Re: NAMD on Linux cluster with Myrinet

From: Gengbin Zheng (gzheng_at_ks.uiuc.edu)
Date: Tue Oct 12 2004 - 00:17:53 CDT

Things became tricky when GM version 2 came. GM2 has multithread support
that uses pthread in its implementation, which unfortunately does not
work smoothly with Charm++. I think there is some way to disable
pthreads in GM compilation (but I don't know how) to workaround this.

Otherwise, as most users have found out, they have good luck with "gm2"
charm++ build option I created specially for GM2 that uses pthread as
well internally for Charm++.

Unfortunately, as I found out, the "gm2" option may not work all the
times on all platforms, which is annoying. And using pthreads in
Charm++/NAMD can be a little bit slower in terms of the thread context
switching, and also suffer from the fact that only around 256 pthreads
can be created on a Linux box, which means you can not simulate more
than 256 home patches on one processor (but this should not be a problem
for most simulations I guess)

Gengbin

Brian Bennion wrote:

>Ah now we are getting somewhere, can you do a stack trace on the core
>file that was dumped? Which I guess means that you compiled with debug
>symbols right... sorry should have asked for you to do that first.
>
>Gengbin, do you have any suggestions?
>
>Brian
>
>
>On Mon, 11 Oct 2004, Alberto Santana wrote:
>
>
>
>>Hello Brian,
>>I did run the megatest and this is what I got
>>
>>node2:.../megatest> charmrun +p4 ./pgm
>>Charm++: scheduler running in netpoll mode.
>>test 0: initiated [groupring (milind)]
>>[0] Assertion "CkMyPe()!=0" failed in file init.C line 389.
>>------------- Processor 0 Exiting: Called CmiAbort ------------
>>Reason:
>>[1] Assertion "getMsgtype()==3 || getMsgtype()==5 || getMsgtype()==13 ||
>>getMsgtype()==14" failed in file envelope.h line 209.
>>------------- Processor 1 Exiting: Called CmiAbort ------------
>>Reason:
>>req_handle_abort called
>>Fatal error on PE 1>
>>
>>I need to figure this out. I don't like the idea of having a Myrinet that
>>NAMD can't use
>>
>>Alberto
>>--
>> Best wishes,
>> Alberto
>>+-------------------------------------------------------------------+
>>Alberto Santana, PhD
>>Department of Chemistry
>>University of Puerto Rico - Mayaguez
>>P.O. Box 9019
>>Mayaguez PR 00681-9019
>>Phone: (787) 832-4040 x3760
>>http://www.uprm.edu/~asantana
>>+-------------------------------------------------------------------+
>>Powered by SuSE Linux 9.1 (http://www.suse.com)
>>
>>
>>
>
>*****************************************************************
>**Brian Bennion, Ph.D. **
>**Computational and Systems Biology Division **
>**Biology and Biotechnology Research Program **
>**Lawrence Livermore National Laboratory **
>**P.O. Box 808, L-448 bennion1_at_llnl.gov **
>**7000 East Avenue phone: (925) 422-5722 **
>**Livermore, CA 94550 fax: (925) 424-6605 **
>*****************************************************************
>
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:37:54 CST