Charm++ parallel programming system

["Kale, Laxmikant V" <kale AT illinois.edu>]

In some applications, the programmer may want some chares to not migrate.
The reasons can be varied. E.g. In NAMD, it was considered that one class
of chares (the "cells" or "patches") stay put, after an initial (possibly
topology aware) mapping, while the others (say force-computation chares)
are migratable.
This makes the load balancing more predictable. To cater to such
situations, we allow users tpo specify which chares are migratable and
which are not. So, if your purpose is a  general purpose balancer used by
many apps, I suggest you take that flag into consideration. Otherwise, IF
the programmer has not written PUP (serialization) routines for those
chares, the program will not work. Of course, if you want to test your
balancer as a reseearch vehicle, then just select benchmarks thatdo not
have non-migratable chares, and it will work fine.

Hope that helps.

-- 
Laxmikant (Sanjay) Kale         http://charm.cs.uiuc.edu
<http://charm.cs.uiuc.edu/>
Professor, Computer Science     
kale AT illinois.edu
201 N. Goodwin Avenue           Ph:  (217) 244-0094
Urbana, IL  61801-2302          FAX: (217) 265-6582






On 2/19/13 3:45 AM, "François Tessier" 
<francois.tessier AT inria.fr>
 wrote:

>Hello!
>
>I'm writing a load balancer for Charm++ and I would like to have more
>information about the concept of "migratable objects". In fact, I saw
>that every chare has a boolean attribute called "migratable". What does
>it main exactly when a chare is not migratable? In my load balancer, I
>don't consider this properties. Is there a problem to move a
>non-migratable chare?
>
>Thanks for your help!
>
>François
>
>-- 
>___________________
>François TESSIER
>PhD Student at University of Bordeaux
>Inria - Runtime Team
>Tel : 0033.5.24.57.41.52
>francois.tessier AT inria.fr
>
>_______________________________________________
>charm mailing list
>charm AT cs.uiuc.edu
>http://lists.cs.uiuc.edu/mailman/listinfo/charm