Small parameters you say... I wonder if a recent change in Java has significantly worsened its behavior with respect to denormalized floats [google for it].

BTW, MASON has the same CERN library available in sim/util/distribution/ though I don't see that method; it may be newer.

Sean

> On Mar 6, 2021, at 3:00 AM, Siamak Sarmady <[log in to unmask]> wrote:
> 
> Hello again,
> 
> I was forced to use a profiler to see what's wrong and it appears that a method in an external random number generator library (CERN Mersenne Twister library) was the culprit.
> 
> The Poisson random number function would fall into a long loop for small Lambda parameters. I had modified the lambda to a small number in my simulation data files (not in the code).
> 
> Removed the function and used my own simple code and I now got a wooping 100x simulation speed....
> 
> Sorry for taking your time guys
> 
> <image.png>
> 
> On Fri, 5 Mar 2021 at 15:18, Robert Zupko <[log in to unmask]> wrote:
> Mac:
> 
> I’ve seen that before when a new Java patch has been installed. We’re there any updates to the system? 
> 
> Regards,
> Robert Zupko
> 
> > On Mar 5, 2021, at 09:46, Mac Sarmady <[log in to unmask]> wrote:
> > 
> > Hello
> > 
> > I am using Java 11 to run mason simulations. A few weeks ago my simulations were reasonably fast (crowd simulation with a few hundreds agents). 
> > 
> > Then I had a pause and when I ran the simulations a few weeks later they are extremely slow (perhaps between 4 to 10 times?)
> > 
> > - Has anyone encountered such a thing on his/her JVM and simulations?
> > 
> > - Is there a benchmark on any of the mason examples that I can test and see whether the problem is from the installation or something inside my own code (I don't remember changing it in the period)?
> > 
> > Regards,
> > Mac