BigFoot: static check placement for dynamic race detection
BigFoot: static check placement for dynamic race detection
复制标题
BigFoot：用于动态种族检测的静态检查位置
DOI：
10.1145/3062341.3062350
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
Stephen N. Freund
中科院分区：
文献类型：
--
作者：
Dustin Rhodes;C. Flanagan;Stephen N. Freund
精确的动态数据竞争检测器提供了强大的正确性保证，但开销很高，因为它们通常将分析状态保存在每个堆内存位置的单独影子位置中，并且它们在每个堆访问上检查（并可能更新）相应的影子位置。 BigFoot 动态数据竞争检测器使用静态和动态分析技术的组合来合并检查和压缩阴影位置。使用 BigFoot，对对象或数组的多次访问通常会引发一次合并检查，该检查操作单个压缩阴影位置，从而使性能比 FastTrack 提高 61%。
Precise dynamic data race detectors provide strong correctness guarantees but have high overheads because they generally keep analysis state in a separate shadow location for each heap memory location, and they check (and potentially update) the corresponding shadow location on each heap access. The BigFoot dynamic data race detector uses a combination of static and dynamic analysis techniques to coalesce checks and compress shadow locations. With BigFoot, multiple accesses to an object or array often induce a single coalesced check that manipulates a single compressed shadow location, resulting in a performance improvement over FastTrack of 61%.