[concurrency-interest] j.u.c/backport performance on Windows

Peter Kovacs peter.kovacs.1.0rc at gmail.com
Sun Apr 1 09:16:23 EDT 2007


Hi,

Does anyone have experience with java.util.concurrent or with the
backport on Windows?

I am testing the Java 5-optimized version of backport  for a specific
application. I already reported on March 20 that I observed
significant performance degradation compared to an old implementation
which makes use exclusively of "primitive" Java language concurrency
constructs (http://altair.cs.oswego.edu/mailman/private/concurrency-interest/2007-March/003784.html).
In the said mail, I hinted at some possible changes in the
backport-based implementation which I thought might reduce the
performance difference. Implementing these changes I achieved a
comparable performance with the backport as with the older "primitive"
implementation.  This was on linux. (I also remember having observed
fairly good results on HP-UX, though I haven't tested that
extensively.)

Now I started testing on windows (on the same hardware) and the
performance degradation is very disappointing. I tried several
variations of my new backport-based implementation and I invariably
get an almost serial execution! (What makes this situation very
difficult to handle is that I could explain a serial behaviour [with a
slow single consumer], if I didn't have more or less acceptable
results on other platforms or with the old "primitive"
implementation.)

I am pretty much out of my wits...

Can anyone comment, please?

Thanks
Peter


More information about the Concurrency-interest mailing list