Explore statistical methods for evaluating software performance changes in this 39-minute conference talk from linux.conf.au 2021. Learn how to determine if performance differences are statistically significant or "down in the noise" using Student's T test and Mann-Whitney U test. Discover insights into the skewed distributions often seen in kernel microbenchmarks and how to interpret results. Gain practical knowledge applicable to software development and performance analysis, presented in an accessible manner for those with basic mathematical understanding.
Overview
Syllabus
"It's (not) down in the noise." - Peter Chubb (LCA 2021 Online)
Taught by
linux.conf.au