This shows you the differences between two versions of the page.
| Next revision | Previous revision |
| software:gromacs:mills [2017-10-23 17:02] – created sraskar | software:gromacs:mills [2021-04-27 16:21] (current) – external edit 127.0.0.1 |
|---|
| ====== Gromacs on Mills ====== | ====== Gromacs on Mills ====== |
| | |
| | Templates for Grid Engine queue submission scripts are available for batch GROMACS jobs in ''/opt/shared/templates/gridengine/gromacs''. Copy and edit the template based on your job by following the usage described in ''gromacs.qs''. |
| | |
| | <note tip>It is a good idea to periodically check in ''/opt/shared/templates/gridengine/gromacs'' on a specific cluster for changes in existing templates, or the addition of new templates, designed to provide the best performance for a particular version of GROMACS on that cluster.</note> |
| |
| ===== Batch job ===== | ===== Batch job ===== |
| |
| On Mills, the template (''/opt/templates/gridengine/gromacs.qs'' specifies a number of resource options to provide the best performance. The number of slots specified for ''NPROC'' should be a multiple of 24 in conjunction with specifying [[/general/jobsched/exclusive|exclusive access]] to allow your job to run on the node(s) required based on the number of slots requested without any other jobs running on those node(s) too. Also specifying ''vpkg_require gromacs/4.5.5-intel64'' (single-precision intel) or ''vpkg_require gromacs/4.5.5-open64'' (double-precision open64) has provided the best results. By default the template uses double-precision (''md_rund''), however uncommenting ''USE_SINGLE_PRECISION=YES'' will specify the single-precision version (''md_run'') instead. Again for best performance, binding cores and only using half of the slots per node are defined as the default due to the shared FP (Floating-Point) unit on Mills. | On Mills, the template (''/opt/shared/templates/gridengine/gromacs.qs'' specifies a number of resource options to provide the best performance. The number of slots specified for ''NPROC'' should be a multiple of 24 in conjunction with specifying [[abstract:mills:runjobs:queues#mills-exclusive-access|exclusive access]] to allow your job to run on the node(s) required based on the number of slots requested without any other jobs running on those node(s) too. Also specifying ''vpkg_require gromacs/4.5.5-intel64'' (single-precision intel) or ''vpkg_require gromacs/4.5.5-open64'' (double-precision open64) has provided the best results. By default the template uses double-precision (''md_rund''), however uncommenting ''USE_SINGLE_PRECISION=YES'' will specify the single-precision version (''md_run'') instead. Again for best performance, binding cores and only using half of the slots per node are defined as the default due to the shared FP (Floating-Point) unit on Mills. |