005) which represents a small
portion of the total run-time. The core work seems to be done in just
over 1/2 a second even though the program runs for 2.5 seconds. The
program seems to be heavily subject to Amdal's law. Perhaps the
protracted locking is responsible for some of these long pauses.
Not all the algorithms work the same, but it seems to take somewhere
between 1 and 1.4 seconds before any actual work starts being done.
Bob
--
Bob Friesenhahn
bfrie...@simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer, http://www.GraphicsMagick.org/
-------------------------------------------
illumos-discuss
Archives: https://www.listbox.com/member/archive/182180/=now
RSS Feed: https://www.listbox.com/member/archive/rss/182180/21175430-2e6923be
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21175430&id_secret=21175430-6a77cda4
Powered by Listbox: http://www.listbox.com
