The goal of this project is to create multicore computer processors with near-mainframe reliability at commodity costs, where the primary costs are additional hardware and additional power consumption. Should this project achieve these goals, it would transform computing. Reliable processors would no longer be a niche product; instead, they would be the commodity products used in desktops, laptops, smartphones, etc.
The research thrusts in this project include error detection, error recovery, diagnosis of permanent faults, and self-repair for tolerating permanent faults. The end result of this project will be a hardware prototype of a low-cost, reliable multicore processor. Building an artifact is a vital part of this project for two reasons. First, a primary project goal is to keep power and area costs low, and these costs cannot be accurately determined using high-level simulation. Second, to transfer this technology to industry and transform the industry, the chip makers must be convinced that the solutions are viable. Chip makers have historically been hesitant to accept ideas that have not been built, not least because important issues can be hidden in a high-level simulator.
