Overview
My primary research interest is in methods and systems for highly irregular
parallel applications. Irregularly structured problems exhibit computation and interaction
patterns which are input-dependent, unstructured, and evolving during the computation.
My research activities address aspects of irregularity at several levels of parallel
systems (e.g. parallel algorithm design, programming models, and systems).
Specifically, I'm involved in the following long-term research projects:
- DOTS
DOTS is a system environment for building and executing distributed parallel C++ applications.
It integrates a wide range of different computing platforms into a homogeneous
parallel environment. DOTS provides parallel programming models especially supporting
task-parallel applications that require fully dynamic problem decomposition.
- Cohesion
Cohesion is a Peer-to-Peer based Desktop Grid computing platform that integrates
advanced methods from high performance computing with state-of-the-art concepts from
the realm of distributed systems. By implementing the microkernel architectural pattern
Cohesion is totally configurable even at the system level.
- ParSAT
This research aims at constructing parallel boolean satisfiability (SAT)
solvers which exhibit a high degree of scalability and robustness. Modern
SAT solving methods apply sophisticated heuristics to speed-up the search
process. However, in the parallel case these heuristics induce a considerable
degree of irregularity which make an efficient parallelization most challenging.
- ParCloth
This project investigates on parallel methods for physical simulation of cloth.
We particularly focus on parallel approaches to collision handling, which is a
typical example of a problem with high intrinsic irregularity.
|
|
 |
