Charm++ parallel programming system

[Abhinav Bhatele <bhatele AT illinoisalumni.org>]

So, if we can get mpi-smp to work with optimization flags, it might be faster. What problems are you seeing when mpi-smp is built with --with-production?

- Abhinav

On Thu, Sep 8, 2011 at 3:30 PM, Bennion, Brian <Bennion1 AT llnl.gov> wrote:

I neglect to state that all were built with --with-production except the mpi-smp which would only compile with debug symbols included.

-----Original Message-----
From: bhatele AT gmail.com [mailto:bhatele AT gmail.com] On Behalf Of Abhinav Bhatele
Sent: Thursday, September 08, 2011 3:25 PM
To: Bennion, Brian
Subject: Re: [ppl] [charm] current results on LLNL cluster

Brian,

Just a minor note. When you are decided on which version of charm to use for your production runs, don't forget to build charm with --with-production. That will kick off all the optimization flags for a given platform.

- Abhinav



On Thu, Sep 8, 2011 at 10:21 AM, Bennion, Brian <Bennion1 AT llnl.gov> wrote:


       Hello,

       Compilation of all the suggested versions of charm++ have been completed.
       They were:
       mpi-linux-x86_64
       mpi-linux-x86_64 smp
       net-linux-x86_64 ivberbs
       net-linux-x86_64 ibverbs smp

       It appears that the first charmm build is the best for namd2.8.  I saw a 25% speedup on my regular jobs using the same processor counts.  No extra tricks were employed in compiling charm++ or namd2.8.

       The ibverbs version did compile for charm++ and namd2.8 and was even able to start.  however, sockets closed prematurely on some node and the whole things burns to the ground.

       Two quick thoughts here.  First, looking at the charm.XXXXX files that are created, they look for some /tmp/charm.XXXX directories that are not going to be available.  The tmp filesystem goes by username it /tmp/bennion1/XXXXX
       Second, I will check with our admins to see if using ib layer is even permissible.

       Brian

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





--
Abhinav Bhatele
Postdoctoral Research Staff Member
Center for Applied Scientific Computing
Lawrence Livermore National Laboratory

-- 
Abhinav Bhatele
Postdoctoral Research Staff Member
Center for Applied Scientific Computing
Lawrence Livermore National Laboratory