As discussed in today’s meeting please allocate 50K core-hours to NeutrinoSim. Thanks, John J. Low Principal Computational Science Specialist Computing, Environment and Life Sciences Building 240, 2143 9700 South Cass Avenue Argonne National Laboratory Argonne, IL 60439. 630-252-0045 www.linkedin.com/pub/john-low/15/8b0/5aa/ From: <Low>, "John J." <[email protected]<mailto:[email protected]>> Date: Wednesday, March 25, 2015 at 4:56 PM To: "Yang, Guang" <[email protected]<mailto:[email protected]>> Subject: Re: [allocations-admins] FW: [LCRC Accounts] Project Request: NeutrinoSim Guang, We will allocate 50K core-hours to your project to parallelize your code with MPI and to get started. When you can demonstrate good parallel scaling up to hundreds or core with your code, we can discuss increasing your allocation. John J. Low Principal Computational Science Specialist Computing, Environment and Life Sciences Building 240, 2143 9700 South Cass Avenue Argonne National Laboratory Argonne, IL 60439. 630-252-0045 www.linkedin.com/pub/john-low/15/8b0/5aa/ From: Guang Yang <[email protected]<mailto:[email protected]>> Date: Tuesday, March 24, 2015 at 7:21 PM To: "Low, John J." <[email protected]<mailto:[email protected]>> Subject: Re: [allocations-admins] FW: [LCRC Accounts] Project Request: NeutrinoSim Hi John, I also learned how to use MPI just now from a guy in our division. I think I got clear with this thing and could run jobs with MPI. So for the non-sequential jobs with only different random seeds, I can efficiently use the cores. Please let me know if we could get the core hours. Thanks, Guang On Tue, Mar 24, 2015 at 3:02 PM, Guang Yang <[email protected]<mailto:[email protected]>> wrote: Hi John, Thank you very much. I read it and I think the task farmer and apron with listing different job scripts methods are quite straightforward to me. My jobs can run in parallel. For example, generating 100,000 random events takes too long. So I would break it into 100 jobs *1000 events. (1,000 events takes 1 hour and half.) They are not sequential. Just make sure that if I can do something like: 1.#PBS -q regular 02.#PBS -l mppwidth=144 03.#PBS -l walltime=12:00:00 04.#PBS -N my_job 05.#PBS -e my_job.$PBS_JOBID.err 06.#PBS -o my_job.$PBS_JOBID.out 07. 08.cd $PBS_O_WORKDIR 09.aprun -n 32 ./a.out & 10.aprun -n 65 ./b.out & 11.aprun -n 24 ./c.out & 12.wait Then I can definitely run them in parallel. I may just have 16 exactly same job scripts. Thanks, Guang On Tue, Mar 24, 2015 at 2:38 PM, Low, John J. <[email protected]<mailto:[email protected]>> wrote: Guang, Shashi Aithal suggested this link to more examples of running parallel jobs in this link. https://www.nersc.gov/users/computational-systems/hopper/running-jobs/examp le-batch-scripts/<https://www.nersc.gov/users/computational-systems/hopper/running-jobs/example-batch-scripts/> John J. Low Principal Computational Science Specialist Computing, Environment and Life Sciences Building 240, 2143 9700 South Cass Avenue Argonne National Laboratory Argonne, IL 60439. 630-252-0045<tel:630-252-0045> www.linkedin.com/pub/john-low/15/8b0/5aa/<http://www.linkedin.com/pub/john-low/15/8b0/5aa/>