Closed Baltoli closed 7 years ago
Robert mentioned that it might be worthwhile to have this benchmark perform some actual meaningful computation rather than just spinning - doing this means that end-to-end time can be a better measure. The computation needs to be selected to take enough time and generate enough contention.
This benchmark should be a synthetic one that has a lot of contention on locks. Needs to run for long enough that an improvement is obvious, and should also think about how best to profile the benchmark.