[CP2K:1359] Re: Wrong tests

Laino Teodoro teodor... at gmail.com
Sat Sep 27 16:02:42 UTC 2008


the point is that using NUMERICALLY those numbers does not make any  
Your libraries, compilers and OS are different, and you can't expect  
to get the same numbers.

You should use those results qualitatively. In your case for example  
you have several RUNTIME FAILURES,
while the official regtests is clean.

This is enough to point that the problem is in your settings  
(compiler, mpi, stack settings or whatever).
That's the best use you can do of those informations.

About possible miscompilation.. well.. that's another story.. but you  
would very probably realize that running
a system and analyzing the data..


On 27 Sep 2008, at 17:54, ilya wrote:

> That's exactly what I used.
> Ilya.
> On Sep 27, 7:49 pm, Laino Teodoro <teodor... at gmail.com> wrote:
>> There's an official regtest that tells you if the CVS snapshot of
>> CP2K is giving correct numbers or not:
>> http://cp2k.berlios.de/regtest.html
>> of course this does not guarantee anything on your compiler.. but
>> that's another story..
>> teo
>> On 27 Sep 2008, at 17:45, ilya wrote:
>>> But I should have some reference to say that my first regtest gave
>>> correct results.
> >

More information about the CP2K-user mailing list