Necessity of Collaboration
Invasive computing is a new paradigm for parallel processing. In order to fully exploit all of its enormous benefits, new programming and architectural issues of many-core computing architectures need to be jointly investigated. Therefore, this TCRC requires a wide range of expertise that goes far beyond pure computer architecture, pure programming and pure algorithmic issues of parallel computing. In fact, one of the key novelties in invasive computing is that both hardware and software of an invasive computing system adapt to each other in order to deliver a high degree of efficiency in many-core computing where efficiency can be measured in performance/power-consumption, performance/hardware-footprint and fault-tolerance. As such, the expertise needed to build an invasive computing system spans from algorithms through compilation and operating systems to diverse high performance architectures as well as to necessary principles like run-time adaptation and hardware reconfiguration. A TCRC with three sites involved was the way to go in order to start this project. All the areas of expertise should be covered with the top experts in their respective fields. In the following, we will give more details showing that the three chosen sites provide excellence in all required key competencies.
Key Competences
The three sites Karlsruhe, Munich and Erlangen offer a unique combination of key competences which comprehensively cover the invasive research vision. Roughly, the three sites offer the following competencies necessary for InvasIC: 1) KIT: i) Adaptive Embedded Computing (Becker, Henkel), ii) Algorithmic Engineering (Sanders), iii) Programming Languages and Software (Snelting), iv) Embedded Architectures and Low Power Design (Henkel). 2) TUM: i) Network Processors and On-Chip Communication (Herkersdorf), ii) Technology Abstraction and Dependability (Schlichtmann, Schmitt-Landsiedel, iii) High Performance Computing (Gerndt). FAU: i) Reconfigurable Computing and Massively Parallel Processor Design (Teich), ii) Operating Systems (Schröder-Preikschat).
Besides, each of the three site has excellent research in a wide range of challenging applications that will show the benefits of invasive computing (Bungartz, Stechele, TUM; Dillmann, KIT;).
Expertise
In terms of large-scale research programmes, fortunately, our TCRC can heavily rely on a 6 year expertise gained in the DFG Priority Programme 1148 (Rekonfigurierbare Rechensysteme) that is coordinated by Prof. Teich. Here, Prof. Becker, Prof. Herkersdorf, Prof. Teich and Prof. Stechele have been funded and cooperated successfully; their results, published in international venues, create an important seed for the TCRC. Prof. Sanders is internationally visible expert in the field of parallel programming and algorithm engineering, in particular for parallel systems. Prof. Sanders is also coordinator of DFG Priority Programme 1307 ìAlgorithm Engineeringî, and the results from this programme will serve as valuable input to invasive algorithm techniques. The group of Prof. Snelting is renowned for programming language technology and its applications. For developing the necessary hardware structures, Prof. Herkersdorf and Prof. Schmitt-Landsiedel have long-standing experience in designing and characterising integratedMPSoCs in silicon, with particular focus on techniques for handling and correction of errors caused by steadily unreliable transistor and component technologies and on hardware monitoring methods. Profs. Becker, Herkersdorf, Schlichtmann and Henkel are experts for simulation, prototyping and synthesis of real-working MPSoCs. Prof. Schröder-Preikschat is a leading German operating systems expert and, in particular, looks back on distinguished knowledge regarding the design and the development of parallel operating systems. Similarly, Prof. Teich and Prof. Herkersdorf cooperate already in AIS, a project sponsored by the BMBF and dealing with new design processes for adaptive MPSoCs that are able to detect transient and performance errors, faults and interferences in the data and control paths of RISC processors with the goal to enhance the reliability of single components and complete MPSoCs. Prof. Henkel is an internationally renowned expert on embedded systems. He is the speaker of the DFG Priority Programme 1500 ìDesign and Architectures of Dependable Embedded Systemsî that will start in Spring 2010. Prof. Becker and Prof. Henkel have also already shown successful collaboration in joint projects like in the DFG KAHRISMA project and the DFG DodOrg project.
Experience
Concerning experience with industrial partners, Prof. Teich runs an associated working group in the well known Fraunhofer IIS (Integrated Circuits) and has ongoing research projects with major semiconductor and chip design companies such as IBM and Alcatel-Lucent. Karlsruhe operates in close cooperation to an Institute on Nanoelectronics and the FZI. Profs. Schlichtmann and Schmitt-Landsiedel work in close collaboration with industry on robust design in sub-100 nm-processes to cope with increasing variability and reliability issues. Prof. Dillmann, leading an SFB in Karlsruhe on Humanoid Robotics, as well as Prof. Herkersdorf and Prof. Stechele, working on image processing on MPSoCs, will contribute applications in robotics and image processing which are likely to benefit heavily from the invasive computing paradigm with focus on improving real-time behaviour. Finally, the project by Profs. Bungartz and Gerndt will investigate invasive computing with focus on high performance computing applications where performance and resource utilisation gains of invasive computing shall be shown. In summary, it can be said that the diversity of the areas of expertise and their excellence for InvasIC is comprehensively covered by the three sites Erlangen, Karlsruhe and Munich. It offers the best combination to make the new paradigm of invasive computing a reality.