Embedded systems are special-purpose computers within devices such as cell phones, cars, and medical equipment. In the future, embedded systems will dominate general-purpose computers in volume, economic value, and frequency of user interaction. This seminar summarizes recent work to improve the quality of embedded systems and ease their design.
Memory is a scarce resource for embedded systems such as cell phones. Software applications are demanding increasing amounts of memory. However, adding RAM increases packaging cost, size, and power consumption. In this talk, we first describe CRAMES, an efficient software-based RAM compression technique for embedded systems. CRAMES more than doubles usable RAM without changes to hardware, without changes to applications, and with negligible performance and energy consumption penalties. It is used in the cell phones of a major vendor. This technology won a Computerworld Horizon Award.
Power is the source of the greatest problems facing both embedded and general-purpose processor designers. High-power processors rapidly deplete battery energy. High spatial and temporal power densities bring high temperatures, which result in decreased lifetime reliability. High temperatures also increase leakage power consumption, thereby closing a self-reinforcing power-temperature feedback loop. The wages of power are bulky short-lived batteries, huge heatsinks, large on-die capacitors, high server electric bills, and unreliable integrated circuits. The alternative is optimizing integrated circuit power consumption, temperature, and reliability.
We will survey modeling, analysis, design, and synthesis techniques to solve these problems systematically.