Re: Dynamic schedules and execution speed

[Roger M. Burkhart]

Rob Kewley writes:

> I was able to find a workaround to the problem with the dynamic schedule
> (through trial and error, aka "Brute force and ignorance!).  My model
> uses three schedules - two repeating and one dynamic.  When I created a
> separate sub-swarm for only the dynamic schedule, activated the sub
> swarm in the model swarm, activated the dynamic schedule in the sub
> swarm, and activated the repeating schedules in the model swarm, the
> entire simulation ran and iterated without a slow down.  However, I
> still believe there is a bug in the dynamic schedule or swarm schedule
> merging machinery that causes the slow down.  I tried to find it but got
> lost in the source code.

Thanks very much for your problem reports and the sample template code
you provided!  Using that starting point, I was able to track down the
schedule merging bug I've just summarized in a separate message.  The
problem was in fact more general than the one case you found (a dynamic
schedule being merged with a repeating schedule in the model swarm), but a
good reproducible test case is the indispensable starting point to find a
problem.

I've been wondering if the right structure of subswarms could somehow avoid
the merging bug I've discovered, and from your case it evidently can.  As
best as I understand right now, it must create a leapfrog effect in which
the repeated concurrent actions are avoided, at least in part.  But I'll
do some more checking to see just how that works.  In the meantime, glad
to hear you were able to find a way to avoid the problem!  If you can test
the sample bug fix, please let me know if it also fixes everything.  You
were absolutely right about there being a problem, and your extremely
dynamic schedule example in the model swarm was the best way to force it
to the surface.  Thanks again, and I hope things will get less in your way
in the future.

--Roger


                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.
                  ==================================