Finding Code that Counts with Causal Profiling

 Coz is a code profiler that find optimization opportunities
 missed by traditional profilers. Coz employs a technique called
 causal profiling that measures optimization potential. This measurement
 matches developers' assumptions about profilers: that optimizing
 highly-ranked code will have the greatest impact on performance. Causal
 profiling measures optimization potential for serial, parallel, and
 asynchronous programs without instrumentation of special handling for
 library calls and concurrency primitives. Instead, a causal profiler
 uses performance experiments to predict the effect of
 optimizations. This allows the profiler to establish causality:
 "optimizing function X will have effect Y," exactly the measurement
 developers had assumed they were getting all along.