Link: https://www.eecs.northwestern.edu/~robby/courses/322-2013-sp...

Great paper, yes. Immediately useful? More like disheartening, because it doesn't really tell you how to be sure your measurements are OK.

Our lab addressed some of the issues with Stabilizer [0], which "eliminates measurement bias by comprehensively and repeatedly randomizing the placement of functions, stack frames, and heap objects in memory".

[0] http://plasma.cs.umass.edu/emery/stabilizer.html "Stabilizer: Statistically Sound Performance Evaluation" by Charlie Curtsinger and Emery Berger, ASPLOS 2013