Prof. Dr. rer. nat. Christian Schröder
MPS Kronos - A Tool for Model-based Development of Distributed Applications on Parallel Computers with Automatic Code-Generation of the Communication Infrastructure Using Standard MPI.
The international standard MPI (Message passing Interface) provides a powerful and efficient way to program parallel systems for high performance computing. There exists a variety of non-commercial as well as commercial implementations for C/C++ and FORTRAN that support the standard functionality as well as extensions that might be useful for specifc purposes. However, writing parallel programs using MPI means a challange for every programmer since (usually) the MPI program is a unique piece of software which is executed simultanously on an arbitrary number of processors. Thus, inter- and intra-process communication have to be handled simultanously on the same logical level which not only increases the complexity but also yields higher error probabilities, especially with respect to process communication, i.e. deadlocks and lifelocks.
The purpose of the project is to examine possibilities of seperating the tasks for inter- and intra-process programming by shifting the description of the inter-process communication to a higher logical, i.e. more abstract level. One way to do so is to use graphical modeling languages with a subsequent code-generation. Here we will use a subset of the upcoming standard modeling language UML 2.0.
People: Jan Gatting and André Wehe
Spinhenge@home - Discovering Novel Physics of Molecular Magnets using Public Resources High Performance Computing
The word's computing power and disk space is no longer primarily concentrated in supercomputer centers and machine rooms, but distributed in hundreds of millions of personal computers and game consoles belonging to the general public. Public-resource computing, sometimes called "global computing" or "Peer-to-peer computing", uses these resources to do scientific supercomputing. This paradigm enables previously infeasible research and encourages public awareness of current scientific research.
The purpose of the project Spinhenge@home is to use these gigantic computational resources to drive theoretical investigations on that relatively new research field of molecular magnets. For the communication infrastructure we use the "Berkeley Open Infrastructure for Network Computing" (BOINC). For more details click here.
People: Thomas Hilbig
Send mail to
questions or comments about this web site.