CONTINUOUS PROFILING - WHERE HAVE ALL THE CYCLES GONE

This article describes the Digital Continuous Profiling Infrastructure , a sampling-based profiling system designed to run continuously on pr oduction systems. The system supports multiprocessors, works on unmodi fied executables, and collects profiles for entire systems, including user programs, shared libraries, and the operating system kernel. Samp les are collected at a high rate (over 5200 samples/sec, per 333MHz pr ocessor), yet with low overhead (1-3% slowdown for most workloads). An alysis tools supplied with the profiling system use the sample data to produce a precise and accurate accounting, down to the level of pipel ine stalls incurred by individual instructions, of where time is being spent, When instructions incur stalls, the tools identify possible re asons, such as cache misses, branch mispredictions, and functional uni t contention. The fine-grained instruction-level analysis guides users and automated optimizers to the causes of performance problems and pr ovides important insights for fixing them.