[ESPResSo-devel] Results of the coding week

[Olaf Lenz]

Hello ESPResSo developers!

Between January 9 and 13, we held a coding week at the ICP in Stuttgart.
In this mailing I want to summarize what has happened to ESPResSo in
that time:

New Home Page
-------------
So far, we have been using a simple mediawiki as the homepage of
ESPResSo. Unfortunately, it was not very easy to find your way around on
the page, and it was actually missing some information. Therefore we
have decided to redesign the homepage, using WordPress.
The new home page for ESPResSo is almost ready to be deployed. You can
have a look at it here:

        http://espressomd.org/wordpress/

We will switch to the new homepage probably next week.

ESPResSo Wiki
-------------
A side effect of creating a new homepage is that we can now use the wiki
for other purposes. We have decided to simply open up the wiki to the
community. Already now, you can create your own account and start
editing. You are invited to put anything there that you think is missing
in the documentation or on the new web site! The wiki is accessible here:

        http://espressomd.org/wiki/

At the moment, this URL will show the old homepage.

Various Improvements for 3.1
----------------------------
We have made a number of significant improvements to the ESPResSo
development code, that will show up in ESPResSo 3.1 which we will
hopefully release soon:

* The NPT barostat now works in many more combinations of
  algorithms than so far (e.g. MMM2D, ELC, ...)

* The CPU Lattice-Boltzmann code now work also when Verlet lists are
  used.

* Removed unneccessary overhead of neutral particles in simulations
  with P3M.

* Removed unneccessary overhead of activated but ununsed interaction
  features.

* Some changes in the Verlet list construction yield a performance
  gain of up to 90% in systems where the interaction ranges differ
  significantly, or where some particle types do not interact at all
  (phantom particles).

* The domain decomposition, P3M, LB and MEMD now use MPI cartesian
  communicators. This will hopefully speed up the simulations on some
  platforms.

* Split interaction code into .c and .h files.

* Split off most the Tcl interface (into src/tcl/) from the C core
  code (in src/). Still incomplete.

New Build System: Cmake
-----------------------
Another big issue of the coding week was to change the build system from
the GNU Autotools to Cmake.

  http://www.cmake.org/

The reason for switching are that a cmake-based build system is
significantly easier to maintain than a system based on the autotools.
The cmake-based build system is almost ready to be merged to the master
branch, but a few details and tests are still missing. We expect the new
cmake system to go into the 3.1 release! If you are interested to have a
preview or to help us, you can get the code from the "cmake" branch in
the developers git repo at:

  https://github.com/espressomd/espresso/tree/cmake

New Interface: Python/Cython
----------------------------
The biggest project that was started on the coding week was the
conversion of ESPResSo's interface to Python. To do this, we decided to
use Cython

  http://cython.org/

Although the conversion is by far not complete, we have done some huge
steps towards the goal. The first step was to split off the old Tcl
interface from the C core code. Only after this was mostly done could we
start to work on actually writing some Cython code to interface ESPResSo
functions from Python. At the end of the week, most ESPResSo functions
can now be called from Python.
Please note that we have not yet tried to design a good Pythonic
interface for ESPResSo. Instead we first focussed on creating a basic,
functional interface that allows to call all functions. Once this is
done, we can start to actually design a good interface and directly try
out our ideas.
It might happen that we include the code for the Python interface into
release 3.1, but Tcl will still remain the default for some time to be.
If you are interested to have a look at the current code, you an access
it at

  https://github.com/espressomd/espresso/tree/cython    


To summarize, I believe that I can say that the week has been an
enjoyable and pretty productive experience for all participants, and we
will soon repeat it!

Best regards
  Olaf

-- 
Dr. rer. nat. Olaf Lenz
Institut für Computerphysik, Pfaffenwaldring 27, D-70569 Stuttgart
Phone: +49-711-685-63607

olenz.vcf
Description: Vcard

signature.asc
Description: OpenPGP digital signature