[CP2K:3505] Re: CI BAND calculation

Urban Borštnik urban.b... at gmail.com
Tue Sep 20 09:45:07 CEST 2011


Hello,

On Mon, 2011-09-19 at 09:56 -0700, Axel wrote:
> [...]
> you have to help yourself. apparently the segfault 
> originates in your BLACS installation. there are multiple
> reasons why that can happen. my first guess would be
> that it was configured incorrectly. BLACS has its own 
> testers and you should run those tests with the same
> 50 processors, that crash its call in cp2k. testing 
> ScaLAPACK and PBLAS/BLACS properly requires
> more than just compiling it and running some blind 
> tests.

Since ran into a similar BLACS/OpenMPI segfault, I would just like to
give a hint where to look (besides OpenMPI's documentation on compiling
BLACS with it, of course!):
TRANSCOMM = -DUseMpi2 -DPOINTER_64_BITS=1
in Bmake.inc solved the problem for me.

But as Axel wrote, do run the tests included with BLACS.

Cheers,
Urban

> 
> 
> cheers,
>      axel.
>  
>         Best regards,
>         Nikita
>         
>         





More information about the CP2K-user mailing list