[Pw_forum] MPI_Bcast error from cp.x

[Omololu Akin-Ojo]

Thanks, Paolo.
It turns out that if one is NOT using orthogonalization='ortho' in
optimizing the electrons, the eigenvalues should not be computed:
NOTE: eigenvalues are not computed without ortho
(I got this warning on a 64bit Linux while it crashes on AIX machine with
the error I reported previously: ERROR: 0032-117 User pack or receive buffer
is too small)

My way of getting around it is to avoid printing "physical quantities to
standard output" as it seems the error comes at the printing step. So, I set
iprint (and isave) in the &control section to something larger than my
maximum number of steps, nstep. It is still running so I don't know what it
will do at the end.

Grazie ancora.

o.

On Tue, Nov 16, 2010 at 7:19 AM, Paolo Giannozzi <giannozz at democritos.it>wrote:

> Omololu Akin-Ojo wrote:
>
> > I actually don't have a clue in which part of the program
> > I should increase/decrease buffer/message size.
>
> you shouldn't increase/decrease anything. You should look in the
> output file if there are other error messages. Then you should
> verify whether this error is reproducible ([referrably by running
> it on a different machine). If it is, please provide code version,
> input data, execution detail so that the error can be reproduced.
> If it is not: your compiler or MPI libraries are, with very high
> probability, buggy
>
> P.
> --
> Paolo Giannozzi, Democritos and University of Udine, Italy
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>



-- 
***************** Seek GOD! *************
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.democritos.it/pipermail/pw_forum/attachments/20101116/95d7983a/attachment.htm