Design of heterogeneous manycore platforms
While performance requirements for integrated circuits continue to increase, the semiconductor industry is struggling to keep up. Problems with power density and static leakage came to the fore, while earlier effects such as Dennard scaling faded out. Instead, to gain further performance, the number of processor cores is increased. Such manycore systems with hundreds of processors require a different hardware design than classical embedded systems. The programmability of such systems also differs greatly, since problems have to be parallelized to a high degree.
Runtime adaptive hardware architectures
A modern system-on-chip must save chip area and power to allow the design of small, mobile devices with long battery life. Yet users expect very high computing power for very specific problems, such as media processing or neural network inference. By developing hardware architectures that are reconfigurable during runtime, they become more versatile and even future-proof for use cases that are not yet known at the time of development - with smaller chip area and fewer dark silicon effects. When developing such adaptive hardware architectures, various design parameters must be carefully weighed to ensure that the system is suitable for as many purposes as possible.
Hypervisor-based systems for safety-critical applications
Safety-critical applications are found in particular in the area of control and regulation of vehicles and systems. Such systems are complex to develop and maintain, since flawless functionality must be guaranteed - even after an update. By using modern virtualization technologies, the isolation of individual system components can be improved. This increases the complexity of the overall system and makes it more difficult to assess security. The aim of the research is to reduce the impediments of using a hypervisor in these areas.
|Implementierung eines approximativen Beschleunigers||Bachelor-/ Masterarbeit|
|An Implementation of Dynamic Intra-Tile Reallocatable L1 Cache in homogeneous GPPs||Bachelor-/ Masterarbeit|
|Aufbau und Weiterentwicklung eines Demonstrators für Over The Air Updates auf hochautomatisierten Fahrzeugen|
Supervised student works (selection)
- MA: "Verteiltes Echtzeit-Monitoring von Manycores"
- MA: "Design eines prozessorbasierten Near-Memory-Beschleunigers"
- MA: "Entwicklung eines dynamischen und flexiblen Speichercontrollers für Manycore-Prototypen"
- MA: "Channel Estimation of Dynamic Intra-Tile Reallocatable Cache in Homogeneous GPPs"
- MA: "Hypervisor-Gastüberwachung in Safety-Critical Systems"
- MA: "Implementing an AI Accelerator as a Special Instruction"
- BA: "ARM-Debugger für den Xen-Hypervisor"
- BA: "Implementierung eines approximativen Beschleunigers"
- BA: "Dynamische CNN-Beschleunigung für den i-Core Prozessor"
Lesniak, F.; Kreß, F.; Becker, J.
2021. Applied Reconfigurable Computing. Ed.: S. Derrien, 221–231, Springer Nature Switzerland AG. doi:10.1007/978-3-030-79025-7_15
Anantharajaiah, N.; Kempf, F.; Masing, L.; Lesniak, F. M.; Becker, J.
2019. Proceedings of the 12th International Workshop on Network on Chip Architectures (NoCArc 2019), Columbus, OH, Ocober 12-13, 2019, 1–6, Association for Computing Machinery (ACM). doi:10.1145/3356045.3360718
Masing, L.; Lesniak, F.; Becker, J.
2019. 15th International Symposium on Applied Reconfigurable Computing, ARC 2019; Darmstadt; Germany; 9 April 2019 through 11 April 2019, 319–333. doi:10.1007/978-3-030-17227-5_23
Werner, S.; Masing, L.; Lesniak, F.; Becker, J.
2015. 2015 25th International Conference on Field Programmable Logic and Applications (FPL), London, United Kingdom, 2–4 September 2015, Art.Nr. 7294020, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/FPL.2015.7294020