Question regarding regtests

[Jörg Saßmannshausen]

Dear Teo,

thanks for the detailed answer, this was exactly the information I wanted to
know.
I agree, it takes some time until you get a feeling for your build and 
you can
trust it (or not).

Best wishes from Glasgow

Jörg

On Dienstag 17 Februar 2009 cp2k group wrote:
> Hi Joerg,
>
> I wouldn't really worry for these cases. Reasons:
>
> 1) As you said NMR (and all linear response part) is quite floppy to
> numerical errors.. it is enough that you rearrange a bit
> the instructions and you get quite different numbers (even in the range
> of 10^-1!!).
> 2) The graphite example is quite similar to NMR.. that's just a matter
> of having built a quite instable example..
> 3) The multiple-walkers metadyn regtest is another one for two different
> reasons: (1) it is based on NH3 which handled with FIST turns out to be
> a quite numerically instable example. (2) relies on information written
> on files. If the file system is not fast enough you may easily end up
> with different results because what you would expect to read is not at
> all written in the file.
>
> As usual perfection is out of this world and let's say that if you have
> only troubles with these examples you can safely ignore them and
> consider your executable well built!
> It's good you posted this question since one has to develop a certain
> feeling in order to rely on an executable.
>
> Ciao,
> Teo



--
*************************************************************
Jörg Saßmannshausen
Research Fellow
University of Strathclyde
Department of Pure and Applied Chemistry
295 Cathedral St.
Glasgow
G1 1XL

email: jorg.sassm... at strath.ac.uk
web: http://sassy.formativ.net

Please avoid sending me Word or PowerPoint attachments.
See http://www.gnu.org/philosophy/no-word-attachments.html