[CPMD-list] Running cpmd on Altix (again)
Axel Kohlmeyer
akohlmey at cmm.chem.upenn.edu
Mon May 7 21:55:34 CEST 2007
On Mon, 7 May 2007, Jiri Houska wrote:
hi jiri (and cpmd-list for completeness),
i ran your input on my machine (intel em64t
with intel compiler 9.1) and it works fine
without any modifications, so in principle
it has to work on the altix.
my next two suggestions of possible reasons for the failure are:
1) your compiler miscompiles the detsp.F/ratom.F file
(and probably others, too).
2) your version (you are using 3.9.1 which is way outdated)
has a bug in destsp.F/ratom.F
in fact, it may even be a combination of both.
i remember spending significant time chasing down
warnings from the intel and other compilers between
version 3.9.1 and 3.9.2, so this might be solved
when compiling up-to-date sources with up-to-date
compilers or perhaps as a first test with using
-O0 optimization level. some patchlevels of the
intel compilers were pretty bad with miscompiling
code at higher optimization. particularly the 8.0
and the 9.0 initial release versions.
cheers,
axel.
JH> Hi Axel,
JH> thanks for the answer. I really believe that the input is correct (in
JH> fact I did not write it from scratch, but copied from directory with
JH> files succesfully used elsewhere ...). The files were copied from
JH> windows but I performed
JH> tr -d "\015" < oldfile > newfile (and now also dos2unix)
JH> for all of them. The pseudo. file is in place. Running it
JH> interactively leads to same results (i.e. no results :( ). Please find
JH> the input and output in the attached tarfile if you will be so kind
JH> ... .
JH> Regards,
JH> Jiri
JH>
JH>
JH> Quoting Axel Kohlmeyer <akohlmey at cmm.chem.upenn.edu>:
JH>
JH> > On Mon, 7 May 2007, Jiri Houska wrote:
JH> >
JH> > jiri,
JH> >
JH> > first, please try running interactively in serial
JH> > to debug your input, if you are not doing this already.
JH> >
JH> > JH> Dear Prof. Hutter (and others),
JH> > JH> thanks for the answer. The output file (copied below) ends without any
JH> > JH> error message just below the first section repeating basic information
JH> > JH> from the input file (the first missing line looks to be "DETSP| THE
JH> > JH> NEW SIZE OF THE PROGRAM IS ..."). For my life I am not able to find
JH> >
JH> > is the pseudopotential in place and the name spelled correctly
JH> > in the input.
JH> >
JH> > is your input in unix-style ascii? some fortran compilers have
JH> > trouble parsing text which has dos/windows style end-of-line
JH> > conventions (in case you wrote your input file on a pc and
JH> > transferred it to the altix). you can try running dos2unix
JH> > on your input to correct that.
JH> >
JH> > otherwise, please send the complete input attached as a gzipped
JH> > file, so that no mailer can mess with it...
JH> >
JH> > cheers,
JH> > axel.
JH> >
JH> > JH> any arror in the simple input file ... . Any help is very welcomed!
JH> > JH> Jiri
JH> > JH>
JH> > JH> ******************output:******************
JH> >
JH> > [...]
JH> >
JH> > --
JH> > =======================================================================
JH> > Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu http://www.cmm.upenn.edu
JH> > Center for Molecular Modeling -- University of Pennsylvania
JH> > Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
JH> > tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel: 1-215-898-5425
JH> > =======================================================================
JH> > If you make something idiot-proof, the universe creates a better idiot.
JH> >
JH> >
JH> >
JH>
JH>
JH>
--
=======================================================================
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 CPMD-list
mailing list