Closed ppxasjsm closed 9 years ago
It's because the head node has intel cores whereas nodes001-003 have opterons. If you set CPU optimisation to SSE4.1 or SSE2 the resulting binary should run on every node (but not at maximum performance).
To squeeze maximum performance we probably would need to provide two types of gromacs binaries. For now I suggest to just use the CPU performance optimisation that sits at the lowest common denominator.
J
Dr. Julien Michel, Royal Society University Research Fellow Room 263, School of Chemistry Joseph Black Building, University of Edinburgh David Brewster Road
Edinburgh
EH9 3FJ United Kingdom phone: +44 (0)131 650 4797
On Sun, Jul 19, 2015 at 2:50 PM, ppxasjsm notifications@github.com wrote:
Compiling gromacs on the headnode and installing to /home/common/gromacs, does not seem to allow for gromacs to run on the compute nodes. (It runs ok on the headnode. ) Not quite sure what the issue might be here, but I haven't looked into it yet.
— Reply to this email directly or view it on GitHub https://github.com/michellab/Cluster/issues/8.
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
Ok. What exactly is SSE? I suppose this is something to google. But yes for the compilation this can of course be done.
Ok still not sure what SSE does exactly, but Gromacs module is now working, with version 4.6. Probably should also provide version 5.0* at some point.
Compiling gromacs on the headnode and installing to /home/common/gromacs, does not seem to allow for gromacs to run on the compute nodes. (It runs ok on the headnode. ) Not quite sure what the issue might be here, but I haven't looked into it yet.