SCoRPiO

Manufacturers go to great lengths to guarantee fault-free operation of their products by introducing redundancy in voltage margins, conservative layout rules, and extra protection circuitry. However, such redundancy may not be really required, given that many modern workloads, such as multimedia, machine learning, visualization, etc. can tolerate a degree of imprecision in computations and data. The project aimed to relax reliability requirements for the hardware layer by researching methods that allow the system- and application-software layers to synergistically characterize the significance of various parts of the program for the quality of the end result, and their tolerance to faults. Based on this information, the system software steers computations and data to either low-power, yet potentially unreliable or higher-power and reliable functional and storage components. Also, the relative significance of different parts of the computation is exploited to support approximate computing in a targeted way.