Nothing ran under measurement, usually because the compiler removed the code under test. These results are not comparable, so they count as unchanged.
Preventing compiler optimizations