JMH indicates that M1 is faster than M2 but M1 delegates to M2

cowwoc cowwoc at bbs.darktech.org
Mon Aug 8 22:21:13 UTC 2016


Hi,

I created a benchmark in JMH that compares the relative speed of two 
methods, M1 and M2. M1 invokes M2 under the hood so I expected M1 to run 
slower than M2, but in actuality JMH reports the opposite is the case.

I posted this question to http://stackoverflow.com/q/38686958/14731. 
Users correctly pointed out some errors in the original benchmark (I was 
forgetting to sink my outputs) but the relative speeds remain: M1 is 
still running faster than M2.

I tried increasing the number of forks, warmup and measurement 
iterations to no avail.

Can someone please take a look at the source-code and detailed output I 
posted to https://bitbucket.org/cowwoc/requirements/downloads and let me 
know what (if anything) I am doing wrong here?

Thank you,
Gili


More information about the jmh-dev mailing list