<div dir="ltr">Dear Iain, Thank you very much for your rapid response and detailed explanation as well as suggestion. Cheers, Huan On Thursday, October 3, 2013 2:45:05 PM UTC+2, IBethune wrote:<blockquote class="gmail_quote" style="margin: 0;margin-left: 0.8ex;border-left: 1px #ccc solid;padding-left: 1ex;">Dear Huan, Like other parallel programs which make use of MPI, CP2K does not know or care exactly where each MPI process is running. Allocation and placement of MPI processes onto physical nodes is up to the MPI library you have installed on your cluster. We regularly run CP2K on several thousands of CPU cores on large-scale HPC machines, so depending on the size of system you wish to study you will find CP2K to be very scalable. Without knowing more, I would suggest that you benchmark CP2K with your existing system to better understand the performance and scalability. How to run a parallel job across multiple compute nodes across depends exactly on which MPI library (e.g. OpenMPI, MPICH2, Intel MPI...) you use and if you use any batch job submission system (e.g. SLURM, PBS, Sun Grid Engine...). Typically, one needs to have the mpi deamon running on each node and all the nodes added to a node-list file or similar, then the 'mpirun' command will launch MPI ranks across your two nodes. Best thing is to speak to your local systems administrators who may be able to help more specifically. Also, you might consider upgrading to a more recent release of CP2K as 2.3 is over a year old and there have been many improvements, new features and performance improvements since. Cheers - Iain -- Iain Bethune Project Manager, EPCC Email: <a href="javascript:" target="_blank" gdf-obfuscated-mailto="PKQKiUb20HUJ">ibe...@epcc.ed.ac.uk</a> Twitter: @IainBethune Web: <a href="http://www2.epcc.ed.ac.uk/%7EibethuneTel/Fax" target="_blank">http://www2.epcc.ed.ac.uk/~ibethune Tel/Fax</a>: +44 (0)131 650 5201/6555 Mob: +44 (0)7598317015 Addr: 2404 JCMB, The King's Buildings, Mayfield Road, Edinburgh, EH9 3JZ On 3 Oct 2013, at 13:30, <a href="javascript:" target="_blank" gdf-obfuscated-mailto="PKQKiUb20HUJ">hua...@mail.huji.ac.il</a> wrote: > Dear All, > > I would like to know is there an up-limit number of cores for CP2K parallel computing. My supervisor want to purchase some new computers for running CP2K/QuickStep. So we need to make sure that is it possible to run parallel computing with, such as, 40 or more cores? > > Another question is that Is it possible for sharing the CPU in different nodes for parallel computing? For example, if it is possible to use 2 nodes (each node has 40 processors) for one MD simulation? If it available, how to do it? Do I need to write some special codes in submitting script? The version of CP2K I used is 2.3. > > I appreciated it very much if someone could answer my questions. Thanks in advance. > > Best regards, > Huan > > -- > 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 <a href="javascript:" target="_blank" gdf-obfuscated-mailto="PKQKiUb20HUJ">cp2k+uns...@googlegroups.com</a>. > To post to this group, send email to <a href="javascript:" target="_blank" gdf-obfuscated-mailto="PKQKiUb20HUJ">cp...@googlegroups.com</a>. > Visit this group at <a href="http://groups.google.com/group/cp2k" target="_blank">http://groups.google.com/group/cp2k</a>. > For more options, visit <a href="https://groups.google.com/groups/opt_out" target="_blank">https://groups.google.com/groups/opt_out</a>. -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. </blockquote></div>