Dear Bartek
thank you very much for your fast response.

Here are the details you asked for:
At the moment i use serial siesta (vers 3.0.b), thus i do not think this
related to mpi. I know 10 fs are a long time but i only try to "heat" up
my system and cool it down after it was sufficent long at 1500 K. Thus i
am not interested in the dynamics in this part of simulation. But i will
try to minimize the timestep- hopefully this will work.
I think my memory consumption is not that high, because siesta uses only
900Mb
on a 4Gb Machine and CG or Anneal do run without any problems.

Greeds
Michael
P.S. How i could get the data in stderr?

>> But i have some serious trouble to do it. I have a system with 184 Atoms
>> and I want to simulate it a distinct time at 1500 K and ambient pressure
>> with a time step of 10 fs. Using the Anneal option everything seems to
>> run
>> fine. But switching it to Nose or NoseParrinelloRahman siesta performs
>> the
>> SCF calculation, calculates the forces- and than nothing. It still runs
>> but it do not produce any output. Has anyone a idea whats going wrong?
>
> I think the timestep is way to much, I hope you know what you are doing.
>
> Regarding your problems - I can imagine quite a few possible reasons.
> Can you get any info from stderr? is it serial or parallel run? I
> experienced some implementations of mpi didn't kill all child processes
> when crashed. Are you sure there is enough memory in your machine?
>
> Cheers,
> Bartek
>
>

Responder a