ZURUECK HOCH VOR INHALT SUCHEN

» Back to overview
Proposing Institution

Karlsruhe Institute of Technolgy, Institute of Theoretical Informatics
Project Manager

Michael Axtmann
Am Fasanengarten 5
76131 Karlsruhe
Abstract
In this project we want to look at novel load balancing techniques for complex problems. This includes a number of techniques. The first area we want to explore is scalable high quality graph partitioning. Here, we have already developed a parallel code that can partition highly unstructed graphs such as web graphs and social networks using much less time compared to ParMetis while improving solution quality. Moreover, the algorithm is very scalable on a smaller machine (with up to 2048 cores usable by a user of the machine). The second subtopic that we want to look at is scalable parallel sorting to efficiently compute space filling curves that can be used as a fast locallity increasing preprocessing step for partitioning. To start with this project, we would like to get a developer account on the SuperMUC.

Impressum, Conny Wendler