[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[ESPResSo] Addendum: Bug in Lennard-Jones interaction in ESPResSo versio
[ESPResSo] Addendum: Bug in Lennard-Jones interaction in ESPResSo versions 2.1.1u up to 2.1.2d
Tue, 10 Mar 2009 17:15:25 +0100
Thunderbird 18.104.22.168 (X11/20080213)
here are some details, that will help you to find out if you were affected by
* if you had only one short range non bonded lennard-jones interaction e.g.:
inter $type_1 $type_2 lennard-jones $eps $sig $rcut $shift $offset
, where $type_1==0 and $type_2==0 (important !)
you were _not_ affected by the bug.
As soon as you had additional short range non bonded interactions ($type_1 > 0 ||
$type_2 > 0)
e.g. "inter 0 1 <whatever> ... "
you were affected by the bug (or at least it was purely random, if you were
affected or not,
because the variable IA_parameters->LJ_min was not copied in copy_ia_params
and thus was uninitialized).
Bonded interactions e.g. "inter 0 fene ..." or long range interactions e.g. "inter coulomb ..."
or weather you ran ESPResSo in parallel or on one CPU only had no influence on the bug at all.
* and again: the bug was present _only_ in the "bleeding egde" CVS versions
2.1.1u to 2.1.2d
* because somebody asked: yes, the ESPResSo testsuite detected the bug
(testcase mass.tcl did fail)
but somehow it could escape our attention until yesterday.
Torsten Stuehn wrote:
due to a bug, which appeared only in the CVS versions 2.1.1u (2009-02-09)
up to 2.1.2d (2009-03-09) some of the lennard-jones pair-interactions
calculated correctly. All simulations that involved lennard-jones
(which is most likely all simulations !) and were done with these
ESPResSo are most likely wrong.
The bug is fixed in version 2.1.2e (2009-03-09)
PLEASE double check your simulations, if you were using the above
mentioned ESPResSo versions !!!