GGrantIndex
← Search

CAREER: Thermal-Aware Synthesis of Embedded Processors

$428,000FY2006CSENSF

Northwestern University, Evanston IL

Investigators

Abstract

The demand for embedded processors is spreading from mobile devices to entertainment systems, appliances, and cars. These embedded processors are designed for increasingly complex tasks and have to deliver ever-higher levels of performance. To fulfill these demands, designers rely on steady miniaturization and large-scale integration. However, this results in rising power densities and failure rates. Power dissipated on a chip is converted to heat. When the heat output from the processor combines with the high temperatures frequently found in embedded system environments, expensive processes can be required to keep the system cool enough to run. Such solutions may not be feasible for embedded systems due to cost and/or size constraints. This NSF Career project develops thermal-aware synthesis algorithms and tools for embedded processor design. Throughout the various stages of hardware/software co-synthesis, this research is developing design techniques for effective control of peak temperature and uniform thermal profiles. This includes thermal-awareness in partitioning of functionality between hardware and software, synthesis (such as resource selection, allocation, and assignment, task scheduling, and memory allocation), and software optimizations (such as data layout optimization) within the co-synthesis design flow. Effective management of temperature in embedded processor design will help sustain the feasibility of technology scaling for highly sophisticated, high performance embedded processors that enable emerging consumer products. As the improved design methodologies and tools that address thermal design issues aid the production of low cost, reliable embedded processors, a variety of embedded applications can be serviced more effectively and new, sophisticated applications can be conceived in a variety of domains from everyday consumer applications to mission-critical engineered systems.

View original record on NSF Award Search →