XPS: CLCCA: Improving Parallel Program Reliability Through Novel Approaches to Precise Dynamic Data Race Detection

点击翻译按钮获取中文摘要

The ubiquity of multi-core processors in everything from servers tosmartphones has demanded a similar prevalence of multi-threadedprograms to take advantage of multiple cores. Unfortunately, writingmulti-threaded code is still in the Wild West era of error-prone manualsynchronization, unchecked concurrency bugs, and undefined semantics.One common symptom of an error in a multi-threaded program is a datarace. Data races arise when a program performs concurrent updates tosome location without synchronization. Automatically detecting dataraces during program execution enforces strong safety properties formulti-threaded programs. While techniques for data race detection existthey slow program execution too much to be viable for always-onenforcement.To make always-on detection of data races practical, the project aimsto develop new algorithms, language extensions, runtime systems, andhardware support to improve the efficiency of data race detection. Theresearch includes validation of these techniques via formal proofs,experiments with multi-threaded benchmark suites, and detailed hardwaresimulation. The researchers plan to openly distribute the systemsbuilt for this project to facilitate examination by other researchersand integrate the research results into the computer architecturecourses they teach. If successful, the proposed technology willimprove the safety and quality of the vast amounts of multi-threadedcode running on today's and tomorrow?s multi-core devices.