Re[2]: AMBER: mpich problem

From: Qiang Lu <qiangl.uci.edu>
Date: Mon, 10 Nov 2003 16:07:06 -0800

Hi Viktor,

This time it works. I compiled mpich and amber7 on another machine
with g77 & gcc -static, then move the mpich and sander to the cluster.
I can run parallel jobs now. Thanks David, Yong and Ray for the help.

-- 
Best regards,
 Qiang                            mailto:qiangl.uci.edu
=================Original message text===============
Qiang,
Did you try compiling with g77? There may be issues with amber7 (is that 
the version you're referring to?) and mpich compiled with ifc. You lose 
the higher optimiziation when not compiling with ifc, but it'll help to 
troubleshoot the problem. I recall that compiling non-parallel amber7 
with ifc worked fine but there were some issues with mpich compiled version.
Cheers,
-Viktor
Qiang Lu wrote:
> Hi amber,
> 
>    I compiled mpich and sander with ifc on dual cpu xoen(has been set to 4 cpu).
> When I ran benchmark jac program.
> 
> mpirun -np 2 /home/qiangl/bin/sander_ifcmpi -O -i mdin -c inpcrd.equil -p prmtop -o bench.jac.out
> 
> I got the following error message:
> 
> p0_18622:  p4_error: Child process exited while making connection to remote process on syngene34.ics.uci.edu: 0
> qiangl.syngene33 % /home/qiangl/mpich/bin/mpirun: line 1: 18442 Broken pipe
> 
> I did configure the mpirun with ssh i.e./configure -rsh=ssh -cc=$CC
> -fc=$FC, because the rsh is not allowed.
> and I can ssh to syngene34.ics.uci.edu manually with no password. Also
> syngene34.ics.uci.edu has been added to known_hosts.
> 
> Also the machine file has been set as following
> In $MPICH_HOME/share/machines.LINUX,
> syngene33.ics.uci.edu:2
> syngene34.ics.uci.edu:2
> syngene35.ics.uci.edu:2
> syngene36.ics.uci.edu:2
> syngene37.ics.uci.edu:2
> syngene38.ics.uci.edu:2
> ....
> 
> Does any one knows what happening here. What's wrong with mpirun?
> 
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
==============End of original message text===========
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu

Received on Mon Nov 10 2003 - 23:53:01 PST