DBCSR assertion failed

[Ondrej Marsalek]

Hey again,

new day, new build. And it does not have the problem reported below,
one of the bugfixes since yesterday did the trick. Thanks guys,
whichever one it was.

Regards,
Ondrej


On Fri, Feb 19, 2010 at 00:19, Ondrej Marsalek
<ondrej.... at gmail.com> wrote:
> Dear all,
>
> I attach a sample input that produces an error with CP2K from
> 2010-02-18. My output is also attached. I have tested that the error
> occurs on 1, 4 and 32 cores. The error occurs just before any output
> of SCF iterations. This is it:
>
>  ASSERTION FAILED:         9.EQ.        8
>  dbcsr_error_print_stack
>  error%msg=source/new bin array sizes not equal
>  error%routine=rebin_distribution
>  error%etype=          -1
>  error%level=           1
>  error%line=        6726
>  error%stack%level=           0
>  dbcsr_error_stack_size=          10
>
> The system is a 32 water cluster in a 20 A box. I am sorry that it is
> not smaller, but reducing the number of molecules made the problem
> disappeared. The input is pruned as much as possible to still see the
> error. Getting rid of ROKS and using UKS makes it run fine, but I need
> ROKS, as I need SIC. It fails in a minute or two, but it needs a
> relatively large memory (around 4 GB).
>
> The executable is built with gfortran 4.3.3 and uses MKL 11.1. I will
> provide more information if needed.
>
> Could someone please check if they can reproduce this? And if so,
> could someone please have a look at the issue?
>
> Thanks a lot,
> Ondrej
>