Charm++ parallel programming system

[Phil Miller <mille121 AT illinois.edu>]

Hi Hrushit,

You've understood correctly - execution of Charm++ application code is entirely non-preemptive.

If you have code in which you strongly wish to cooperatively schedule work, an object's method can send itself another message and return to the scheduler, or (if using the [threaded] attribute) yield to the scheduler by calling CthYield().

In general, though, we lean more towards tuning programs so that the grain size (the time any one method takes to execute) is not so long as to demand pre-emption for other work. Note that Charm++ execution in general is driven by the events of message delivery. Messages can be given absolute priority by sending them to methods marked with the [expedited] attribute, or they can be given relative priority as described in the manual:

http://charm.cs.illinois.edu/manuals/html/charm++/10.html#SECTION02130000000000000000

Please feel free to post with any further questions you have.

Good luck and happy coding,

Phil

On Sat, Jun 21, 2014 at 4:42 PM, Hrushit Parikh <parikhhrushit AT hotmail.com> wrote:

Hey,
I have recently started using charm and from the few papers I'v read about the charm language and run time system Iv noticed a mention that a processor(chare) executing a message cannot be interrupted or preemted. Is this correct? Is there any way of interrupt or even driven messaging in charm? Or can a particular message be given the highest priority of execution after the current message finishes execution?
Which papers/manuals could I refer to for more information about the messaging infrastructure in charm?

Thank you,
Hrushit 

_______________________________________________
charm mailing list
charm AT cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/charm