in order to make it even more complex ... also different multithreaded sized clients would be ... great! :)
Please let me know if there is something I can help with.
On Oct 11, 2013, at 22:30, Sean Luke <[log in to unmask]> wrote:
> Ralf, I think it's possible but I have to think about it, because in many cases the job size roughly translates into the "population size" on the slave side, which may complicate things.
> On Oct 11, 2013, at 6:34 AM, Ralf Buschermöhle wrote:
>> currently a multithreaded slave waits until the complete eval.masterproblem.job-size (the master "offers") is processed.
>> Can this behaviour be changed to something where multithreaded slave processes individuals as soon as possible e.g., if the job-queue is still filled?
>> AFAIK the motivation for the current approach is the reduce the required network bandwidth between master/slaves. My motivation is to reduce the memory consumption of each slave (because of a large DB each slave need to "carry").