Research

Real-time embedded systems are everywhere, from phones and tablets to cars and spacecraft, and they keep demanding more processing power. Multiprocessor systems answer that, and bring the problem of multiprocessor scheduling, which multi-core architectures and optimal scheduling algorithms have pushed to the front of the field.

The group works on optimal global scheduling algorithms. In theory they use processors better than partitioned scheduling; in practice they lose out because of the overhead of frequent scheduling points, migrations and preemptions. The group is reducing that overhead, and also studies energy-constrained scheduling.

The second stream targets reconfigurable architectures: the design of next-generation field programmable gate arrays (FPGAs) and the computer aided design (CAD) algorithms that map circuits onto them. FPGAs can implement almost any digital circuit and have grown fast in capacity and market, but compared with ASICs they take more area, run slower and use more power. The aim is to improve FPGAs through both their architecture and their CAD algorithms.