Mike, most of our work lately has been in either asynchronous island
models or (more commonly) in the master/slave evaluator, where we've
put most of our effort. I'd be glad to look into it but I need a
minimal example that we can run here and test.
I will mention that if one of the nodes is stuck for some reason --
some large individual is taking forever to be processed, for example,
a synchronized island model will sit there waiting for it.
On Jun 9, 2006, at 12:28 PM, Michael Solano wrote:
> Hello Everyone,
> I'm having a bit of trouble using the IslandExchanger to perform
> synchronous Island Model experiments in ECJ version 15. It seems to
> work for a while but at some point one of the nodes will output
> "Waiting for synchronization...." and everything seems to stall
> (ecj is still running but none of the nodes are doing anything).
> The current setup I'm using is 3 islands (on 3 separate computers)
> with subpop sizes of 333 individuals each with migration of 10
> individuals every 5 generations. Does anyone know what might be
> causing this? Is there something I may be doing incorrectly? Thanks :)