[Pw_forum] newbie update
akohlmey at cmm.chem.upenn.edu
Thu Mar 5 22:12:02 CET 2009
On Thu, 5 Mar 2009, Carlo Nervi wrote:
CN> > Carlo Nervi wrote:
CN> >> Total wall time (s) spent in this run: 5274.73
CN> >> Reference : 720.04
CN> >> There are few discrepancies, but they seems
CN> >> reasonables..
CN> >> or not?
CN> > they do. What doesn't seem reasonable to me is 2 1/2 h for
CN> > a series of tests that take 12 min not on the "ExaFlop
CN> > Machine", but on a rather standard 3GHz IntelCore2 E6850
CN> > (the one I have in front). There is something really
CN> > strange in your setup.
CN> This is exactly what I was wondering. During the tests 4
CN> CPUs were running at 10%, while the other four were
CN> running at 1-5%.
CN> It must be something wrong in my settings, but I cannot
CN> figure it out.
how much memory does your machine have?
what else is running on that machine?
CN> Anyone could please provide a sort of "semi-standard"
CN> make.sys for ifort and mkl with a working mpi?
if you could run the tests, your make.sys should be ok.
either you do something really strange that limits the
MPI communications (and i cannot imagine what that would
be) or there is something else going on that you have not
paid attention to.
how does the machine behave, if you compile a serial executable?
CN> Carlo Nervi carlo.nervi at unito.it Tel:+39 011 6707507/8
CN> Fax: +39 011 6707855 - Dipartimento di Chimica IFM
CN> via P. Giuria 7, 10125 Torino, Italy
CN> Pw_forum mailing list
CN> Pw_forum at pwscf.org
Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu http://www.cmm.upenn.edu
Center for Molecular Modeling -- University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel: 1-215-898-5425
If you make something idiot-proof, the universe creates a better idiot.
More information about the users