Print

Print


Hi Sadat,

I used Mason to model a football(soccer)-stadium with 13000 agents and
interestingly it made no difference for the runtime if I did the simulation
with or without the gui. I didn't used it, but maybe you should read the
sections of the manual (p.145 ff) where checkpoints are described. Maybe
that is what you want. You could set a checkpoint at those milestones. save
the simulation, run that in a gui environment for some steps till another
checkpoint and than run again without the gui. At least what I understand
from that section of the manual, it should be possible.

If time isn't the limiting factor it is only important that a user can
interact at a certain point, you can also run the entire simulation in the
gui, you can set a step where the simulation has to pause so you won't miss
a certain point.

On a unrelated sidenode:
2 things I discovered with my project:

- if I run my mason model in parrallel without a gui, it leads to states
where some agents where in a live- or deadlock (I hadn't the time to find
out exactly what happened there). When I run the same simulations in a loop
(sequential) then those live-/deadlock didn't occur.
- I tried to capture a video (10fps) of my simulation (one run needs about
2 1/2 hours) with the build in tool but the video output wasn't properly
saved. There was a tmp-output file but it seemed that this wasn't written
back into the original file I chosed in the capture-dialog.


greetings,
Norman


2013/11/6 Sadat Chowdhury <[log in to unmask]>

> Hi all,
>
> Perhaps this is something I have to figure out myself, as it is very
> specific to my application, but I'm still asking this to the MASON
> community at large -- maybe someone has had a similar situation before.
>
> I am simulating an evolutionary system that has a long run-time. If I run
> the simulation without the GUI, it runs very fast, and I encounter
> "interesting" events fairly quickly. If I run the simulation with a GUI, I
> can visualize the process, but the execution is slowed down significantly
> -- and I would have to let the GUI simulator run for a very long time to
> get to those events that I am interested in observing.
>
> I have a way of detecting when I hit certain milestones. What I would like
> to do is, "freeze" the GUI part (viewer part in the MVC framework) until I
> hit those milestones, let the viewer act on the model at that point, for
> certain number of steps, and then disconnect the viewer until I hit the
> next milestone. The purpose is to only wake up the viewer when it is
> signaled -- otherwise run the main simulation in a detached mode.
>
> Are there any particular API calls, examples, patterns I should follow?
>
> Thanks,
> Sadat.
>