Re: Espressomd-devel Digest, Vol 77, Issue 2

[Pierre de Buyl]

Dear Josh,

Some time ago, Sebastian Bindgen (former student at ICP, now PhD student with
Erin Koos at KU Leuven, and with whom I work on simulations) needed shear flow
for rheology simulations. We picked ESPResSo because Sebastian was familiar with
it and there was a Lees-Edwards feature.

After some time, we hit a wall. The feature duplicated a lot of code which made
it hard to maintain or refactor. Then we noticed that the thermostat imposed the
velocity bias, instead of relying on hydrodynamics + shear to obtain the shear
profile, which did not seem physically appropriate.

At this point, we considered that deciphering and fixing the code would be to
much effort and we decided to remove the feature completely and to start from
scratch. To make this clean, I removed the code in the python branch about two
years ago (!).  So, there is no refactor, only new code, which you can find on
Sebastian's fork (or Rudolf's or mine).

We devised a new strategy based on implementing a modified minimum image
distance function. We had help from Rudolf and Florian, among other topics to
design a proper cell system for Lees-Edwards (columnar), and to make sure we
would use "current" API design.

You can check the new code at 
https://github.com/bindgens1/espresso/tree/le_experimental

There is of course a test suite at testsuite/python/lees_edwards.py and a sample
program code/espresso/samples/lees_edwards_vis.py

To model a fluid, we rely currently on DPD because of its ability to thermostat
non-quiescent fluids. We plan to also support LB, which is waiting for ESPResSo
to migrate to Walberla if I understand well. The idea here is that the
integration of LB takes into account the challenges of integrating cleanly the
Lees-Edwards feature.

Before getting the feature back in the main branch of ESPResSo, we are waiting
for some changes to the ghost code as well, which prevents the merge since the
past summer.

I hope the "story" here is faithful, Sebastian, Rudolf, and Florian might have
further comments.

Regards,

Pierre

PS: I just notice that you unearthed emails from the mailing list from when we
started working on this so there should not be that much surprise.

On Wed, Apr 29, 2020 at 02:34:10PM +0200, Josh Berryman wrote:
> Hi I see on the dev tree that there was some effort to rewrite the
> Lees-Edwards code, last commit was Sep 30th 2018.
> 
> Is this still active?  Why did it need a rewrite?  Having written the LE
> stuff in 2014 or so I find that I want to use this functionality again, but
> the test scripts have disappeared and it seems like somebody started an
> ambitious refactor, integrating with Lattice Boltzmann etc, before giving
> up again.
> 
> If I want to get LE back in action, should I just check out from the main
> tree and see what is there?
> 
> Josh
> 
> 
> 
> On Tue, 13 Mar 2018 at 05:51, <address@hidden> wrote:
> 
> > Send Espressomd-devel mailing list submissions to
> >         address@hidden
> >
> > To subscribe or unsubscribe via the World Wide Web, visit
> >         https://lists.nongnu.org/mailman/listinfo/espressomd-devel
> > or, via email, send a message with subject or body 'help' to
> >         address@hidden
> >
> > You can reach the person managing the list at
> >         address@hidden
> >
> > When replying, please edit your Subject line so it is more specific
> > than "Re: Contents of Espressomd-devel digest..."
> >
> >
> > Today's Topics:
> >
> >    1. Re: call tree for system.integrator.run (Henri Menke)
> >
> >
> > ----------------------------------------------------------------------
> >
> > Message: 1
> > Date: Tue, 13 Mar 2018 10:05:13 +1300
> > From: Henri Menke <address@hidden>
> > To: Pierre de Buyl <address@hidden>,
> >         address@hidden
> > Subject: Re: [ESPResSo-devel] call tree for system.integrator.run
> > Message-ID: <address@hidden>
> > Content-Type: text/plain; charset="utf-8"
> >
> > On Mon, 2018-03-12 at 16:41 +0100, Pierre de Buyl wrote:
> > > Hi,
> > >
> > > I'd like to know wheter there is a diagram somewhere, or if a ESPResSo
> > dev
> > > could
> > > outline it by email, of the call tree that is executed when one calls
> > > system.integrator.run
> >
> > You can set CALL_GRAPH in the Doxyfile to YES and build the Doxygen
> > documentation using 'make doxygen'. ?This will give you quite an impressive
> > graph. ?I attached the one I just produced (I also set?DOT_IMAGE_FORMAT =
> > png).
> >
> > >
> > > I am interested in knowing the various ghost, cell and neighbor list
> > updates,
> > > in
> > > order to understand better the Lees-Edwards implementation. I have
> > started to
> > > work with Sebastian Bindgen on this issue and was a bit lost in the
> > source
> > > tree
> > > at times.
> > >
> > > Thank you for any feedback :-)
> > >
> > > Pierre
> > >