[CP2K-user] [CP2K:11953] posting to log file
patri... at gmail.com
Sat Jul 6 15:31:20 CET 2019
given the scale of your calculation, you should be well within optimal
scaling with only 3 nodes (36 cores). Make sure you have a correctly
compiled version of CP2K and a correctly written submission script (do you
spawn 36 mpi processes on a single 12 core node, by any chance?). With 300
atoms and at the semilocal level of theory, I expect you should see linear
speed-up with up to 10 nodes. And certainly, even adding more nodes beyond
that should not slow down the calculation, quite the opposite, actually.
On Sat, 6 Jul 2019 at 14:36, Raitis <raitis... at gmail.com> wrote:
> Thank you for the answer, Patrick.
> Here is the output of one of the simulations (i cut out most of the
> non-interesting data as the file was to large to attach).
> I am running on 1 node with 12 cores. I tried to to run on more nodes, but
> by running with mpirun -np 36 cp2k.popt it was even slower. I will get in
> touch with our server administrator to understand how to run it properly on
> more than one node.
> Thank you for the suggestion. I thought that this might have something to
> do with wave function recalculations at the restart, but i could not figure
> out how to write this information from the first simulation. Will modify
> the input and try it this way.
> On Saturday, July 6, 2019 at 2:05:48 PM UTC+3, Patrick Gono wrote:
>> Dear Raitis,
>> Could you please attach the output file as well? So that we can get some
>> idea of the timings.
>> How many nodes are you using? Is it possible to use more, to speed up the
>> calculations? Are you limited by memory or the lack of CPUs?
>> Finally, if you have to restart a calculation, especially an MD, you
>> should choose to start from an already computed wavefunction file (if
>> you're not already doing it):
>> WFN_RESTART_FILE_NAME *project name*-RESTART.wfn !
>> default naming convention
>> SCF_GUESS RESTART
>> &END SCF
>> &END DFT
>> This should speed up any restarted simulation, as you don't have to solve
>> the KS equations from scratch at the beginning.
>> On Thu, 4 Jul 2019 at 15:41, Raitis <rai... at gmail.com> wrote:
>>> I am new to the cp2k and I am trying to calculate the Raman spectra by
>>> following this tutorial
>>> <https://brehm-research.de/files/spec_tutorial_2018.pdf>. I am having
>>> several questions about cp2k performance/way of calculating things:
>>> 1. Is it normal that for several first hours of my calculations there
>>> are nearly no output in the log file (it prints all the cp2k and quickstep
>>> related information and then nothing for a while). Is it supposed be like
>>> that or something might be wrong? What is happening during this time.
>>> 2. The walltime on my server is 2 days and I cannot finish all the
>>> simulation in this time, therefore I am continuing it using the .restart
>>> file. But than again, once I restart the simulation, there is this long few
>>> hour time were nothing is written in the log file, and only later the
>>> actual simulation data are printed.
>>> You received this message because you are subscribed to the Google
>>> Groups "cp2k" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
>>> an email to c... at googlegroups.com.
>>> To post to this group, send email to c... at googlegroups.com.
>>> Visit this group at https://groups.google.com/group/cp2k.
>>> To view this discussion on the web visit
>>> For more options, visit https://groups.google.com/d/optout.
> You received this message because you are subscribed to the Google Groups
> "cp2k" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to cp... at googlegroups.com.
> To post to this group, send email to cp... at googlegroups.com.
> Visit this group at https://groups.google.com/group/cp2k.
> To view this discussion on the web visit
> For more options, visit https://groups.google.com/d/optout.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the CP2K-user