Re: [ESPResSo-devel] A bug in Espresso

[Axel Arnold]

Hi!

There was a small change in Espresso 3.0.2-> 3.1.0. In the older version, 
lattice grid points were located at integer positions, now they are shifted by 
half a grid spacing. This has big advantages with the parallelization, e.g. 
the "cellsystem domain_decomposition -no_verlet_list" no longer required. 
However, because of this, your imposed flow is now half a lattice spacing off 
compared to the old code.

Unfortunately, we forgot to put that into the NEWS /CHANGES file.

Axel


On Friday 29 June 2012 15:25:03 Ulf Schiller wrote:
> Hi Ivan,
> 
> On 06/29/2012 01:37 PM, Ivan Cimrak wrote:
> > Hello,
> > 
> > I have been using Espresso quite some time. Together with Markus
> > Gusenbauer from FH Sankt Poelten we have done some extensions. During
> > my tests I found out that the version 3.0.2 does not compute the same
> > results as the 3.1.0 does.
> > 
> > I did the following test:
> > 2 particles bonded with harmonic interaction. I put them in the
> > lbfluid, I set the velocity of the fluid on the left side of the
> > channel  and I look at the positions of both particles and their
> > distance. I compiled two versions from your sources available at
> > http://espressomd.org/wordpress/download/. In both cases I compiled
> > only the LB and LB_BOUNDARIES in. With these two versions I obtained
> > different results.
> > 
> > I attach two output files.
> > 
> > Do you have any idea why is it so?
> 
> I had a quick look at your data and at first glance it seems there is
> only a numerical difference. Do you think the results are physically
> incorrect? The difference seems to be that the coupling between
> particles and fluid sets in at different time steps (i=11 in 310.dat and
> i=12 in 302.dat). This could be due to round-off differences in the two
> versions of ESPResSo. The results should converge for dt->0.
> 
> Hope this helps,
> Ulf
> 
> > TCL code:
> > set vmd "n"
> > setmd time_step 0.1;
> > setmd skin 0.2;
> > thermostat off;
> > 
> > set pipeX 50; set pipeY 20; set pipeZ 20;
> > setmd box_l $pipeX $pipeY $pipeZ;
> > 
> > part 0 pos 10 10 10 type 1;
> > part 1 pos 15 10 10 type 1;
> > inter 0 harmonic 0.1 5.0;
> > part 0 bond 0 1;
> > 
> > cellsystem domain_decomposition -no_verlet_list
> > lbfluid grid 1.0 dens 1.0 visc 1.0 tau 0.1 friction 1.0
> > 
> > set upscale 5.;
> > if { $vmd == "y" } { prepare_vmd_connection simEspresso 3000 1
> > $upscale; exec sleep 1; imd positions $upscale; }
> > 
> > set fp [open "output.dat" "w"]
> > set i 0
> > while { $i<  500 } {
> > 
> >     puts "$i";
> >     
> >      if { $vmd == "y"} { imd positions $upscale; }
> >     
> >     for { set ii 0 } { $ii<  1} { incr ii } { for { set j 0 } { $j<
> > 
> > $pipeY } { incr j } { for { set k 0 } { $k<  $pipeZ } { incr k } {
> > 
> >                                     lbnode $ii $j $k set u 0.5 0.0 0.0;     
> > } } }
> >     
> >     set aa [part 0 print pos]; set coraa [expr [lindex $aa 0]];
> >     set bb [part 1 print pos]; set corbb [expr [lindex $bb 0]];
> >     
> >     puts $fp "$i:   $coraa, $corbb, [expr $corbb - $coraa]";
> >     
> >      integrate 1;
> >      incr i
> > 
> > }
> > close $fp;
> > 
> > 
> > -------------------------
> > Dr. Ivan Cimrak
> > Dep. Software Technologies, University of Zilina, Slovakia
> > web: www.kst.fri.uniza.sk/~icimrak
-- 
JP Dr. Axel Arnold
ICP, Universität Stuttgart
Pfaffenwaldring 27
70569 Stuttgart, Germany
Email: address@hidden
Tel: +49 711 685 67609