no way to compare when less than two revisions

Differences

This shows you the differences between two versions of the page.

@@ Line 1: / Line 1: @@
+\\
+**[[cluster:0|Back]]**
+===== NAMD =====
+Most of your question can be answered on the web site [[http://www.ks.uiuc.edu/Research/namd/| NAMD]] or subscribe to their community supported list ''namd-l''.
+The rest of this page are simple instruction to get you going.
+===== Jobs =====
+The NAMD binary was compiled against the Topspin libraries, hence can only run on the ''imw'' queue (Infiniband switch).  You need to be a member of the "inf_nodes" unix group to successfully run the bianry.
+Here is a sample job.
+<code>
+#!/bin/bash
+#BSUB -q imw
+#BSUB -R "span[ptile=1]"
+#BSUB -J namd
+#BSUB -o out
+#BSUB -e err
+#BSUB -n 4
+MPIRUN="/share/apps/bin/lsf.topspin.wrapper"
+PATH=/usr/local/topspin/mpi/mpich/bin:$PATH
+export PATH
+# scratch dirs
+MYSANSCRATCH=/sanscratch/$LSB_JOBID
+MYLOCALSCRATCH=/localscratch/$LSB_JOBID
+#cd $MYSANSCRATCH
+rm -rf err out
+export LD_LIBRARY_PATH=/usr/local/topspin/mpi/mpich/lib64
+$MPIRUN /share/apps/NAMD/NAMD_2.6_Source/Linux-amd64-MPI/namd2 /home/hmeij/1g6r/named-pgm/alanin
+</code>
+You may browse that directory to view the type of output it generates.
+===== OpenMPI =====
+Axel's commentary on how to get NAMD to run with OpenMPI.
+the way i'm compiling openmpi is the following (with support for infiniband, myrinet, and job launch directly through OpenPBS without host/nodefiles):
+<code>
+./configure --prefix=/cmm/pkg/openmpi-1.2.7 \
+            --enable-orterun-prefix-by-default \
+            --with-gm=/opt/gm-2.1.29 \
+            --with-openib=/opt/ofed-1.3.1 \
+            --with-tm=/opt
+make -j8
+make install
+</code>
+now i can select what communication to use by default through /cmm/pkg/openmpi/etc/openmpi-mca-params.conf , which has either **btl = openib,self**  or **btl = gm,self**
+if i run on the frontend, i override that with **-mca btl tcp,self**
+the very same charm++ megatest binary works fine on all four clusters and the front end without any other modifications.
+one of the nicest things about a (normal) openmpi installation is that the resulting executables only depend on the openmpi libraries and _not_ on any other lowlevel communication libs, so one can actually use the same binary on a different installation of openmpi with different lowlevel protocols, provided it is of the same openmpi version and uses the same gcc/libc and so on.
+\\
+**[[cluster:0|Back]]**