635fc9eae0
Motivation: Results are just wrong for small delays. Modifications: Switching to AvarageTime avoid to rely on OS nanoTime granularity. Result: Uncontended low delay results are not reliable
Microbenchmark tests
See our wiki page.