espressomd-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ESPResSo-users] cell_grid $ verlet_reuse


From: Ulf Schiller
Subject: Re: [ESPResSo-users] cell_grid $ verlet_reuse
Date: Fri, 07 Jan 2011 08:47:33 +0100
User-agent: Mozilla/5.0 (Windows; U; Windows NT 6.1; en-US; rv:1.9.2.13) Gecko/20101207 Thunderbird/3.1.7

Hi,

On 1/6/2011 10:54 PM, Mikheil Azatov wrote:
Hi, yeah the max_num_cells was set to MAX_INT = 32767. After I changed
it to a bigger number everything seems to work fine.

Do you know what is an optimal for speed number of verlet_reuse per
number of integration steps in integrate command ? Also does it matter
for speed how many integration steps I do during integrate command ?

In general it is rather difficult to tell any optimal parameters as they will depend on the specific system you simulate. For verlet_reuse, it depends on the trade-off between the time needed for reconstructing the verlet list and the time saved by reducing the number of particle pairs in the force calculation. This may be very much affected by the density distribution in the system, so if you really want to optimize your parameters you will have to try for your system.

More or less the same holds for the number of integrations performed during the integrate command. Depending on what you do in your script, an additional force calculation may become necessary. As a general guideline, you would usually run the integrate command with the number of steps between measuring observables. However, there are potentially many other reasons where you want control back on the script level. So again, the best advice I can give you is to try for your system.

Best,
Ulf

Thanks,
MIkheil


On Thu, Jan 6, 2011 at 12:56 PM, Axel Arnold <address@hidden
<mailto:address@hidden>> wrote:

    On Tuesday 04 January 2011, Mikheil Azatov wrote:
     > Hi,
     >
     > I noticed my cell_grid size doesn't go above curtain size. Is
    there a limit
     > on the number of cells or something like this ? I'm running the
    code on 2
     > CPUs and the maximum cell_grid I get is 20 40 40. If I increase the
     > simulation box size or decrease the skin parameter it stays the
    same...

    There was a long time a limit, which was lifted only in the latest
    version.
    However, the limit was 512 cells, which is less than what you get,
    so it seems
    that for you the limit is basically infinite. To check, use "puts [setmd
    max_num_cells]", or set this variable to a high value.

     > Also I noticed that my verlet_reuse parameter can only be 0 or 1
    which is
     > obviously not good for parallel computing. If I decrease my skin
    to like
     > 0.1 then verlet_reuse is always 1.0 and if I start increasing the
    skin.
     > the verlet _reuse starts to become 0 more and more integration
    steps. At
     > large skin like 2 or 3, verlet_reuse is 0 every integration step.
    I'm a
     > little bit confused here how is that possible... Shouldn't
    verlet_reuse
     > increase when I increase the skin ?

    verlet_reuse tells you how often _during the last integrate command_ the
    verlet lists where updated on average. Since you do only a single
    time step,
    the answer can be only 0 (if no update occured) or 1 (if an update
    occured),
    and the number of 0s should indeed increase. If you do the typicall
    100 or
    more time steps at once, the output will be more what you expect.

    Axel

     > for {set i 0} { $i < 10000000 } { incr i} {
     >   integrate 1
     >   #setmd time [expr $i*$time_step]
     >     imd positions
     >      puts "verlet_reuse =  [setmd verlet_reuse] "
     >      puts [setmd cell_grid]
     >      if {[expr $i%$pdbstep] == 0 } {
     >      incr pdbindex
     >      writepsf "$name$ident/$name$ident$pdbindex.psf"
     >      writepdb "$name$ident/$name$ident$pdbindex.pdb"
     >      puts " i = $i "
     >      puts "verlet_reuse =  [setmd verlet_reuse] "
     >      }
     > }


    --
    JP Dr. Axel Arnold      Tel: +49 711 685 67609
    ICP, Universit├Ąt Stuttgart      Email: address@hidden
    <mailto:address@hidden>
    Pfaffenwaldring 27
    70569 Stuttgart, Germany






reply via email to

[Prev in Thread] Current Thread [Next in Thread]