ApproxBench is a resource for evaluating approximate computing research.
Work in approximate computing studies complex application-level interactions between efficiency and quality. Convincing evaluations need to study applications that are known to error-resilient.
ApproxBench aims to make approximate-computing evaluations easier and better by providing:
The ApproxBench harnesses are in development. In the mean time, check out EnerJ’s Java benchmark suite and ACCEPT’s C/C++ suite. You may also be interested in AxBench from Georgia Tech’s ACT Lab.
ApproxBench is a project of the University of Washington Department of Computer Science & Engineering department in the Sampa lab. Please contact Adrian Sampson with inquiries.