Alex Gaynor • 7/15/2013

Your tests are not a benchmark

This article clarifies why running a test suite under PyPy often yields slower results than the actual application. It details how PyPy's Just-In-Time (JIT) compiler requires repeated code execution to optimize, a condition rarely met in short, one-off tests. It also covers how test-specific activities like monkey-patching cause deoptimization, making tests unsuitable for benchmarking PyPy's real-world performance.

0 comments

#unit testing #performance optimization #benchmarking