Skip to Content.
Sympa Menu

charm - Re: [charm] Charm++ SMP runtime problem

charm AT lists.cs.illinois.edu

Subject: Charm++ parallel programming system

List archive

Re: [charm] Charm++ SMP runtime problem


Chronological Thread 
  • From: Marius Micluta <marius AT biochim.ro>
  • To: Filippo Gioachin <gioachin AT uiuc.edu>
  • Cc: charm AT cs.uiuc.edu
  • Subject: Re: [charm] Charm++ SMP runtime problem
  • Date: Tue, 2 Jun 2009 15:44:08 +0300 (EEST)
  • List-archive: <http://lists.cs.uiuc.edu/pipermail/charm>
  • List-id: CHARM parallel programming system <charm.cs.uiuc.edu>


Hi Filippo,

Thank you for clarifying the fact that an SMP build of charm++ is incompatible with charmdebug. I also tested the offending command with the precompiled non-SMP binary before posting my previous message and it didn't crash, but I did not run further a debug session, as my concern is with the SMP version.

however, if charmdebug is not appropriate for my situation, how could I find and eliminate the cause of the freeze at startup, which occurs from time to time when running the net-linux-x86_64-smp-icc version of charm++/NAMD? The NAMD documentation states that "Additional performance may be gained by building NAMD against an SMP version of Charm++ such as net-linux-smp or net-linux-smp-icc. This will use a communication thread for each process to respond to network activity more rapidly." I also found that the SMP build is significantly faster than the net-linux-x86_64-icc version. The problem is that sometimes it just fails to start properly. On the other hand, when the startup succeeds, it runs fine and exits gracefully.

The NAMD documentation is somewhat obscure to me when referring to clusters of SMP machines. Besides, it addresses older hardware - single-core dual-processor and single hyperthreaded processor architectures. Should I run fewer processes/threads per node than the number of available CPU cores?


Sincerely,

Marius




Archive powered by MHonArc 2.6.16.

Top of Page