Dynamic data race detectors are effective at finding concurrency bugs, but their high runtime overhead has limited adoption. According to a new arXiv paper, this overhead stems primarily from pervasive instrumentation of memory accesses, which slows programs down significantly.

The paper, "Instrumentation Optimization for Practical Dynamic Race Detection," addresses this bottleneck directly. While the abstract does not detail specific methods, it signals that the work targets the instrumentation layer itself, aiming to reduce the cost that makes these detectors impractical in many real-world scenarios.

If successful, such optimizations could make dynamic race detection a more routine part of development, allowing programmers to catch data races without paying a severe performance penalty. The paper is listed as a replace-cross version on arXiv, suggesting ongoing refinement of the research.