How Not to Write a Microbenchmark (2002)

Also of interest: 'Why Aren’t More Users More Happy With Our VMs?'—https://tratt.net/laurie/blog/2018/why_arent_more_users_more...

Also excellent:

"A JVM Does That???" by Dr Cliff Click (Video, Presentation, 2016) https://youtu.be/-vizTDSz8NU

Any chance this could be shared somewhere that's not Slideshare?

It was originally a PDF, archive.org to the rescue: https://web.archive.org/web/20120216164427/http://www.azulsy...

I had not realized just how bad SlideShare had become over the years. One full screen ad every 6 slides is crazy

The "know what you test" slide is the real painful part. Benchmarks should ideally always be run periodically under a profiler to help you understand what's actually being measured and question your assumptions about your application.

Slide 6 is wonderful.

As part of an admonition about measuring things wrong, Dr. Click pulls numbers out of thin air. Very convincing!

/s