jeudi 14 Décembre, 14h en salle Minquiers
Shuvra S. Bhattacharyya, University of Maryland and Tampere University of Technology
The DSPCAD Framework for Dataflow-based Design and Implementation of Signal Processing Systems
Dataflow-based design methodologies are important for addressing the growing complexity of signal processing systems, including systems for wireless communications, machine learning, video processing, and biomedical instrumentation. This complexity is compounded by the need to exploit parallelism effectively, satisfy stringent power consumption requirements, and work with heterogeneous processing resources and sensing modalities. The DSPCAD Framework is a computer-aided design (CAD) framework that helps designers to apply the formalisms of the dataflow paradigm in software/hardware design processes for signal and information systems. The DSPCAD Framework is specifically oriented toward exploration of interactions and optimizations across different signal processing application areas; alternative dataflow models of computation; and alternative target platforms along with their associated platform-based tools (e.g., field programmable gate arrays, graphics processing units, multicore digital signal processors, and low power microcontrollers). In this talk, I will first review challenges in dataflow-based design methodologies for signal processing systems, and then present key aspects of the DSPCAD Framework that help to address these challenges.