16 Oct
2010
16 Oct
'10
5:24 a.m.
At Fri, 15 Oct 2010 17:43:23 -0400, Bryce Lelbach wrote:
How do I write accurate, empirical and reproducible performance tests? I realize this is a very broad question, but my knowledge on this subject is lacking, so I'd appreciate a comprehensive answer (e.g. I'm not looking to benchmark a particular program/library, just looking for a strategy for designing good performance tests). I have read up on the subject, but I'm interested in what you guys (e.g. Boost users/developers) do.
You might want to read through http://github.com/boost-lib/parameter/blob/master/test/efficiency.cpp, (which is in the Boost distro as libs/parameter/test/efficiency.cpp) -- Dave Abrahams BoostPro Computing http://www.boostpro.com