[ESPResSo-devel] conversion to python

[Ivan Cimrak]

Hi guys,


During the python transition of object-in-fluid part of Espresso, we 
encountered some discrepancies in LB computations. We got different 
results for the movement of particles in LB fluid when computed with tcl 
version and python version. This lead us to design a simple experiment 
with plain Espresso without object-in-fluid: We put one particle in LB 
fluid, move LB fluid with ext_force and output the position after 500 
steps. We got different x-coordinate of the particle, the difference was 
around 0.1 percent. (instead of 6.19593 we got 6.19105, for example.)


Here are the steps that we did:

1. download the current (today's) versions of master branch (commit 
6cbef0e) and python branch (commit 272a361).

2. install tcl version using: bootstrap.sh, configure 
--without-python-interface, myconfig.hpp with EXTERNAL_FORCES and LB, make

3. install python version using: create myconfig.hpp with 
EXTERNAL_FORCES and LB, cmake -DPYTHON_EXECUTABLE=$(which python), make

4. run the attached 018lb-test.tcl and 018lb-test.py



The differences are too high to ignore. Do I miss anything?

Thanks,
Ivan

018lb-test.py
Description: Text Data

018lb-test.tcl
Description: Tcl script