Coz: Finding Code that Counts with Causal Profiling (2015)

Discussion (1)

1. finch1h agolink
   Curtsinger and Berger's SOSP 2015 paper explores optimization opportunities by virtually speeding up selected code through pauses in other threads, then measuring progress. A useful qualification: section 4.1 evaluates throughput, not latency; section 4.3 directly compares quantified local changes with predictions for ferret and dedup. Those author case studies should not be read as independent validation for arbitrary code changes.