[Pw_forum] espresso3.0 different results in serial and paralel mode...

Paolo Giannozzi giannozz at nest.sns.it
Fri Jan 27 09:54:46 CET 2006


On Thursday 26 January 2006 20:17, Konstantin Kudin wrote:

> > I have occasionally seen small numerical differences to build up
> > into rather different convergence pattern. There is nothing obviously
> > wrong in your input and output data.
> >
> > P.
>
>  Is it because of the Davidson diagonalization that the energies start
> out being different already in the 4th decimal? Is there a way to make
> Davidson more consistent on the same architecture across different
> number of cpus ?

maybe there is a way, but I don't know it. As far as I know there is
nothing numerically inaccurate in either the diagonalization or the
charge mixing. 

In this specific case, some starting wavefunctions that are initialized 
with random numbers. There is no warranty that they are the same 
for different machines/number of processors/phases of the moon. 

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



More information about the Pw_forum mailing list