||The group headed by Dr. Marwedel focuses on the design of efficient embedded systems. The talk will provide an overview of the research performed in this group:
- Embedded systems connected to the physical environment must meet timing constraints. Typically, timing verification is performed after software generation, possibly leading to many design iterations until timing constraints are finally met. In order to avoid such loops, compilers can be made timing-aware. An approach for such integration will be presented together with some results.
- The next part of the talk will be about addressing the memory wall problem. In particular, the use of energy-efficient and timing-predictable scratchpad memories will be presented. Results on the worst-case execution time aware compilation for an automotive application using scratchpad memories will be included.
- The final part will cover other research projects in the DAES group. These include two projects of the collaborative research project on resource constrained machine learning (SFB 876). In the first project, resource-efficient image analysis for an advanced real-time virus detector is the subject for design space exploration. The second project aims at speeding up the execution of programs written in R (a language frequently used for statistical analyses) by about two orders of magnitude, thereby enabling more comprehensive statistical analyses, particularly in life sciences. Another project addresses the problem of guaranteeing timeliness for systems based on unreliable hardware.