Enabling workload-driven novel reconfigurable systems leveraging system technology co-optimisation.

With the increasing diversity of domains using AI/ML-based applications, there is a growing diversity of workload requirements from electronic systems. While heterogeneous computing systems (CPU, GPU, and NPU accelerators) provide some level of optimization, the static nature of computation, communication, and memory elements limits the level of utilization for emerging workloads. Similarly, state-of-the-art reconfigurable systems (FPGAs, CGRAs, etc.) focus on runtime reconfiguration targeted at functionality modifications only.

The student will be working on a workload-aware exploration framework. The project will involve identifying system-level designs that can leverage existing reconfigurable technologies, across computation, communication, and memory, to provide optimized PPAC (power, performance, area, and cost) trade-offs for a set of workloads.

Type of project: Internship only

Type of work: 10% literature survey to gain an understanding of the landscape of technology innovations and stat-of-the-art reconfigurable technologies. 90% hands-on modelling and framework development for driving the design and exploration of technology-architecture co-design.