Making parallel programs reliable with stable multithreading

Abstract

Stable multithreading simplifies the interleaving behaviors of parallel programs, offering new hope for making parallel programming easier. To eliminate non-determinism, several groups of researchers, have dedicated themselves to building deterministic multithreading (DMT) systems that force multithreaded programs to always execute the same thread interleaving. By mapping each input to only one schedule, DMT brings determinism, a key property of sequential computing, into multithreading. Stable multithreading, or StableMT, reuses each schedule on a range of inputs, mapping all inputs to a dramatically reduced set of schedules. StableMT can stabilize program behaviors for better reliability, so it works efficiently and deterministically while greatly improving precision of static analysis. Moreover, it promises to help solve the grand challenge of making parallel programming easy.