Daedalus: System-Level Design For Multi-Processor System-on-Chip

Short Introductory Video

Daedalus^RT is an adapted version of the Daedalus framework for designing hard-real-time embedded streaming systems. Daedalus^RT is based on the initial Daedalus methodology. However, it differs in the following aspects: 1) support for multiple applications. The initial Daedalus flow supports only a single application, whereas Daedalus^RT flow supports multiple applications, 2) replacing the complex DSE with very fast yet accurate schedulability analysis to determine the minimum number of processors needed to schedule the applications, and 3) using hard-real-time multiprocessor scheduling algorithms providing temporal isolation to schedule the applications. An overview of Daedalus^RT flow is shown in the figure below:

The DSE phase of Daedalus is replaced with the CSDF model derivation and analysis phases. In the analysis model derivation phase, the CSDF graphs equivalent to the PPN models are derived. The derivation phase uses WCET analysis to determine the WCET of each actor. Determining the WCET can be done through either static code analysis or exact measurements on the target platform. After that, the CSDF models are analyzed to determine the platform and mapping specifications. In Daedalus^RT, the CSDF is used as an analysis model, and PPN is used for code generation.