If you are publishing a JVM GC benchmark, you should really include the parameters you used for the program as well as what steps you took to find those particu

If you are publishing a JVM GC benchmark, you should really include the parameters you used for the program as well as what steps you took to find those particular parameters. While I think it has gotten better in recent years, there are still many knobs for most JVMs which can make a big difference in performance. Ideally, you would want to have the benchmark developers cooperate with the team creating the JVM you are evaluating in order to be as far as possible with the benchmark sort of like how the TPC benchmarks are done.