From: Giovanni Bellesia (gbellesia_at_chem.ucsb.edu)
Date: Mon May 22 2006 - 14:31:14 CDT
With the help of Jim and the development team a solution was found.
To cut a long story short, if you experience problems implementing Tcl
forces or Tcl boundary forces scripts (random crashes with floating
point errors) when running namd in parallel mode, then you should use
the +netpoll command option on the command line or use the binary
versions compiled with TCP.
> Hi all,
> I'm still struggling with a TCLboundaryforces script taht works fine
> in standalone mode and randomly crashes if I run my simulation on
> many processors.
> Is there someone which is working with TCL boundary forces having no
> problems in parallel simulations ?
> Also could someone explain to me this paragraph from the NAMD UG:
> .. page 75
> The script provided in tclBCScript and the calcforces procedure it
> defines are executed in multiple Tcl interpreters, one for every
> processor that owns patches. These tclBC interpreters do not share
> state with the Tcl interpreter used for tclForces or config file
> parsing. The calcforces procedure is passed as arguments the current
> timestep, a �~@~\unique�~@~] flag which is non-zero for exactly one Tcl
> interpreter in the simulation (that on the processor of patch zero),
> and any arguments provided to the most recent tclBCArgs option. The
> �~@~\unique�~@~] flag is useful to limit printing of messages, since the
> command is invoked on multiple processors.
> I'm not sure I understood the multiple tcl interpreters 'thing' and
> also how the 'unique' flag works.
> Thanks for your help
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:42:04 CST