Differences

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

--- md_software [2016/02/25 18:47]
sjabbari [LAMMPS]
+++ md_software [2016/03/21 18:04]
milchev [HOOMD]
@@ Line 22: / Line 22: @@
     * URL(s): [[http://lammps.sandia.gov/|lammps]]
     * Docs:
-    * Example call(s):
+    * Example call(s): To run lammps, you need to provide an input script for example "in.script" and also a a data file that your input script needs to read to start the simulation. In the following example, "16" is the number of cores requested.
-#
-     # To run lammps, you need to provide an input script for example "in.script" and also a a data file that your input script needs to read to start the simulation. In the following example, "16" is the number of cores requested.
-     mpirun  -n 16  -envall lmp_mpi < in.script
   <code bash>
+  mpirun  -n 16  -envall lmp_mpi < in.script
 </code>
     * LSF considerations (e.g. ''bsub'' example): an example of a script for job submission
+     <code bash>
-    *''#!/bin/bash
+#!/bin/bash''
 #BSUB -J lammps-test      #jobname
 #BSUB -o test.log
 #BSUB -e  test.err
-##BSUB -q short
 #BSUB -q nodeshort
 #BSUB -W 300      #runtime in min
 #BSUB -n  64
+mpirun  -n 64  -envall lmp_mpi <in.fene1000x10
+</code>
-# Store working directory to be safe
-SAVEDPWD=`pwd`
-# We define a bash function to do the cleaning when the signal is caught
-cp ${SAVEDPWD}/in.fene1000x10  /jobdir/${LSB_JOBID}/
-cp ${SAVEDPWD}/data.file /jobdir/${LSB_JOBID}/
- cd /jobdir/${LSB_JOBID}
-mpirun  -n 64  -envall lmp_mpi <in.fene1000x10
-mv -f  *.* ${SAVEDPWD} ''
 you can submit the job as ''bsub < lammps_job.sh ''
-    * Please download all the required modules, before starting to run lammps.
+    * Please load all the required modules, before starting to run lammps.
     * contact: ''sjabbari'' (Institut für Physik)
 ==== HOOMD ====
+HOOMD-blue is a general-purpose particle simulation toolkit. It scales from a single CPU core to thousands of GPUs. On a single NVIDIA GPU, HOOMD-blue performs an order of magnitude faster than a multi-core CPU in typical benchmarks.
     * Modules:
+    * gcc/4.6.2 cuda/gcc-4.6.2/6.5.14
+    * mpi/openmpi/1.8.7/gcc_4.4.7-cuda_6.5.14
           * ''software/hoomd/1.0.5/gcc-4.6.2_cuda-6.5.14/nompi''
           * ''software/hoomd/1.0.5/gcc-4.6.2_cuda-6.5.14/openmpi-1.8.7''
-          * ''software/hoomd/1.3.2/gcc-4.6.2_cuda-6.5.14/nompi''
+          * ''software/hoomd/1.3.3/gcc-4.6.2_cuda-6.5.14/nompi''
-    * Description:
+    * Description:                                                                                     HOOMD-blue is a general-purpose particle simulation toolkit. It scales from a single CPU core to thousands of GPUs.
+You define particle initial conditions and interactions in a high-level python script. Then tell HOOMD-blue how you want to execute the job and it takes care of the rest. Python job scripts give you unlimited flexibility to create custom initialization routines, control simulation parameters, and perform in situ analysis.
     * Binaries: ''hoomd'' and ''hoomd-config.sh'' plus those in the module's ''tools'' directory.
     * URL(s): [[https://codeblue.umich.edu/hoomd-blue/doc/index.html|HOOMD]]
     * Docs: [[https://codeblue.umich.edu/hoomd-blue/doc/page_quick_start.html|HOOMD-Tutorial]]
-    * Example call(s):
+    * Example call(s): (could be specified in a shell script, say, 'submit_hoomd.sh')
   <code bash>
+"echo  #!/bin/bash
+#BSUB -e $HOME/my_working_directory/err.log
+#BSUB -o $HOME/my_working_directory/out.log
+#BSUB -W 300/g0001
+#BSUB -q gpushort
+#BSUB -n 1
+#BSUB -R 'rusage[cuda=1]'
+#BSUB -app Reserve1800M
+#BSUB -J hoomd${n}
+timeout -s 2 290m hoomd ../myscript.py
+bash submit_hoomd.sh | bsub"
 </code>
-    * LSF considerations (e.g. ''bsub'' example):
+One can submit a job as: bash submit_hoomd.sh | bsub
+    * LSF considerations (e.g. ''bsub'' example for submitting a job directly [without a bash script] from the command line): bsub -e err.log -o out.log -W 300/g0001 -R 'rusage[cuda=1]' -app Reserve1800M  -q gpushort -n 1 hoomd myscript.py --mode=gpu                                                                            submits a job specified in 'myscript.py' to a single node in the queue gpushort (one could use here the other available GPU queues as well) whereby this job is supposed to run for 300 minutes and requires 1800M operational memory
+    * In the working directory one should have a data file [written in xml-format], which specifies particle positions, velocities, interactions, etc., that the input script, say, 'myscrip.py', needs to read so as to start the simulation.
     * contact: ''milchev'' (Institut für Physik)