NAMD on Intel XEON EM64T

From: Sterling Paramore (paramore_at_hec.utah.edu)
Date: Mon Nov 14 2005 - 16:55:50 CST

• Next message: Uma Mahankali: "freeEnergy of conformational change of a test molecule"
• Previous message: Raj NK: "munch: The input file main.o is not valid in the current object mode"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]

So I'm trying to get NAMD working on an Intel Xeon EM64T cluster with
myrinet (JVN - http://www.arl.hpc.mil/userservices/lnxi_jvn.html). I've
been able to get it to compile and have successfully run some
preliminary tests. However, about 50% of my 24 hour/64 processor jobs
(the system has 83,161 atoms) will crash and neither NAMD nor the system
report any errors (output just stops, but the job keeps running until it
hits its allotted walltime). I'm thinking that NAMD might be causing
some MPI error somewhere and was looking for suggestions about how I
might change my compiling options. Right now I'm using the 32-bit Intel
8.1 compiler (I had tried using the 64-bit compiler, but my success rate
was even lower with that one). Charm is compiled using the following:

./build charm++ mpi-linux icc gm2 --no-shared -DCMK_OPTIMIZE=1
-I/opt/mpi/x86-intel-8.1/mpich-gm-latest/include
-L/opt/mpi/x86-intel-8.1/mpich-gm-latest/lib -lgm -lmpich -lpmpich -lpthread

and NAMD is compiled with

./config tcl fftw plugins Linux-i686-GM-icc

I see that on the NAMD website, there is a version of NAMD compiled for
this type of machine, but without MPI. Any ideas about how I could fix
this problem?

Thanks,
Sterling Paramore

<http://www.ks.uiuc.edu/Development/Download/download.cgi?UserID=&AccessCode=&ArchiveID=791>

• Next message: Uma Mahankali: "freeEnergy of conformational change of a test molecule"
• Previous message: Raj NK: "munch: The input file main.o is not valid in the current object mode"
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:40:09 CST