From: Gengbin Zheng (gzheng_at_ks.uiuc.edu)
Date: Sun Dec 21 2003 - 14:11:56 CST
Again you discovered a bug! Not many people were interested in charm
daemon on non-windows system though. :-)
Specificaly there was a bug that preventing passing the whole argument
list into a node program.
Also I found lots of other bugs such as zombie processes, incompatible
directory names ('\' of windows v.s. '/' of unix)
Current Charm++ does not support running a charm program across
different platforms, largely due to different type representation issues
such as different int size or little-endian big-endian issues. The
messages send from one to the other may be interpreted wrongly.
So I don't know if running NAMD on a mixed windows and linux cluster would
As far as I know, long time ago, we do have a version that support
heterogeneous clusters. I don't know if someone in the group is interested
in revive this feature.
On Sun, 21 Dec 2003, Gadi Oron wrote:
> Dear everyone.
> I am continuing in my struggle to make NAMD run on several networked
> computers; both Windows and Linux.
> My idea is to run charmd as a service on all those computers.
> It seems to me that there is a bug in the implementation for Linux
> computers of charmd.
> Any attempt to run:
> charmrun namd2 equil.namd
> after setting CONV_DAEMON results in:
> FATAL ERROR: No simulation config file specified on command line.
> When the daemon is on a Windows computer, there is no such problem.
> Thank you for any info.
> ================= Gadi ORON, PhD -o- Proteologics ======================
> =========== gadi/at/proteologics.com =========== +972 8 9475666 ===========
> Plate voltage too low on demodulator tube
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:37:15 CST