[Pw_forum] CVS cp.x
Konstantin Kudin
konstantin_kudin at yahoo.com
Wed Nov 2 01:35:44 CET 2005
OK. Here is some indication of where the problem might be.
When I compiled things with -D__USE_INTERNAL_FFTW, I had the same
error message on AMD64.
Once I recompiled with the external FFTW library, the code works.
So either today's fix for "pseudopot.f90" took care of the problem,
or, there is an issue with the internal FFTW library.
It is likely that nobody's build is testing -D__USE_INTERNAL_FFTW, so
things have gotten out of sync.
Kostya
--- Paolo Giannozzi <giannozz at nest.sns.it> wrote:
> On Tuesday 01 November 2005 05:46, Alexander Shaposhnikov wrote:
>
> > It does not work for me [...] The architecture is AMD64,
> > ifort9.0 + icc9.0, goto-blas + ACML, mpich2.
>
> the possibility that a subtle bug shows up only for some compilers
> or some architectures or some values of the phase of the moon
> can never be ruled out. A compiler bug seems to me a more likely
> reason, though. There are a few things you can do to in order to
> clarify this issue:
> - check if later releases of ifort 9.0 (if any) have the same problem
> - check if g95 has the same problem (or a different one)
> - check if the problem disapperas by disabling optimization
> - find the place in the code where the error appears (use grep),
> trace it back (for instance: if it is a badly set variable, look for
> theplace where is set). If it is a real bug, it shoud be tracked to
> a precise location.
>
> In any event, waiting for the problem to disappear is hardly a
> good strategy
>
> Paolo
>
> --
> Paolo Giannozzi e-mail: giannozz at nest.sns.it
> Scuola Normale Superiore Phone: +39/050-509876, Fax:-563513
> Piazza dei Cavalieri 7 I-56126 Pisa, Italy
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>
__________________________________
Yahoo! FareChase: Search multiple travel sites in one click.
http://farechase.yahoo.com
More information about the users
mailing list