On February, 17th, the S/PHI/nX developers met in the scope of the first S/PHI/nX Developer Convention.
In the past two years the development of the S/PHI/nX project has been pushed forward in four different places, i.e., by
- Max-Planck-Institut für Eisenforschung (MPIE),
- Tyndall National Institute,
- HPC computing center Garchin, and by
The contributing institutions introduced awesome new features to their branches. Björn Lange from MPIE focussed on quantitatively optimized atomic orbitals (Quamols) – a new way of expressing plane-waves in a significantly smaller basis-set. The S/PHI/nX Dirac layer has been greatly improved (actually rewritten to a large extend) by Christoph Freysoldt and is now applicable to the projector-augmented waves method (PAW). Further highlights are the implementation of hybrid functionals and the improvement of the convergence performance in the electronic minimization loop.
A highly generalized and incredibly flexible implementation of a multiband k.p Hamiltonian and continuum elasticity models has been presented by Oliver Marqardt (Tyndall). Instead of editing the Hamiltonian directly into the source code (which is common for most k.p codes), the new S/PHI/nX version provides an intuitive user interface which makes working with S/PHI/nX great fun. The latest version also provides excellent performance specs.
Klaus Reuter from RZG Garching presented the latest results about the progress on the MPI parallelization of S/PHI/nX. A vivid discussion about the next steps of the implementation of MPI calls in the Dirac layer followed.
An exciting discussion about how the S/PHI/nX project can benefit from GPUs followed the talk of Robert Spatschek, MPIE.
The hosts Vaclav Bubnik and myself (Gemmantics) which are focussing on the general purpose C++ layout SxAccelerate discussed the latest features which have been contributed such as the new interprocess communication layer, efficient hashing methods, and the new signal/slot engine.
Looking forward to seeing you at SxDevCon 2013!
The slides of the presentations can be found below.
|I – S/PHI/nX Accelerate:|
|Getting rid of S/PHI/nX wrappers||Sixten Boeck||Gemmantics|
|Hashing S/PHI/nX||Vaclav Bubnik||Gemmantics|
|Content Management with SxAccelerate||Sixten Boeck||Gemmantics|
|SxAccelerate – Runtime Dynamics||Sixten Boeck||Gemmantics|
|II – Parallel computing and GPUs:|
|Component-based programming in S/PHI/nX||Sixten Boeck||Gemmantics|
|SxAccelerate/IPC – Sketches, Concepts, and Ideas||Sixten Boeck||Gemmantics|
|SxAccelerate – the new IPC Layer||Vaclav Bubnik||Gemmantics|
|First results on the distributed memory parallelization of the S/PHI/nX library||Karsten Reuter||RZG|
|GPUs beyond gaoming: The poor man’s supercomputer||Robert Spatschek||MPIE|
|III – MPI Branch|
|A flexible, pw-based formulation of continuum elasticity and multiband k . p-models||Oliver Marquardt||Tyndall|
|Quamols and spline-fit||Björn Lange||MPIE|
|PAW and hybrid functionals, future perspectives||Christoph Freysoldt||MPIE|
- Max-Planck-Institut für Eisenforschung
- Yin Cui
- Christoph Freysoldt
- Albert Glensk
- Claas Hueter
- Björn Lange
- Zongrei Pei
- Gernot Pfanner
- Robert Spatschek
- Siyuan Zhang
- Oliver Marquardt
- RZG Garching
- Klaus Reuter
- Sixten Boeck
- Vaclav Bubnik
Please follow the author on Twitter.
You can join the S/PHI/nX discussion forum by sending a blank email to
sxusers-subscribe at mpie dot de. More infos at http://sphinxlib.de.