by Taner Pirim

The optimized version of Amber ver.8 is now installed on redwood at /usr/local/appl/Amber8 by using Intel Math Kernel Library, MKL, version 7.2.1. You should set the AMBERHOME environment variable to /usr/local/appl/Amber8. The executables are in /usr/local/appl/Amber8/exe, and the examples are in /usr/local/appl/Amber8/examples. You may browse this directory and copy example files into your own working directory for execution. You should ensure that the $AMBERHOME/exe/ directory is in your PATH environment variable.

To run a parallel Amber program, use mpirun, and specify the number of CPUs/processes to run on with the -np option to mpirun. For example, to run the Amber "nucgen" program on 4 CPUs, using the sample "nuc.in" input file supplied in the /usr/local/appl/amber7/tutorial/polyA-polyT directory:

1. Ensure AMBERHOME environment variable is set to /user/local/appl/Amber8
2. Ensure PATH environment variable includes $AMBERHOME/exe/
3. Copy the inpcrd, prmtop and inpdip files from the /usr/local/appl/Amber8/test/dna_pol/ directory into your working directory as dna_pol_inpcrd, dna_pol_prmtop and inpdip, respectively.
4. Type: mpirun -verbose -np 4 $AMBERHOME/exe/sander -O -i dna_pol_mdin -o dna_pol.out -c dna_po l_inpcrd -p dna_pol_prmtop -r dna_pol.restrt -ref dna_pol_inpcrd
   Make sure it created a new dna_pol.out
5. Verify results in view dna_pol.out
6. For more optimal use of memory, try using the dplace utility:
   Type: mpirun -np $NCPUS dplace -c1-$NCPUS -s1 $AMBERHOME/exe/sander -O -i dna_pol_mdin -o dna_pol.out.v8_8p -c dna_po l_inpcrd -p dna_pol_prmtop -r dna_pol.restrt -ref dna_pol_inpcrd
   where $NCPUS environment variable has been defined as the number of CPUs you want to run on.

We would be interested in hearing how your Amber8 job performs on redwood. In future weeks, MCSR plans to make Amber 8 available on mimosa as well. Currently, the performances of Amber 8 and Amber 7 are being analyzed for various number of cpus by one of our researchers together with our supercomputer consultant. The next step would be to compare the obtained results with Mimosa Amber 8 benchmarks. If you would like more information on the versions of Amber installed on our systems, please visit our Amber website. For more questions on Amber, please contact us via assist@mcsr.olemiss.edu.