Elastic Coarse-Grained Reconfigurable Arrays

Standard microprocessors are not getting faster anymore, however, consumer and industry demand for faster and lower-power computing is insatiable. The project aims to research and develop a novel computer-hardware platform to be used for accelerated low-power computing versus the use of microprocessors. The new hardware platform, called a coarse-grained reconfigurable array (CGRAs), can be configured accordingly for a specific application. The platform is comprised of a 2D array of user-configurable computational blocks that can be programmably connected to one another. The platform is “elastic” in the sense that its internal computational blocks perform their work whenever input data is ready, rather than performing work at fixed prescheduled times. CGRAs are a promising candidate for realization of future application accelerators, both in the datacentre/cloud, as well as at the edge.

Faculty Supervisor:

Jason Anderson

Student:

Partner:

École polytechnique fédérale de Lausanne

Discipline:

Engineering

Sector:

Education

University:

University of Toronto

Program: