Charm++ parallel programming system

[Hrushit Parikh <parikhhrushit AT hotmail.com>]

Thank you, Phil. That is what I was looking for!

---

From: mille121 AT illinois.edu
Date: Sun, 8 Mar 2015 17:47:49 -0500
Subject: Re: [charm] Updating a chare objects execution time in LB Database
To: parikhhrushit AT hotmail.com
CC: charm AT cs.uiuc.edu

In charm/src/ck-core/cklocation.C there's a method CkLocRec_local::invokeEntry that calls startTiming() and stopTiming(). You can trace the mechanism involved in the call stack down from those.

On Sun, Mar 8, 2015 at 5:13 PM, Hrushit Parikh <parikhhrushit AT hotmail.com> wrote:

Hi,
I am trying to determine the variation in execution time of chare objects in each iteration of iterative benchmarks. I do not want to do this from the application using getObjTime() because I will have to manually insert this function call at relevant locations in every benchmark I test.

Instead charm++ Load Balancer maintains a LBDatabase where the execution time of every object is updated once it runs and the load balancer uses this information from the database to identify overloaded/underloaded processors   
(principle of persistence etc).

So I am trying to figure out where in the charm++ stack is this per chare load information updated once the chare runs. Does it happen in converse layer once a chare from the processors queue is processed?
I was unable to find this in the charm++ layer. I had a look at ck-ldb/lbdb.h and related files which have the struct LBObjData .

Can someone please point me to the right charm++ layer/files?

Thank you very much.

Hrushit

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