On the Validity of Performance Antipatterns at Code Level

Abstract

Performance antipatterns at code level should be avoided to assure good application performance. Performance antipatterns avoidance is hard, since it requires up-to-date knowledge of these antipatterns. Common lists of antipatterns, like the performance rules of the static code checker PMD, only contain limited information about versions and circumstances where the performance antipatterns are valid. We close this gap by prodiving a suite of 30 performance benchmarks. Each of this benchmarks checks whether the performance antipattern is measurable in Java 6, 7, 8, 11 and 12. We find that two of the 30 performance checks are not valid in the current OpenJDK 12.