Performance and energy efficiency analysis for innovative computer architectures
Although high performance computers become more and more powerful, especially in terms of their theoretical peak performance, achieving best performance for scientific applications requires a detailed understanding of the architecture and performance of the underlying computing systems. In addition to conducting detailed studies of the function and performance of new processor architectures and memory hierarchies, ZIH is actively developing benchmark environments to facilitate such research and to significantly improve the understanding of the hardware behavior in research and teaching.
Another essential focus along this area of research is the energy efficiency of computer systems. With the decreasing efficiency gains through improved manufacturing processes, a broader approach is necessary. The performance optimization of parallel applications is promising in terms of energy efficiency, particularly through intelligent control (within the scientific applications) of the underlying hardware.
In terms of novel computer architectures, another research focus concerns the hardware accelerators, in the present mainly represented by general purpose GPUs (GPGPUs) and many-core processors. The research at ZIH in this area ranges from the detailed study of hardware accelerators, to the development of software tools for performance analysis, and to porting and optimizing scientific applications on accelerator architectures.
Current Projects
- FIRESTARTER: A processor stress test utility
- HDEEM: High Definition Energy Efficiency Monitoring
- Dresden GPU Center of Excellence
- cfAED: Center For Advancing Electronics Dresden
- BenchIT: Flexible and portable performance analysis for computer systems
- SPEC: Standard Performance Evaluation Cooperation
- READEX: Runtime Exploitation of Application Dynamism for Energy-Efficient Exascale Computing
- HAEC: Highly Adaptive Energy Efficient Computing
- HI-CFD / Scout: Highly efficient implementation of CFD codes for HPC many-core architectures
-
Score-E: Scalable Tools for the Analysis and Optimization of Energy Consumption in HPC
- Cool Silicon: Energy Efficiency Innovations from Silicon Saxony
- GASPI: Global Address Space Programming Interface
- SIOX: Scalable I/O for Extreme Performance
- eeClust: Energy-efficient cluster computing
- Fast Matrix Multiplication on Cell (SMP) Systems
- Software Tracing on Cell/B.E. Processors
- EP-Cache
- PARbench
© Robert Gommlich
Daniel Hackenberg
Deputy Director (acting) / Scientific Assistant
Send encrypted mail via the SecureMail portal (for TUD external users only).
Center of Information Services and High Performance Computing (ZIH)
Center of Information Services and High Performance Computing (ZIH)
Visitor Address:
Willers-Bau, Room A 204 Zellescher Weg 12
01069 Dresden
- work Tel.
- +49 351 463-32055
- fax Fax
- +49 351 463-37773