problem on cray-xt

[Matt W]

Hi Jun,

>  More problematic is that I am not lucky enough to get another
> similar job running.
> Any other suggestions?

Not much help, but executables from last year (Jun 20 2009 with
pathscale comp and Oct 29 2009 GCC) work. Unfortunately as they're
static builds it's not so easy to see what libraries might have
changed.

I tried GCC and PATHSCALE, and downgrading libsci to xt-libsci/10.2.1
with the current source but no joy.

Matt

>
> Cheers,
> Jun
>
> On Wed, Sep 29, 2010 at 3:24 PM, Matt W <mattwa... at gmail.com> wrote:
> > Hi Jun,
>
> > I can reproduce this problem on xt-6 (CVS of Mon Jul  5 11:54:19 BST
> > 2010).
>
> > Giving the ghost atom a basis set gets rid of the problem (changing
> > BASIS_SET NONE -> DZVP-MOLOPT-SR-GTH).  So some piece of code
> > somewhere is trying to get a basis set for the H1 atom type and not
> > finding it.  What bit of code that is I don't know.
>
> > Cheers,
>
> > Matt
>
> > On Sep 28, 10:40 am, Jun <chen... at gmail.com> wrote:
> > > Dear all,
>
> > > Initially, I got the following error message for a newly set-up input
> > > files on cray-xt6,
> > > {    3,    5}:  On entry to DSTEQR parameter number  -31 had an
> > > illegal value
>
> > > Since that input file is quite complicated, I cut it down and
> > > represent
> > > the problem using a simpler input file, as uploaded (err.inp).
> > > Sometimes, the
> > > error message changed just with small changes of input (really not
> > > important change) or changes of numbers of cpus. I also got NaN error
> > > during scf runs (mostly at the first step). If using DIIS minimizer
> > > for OT, "singular diis matrix"
> > > can also appear. Strangely, I can get some testing jobs run
> > > occasionally
> > > by just decreasing the numbers of cpus or changing the order of atom
> > > coordinates in input file.
>
> > > CP2K on cray-xt6 was built with GCC 4.4.4, Cray MPT 5.0.2, FFTW 3.2.2
> > > and Cray Libsci 10.4.6. The problem persists on cray-xt4 with the same
> > > compilation setting.
> > > Hope some of you can reproduce the error, any suggestions will be
> > > appreciated.
>
> > > Regards
> > > Jun
>
> > --
> > You received this message because you are subscribed to the Google Groups
> > "cp2k" group.
> > To post to this group, send email to cp... at googlegroups.com.
> > To unsubscribe from this group, send email to
> > cp2k+uns... at googlegroups.com <cp2k%2Bun... at googlegroups.com>.
> > For more options, visit this group at
> >http://groups.google.com/group/cp2k?hl=en.