↑ Return to GCoE members

Print this Page

Prof. Nagel – ZIH

Center for Information Services and High Performance Computing (ZIH) @ TU Dresden

Director: Prof. Dr. Wolfgang E. Nagel

Powering Central GCoE Resources

The HPC cluster Taurus at the ZIH Dresden provides a set of nodes equipped with 128 dual-K80 and 88 K20Xm Nvidia GPUs. Furthermore, various GPUs are installed to separate nodes for testing and performance evaluation.

CUDA/OpenCL Teaching

Prof. Nagel and ZIH are including GPU and hardware accelerator architecture and programming in a number of classes. A suggested first course for students is the “Highly parallel computer simulations with CUDA and OpenCL” class offered every winter term.

Parallel Programming Expertise

The GCOE and ZIH offer HPC and GPU consulting for accelerating the performance of the user’s application. Performance analysis with Vampir or Score-P and profile-driven optimization guides help to select the critical paths in the code of the user. By supporting the design process of efficient data structures and parallel algorithms the ZIH and the GCOE shares the experiences how to get most out of the memory and compute resources.

Performance Analysis & Tuning

With the growing demand for parallel computing resources, it is important that researchers have access to software tools to optimize both their programming and the analysis of the performance of their codes. Providing the necessary tools that are both robust and user-friendly, can dramatically reduce the amount of time spent developing the application, diagnosing problems in the applications, and leads to more efficient use of the allocated computing resources.

Therefore, the ZIH sees an important part of its mission as developing and deploying robust software tools to support programming, as well as performance analysis and optimization. The aim of these tools is to help programmers, as they develop and parallelize large-scale applications, to identify problems and support more efficient programming for particular systems.

For example, the performance tracking and visualization tool Vampir was originally developed in a research environment and has gained worldwide acceptance as an effective tool for analysis and optimization of parallel programs.

 

Permanent link to this article: https://gcoe-dresden.de/gcoe-members/prof-nagel-zih/