Abstract
Instrumenting code to collect profiling information can canse substantial execution overhead. This overhead makes instrumentation difficult to perform at runt/me, often preventing many known o]fiine feedback-directed optimizations from being used in online systems. This paper presents a general framework for performing instrumentation sampling to reduce the overhead of previously expensive instrumentation. The framework is simple and effective, using codeduplication and counter-based sampling to allow switching between instrumented and non-instrumented code. Our framework does not rely on any hardware or operating system support, yet provides a high frequency sample rate that is tunable, allowing the tradeoff between overhead and accuracy to be adjusted easily at runt/me. Experimental results are presented to validate that our technique can collect accurate profiles (93-98% overlap with a perfect profile) with low overhead (averaging ,-,6% total overhead with a naive implementation). A Jalapefio-specific optimization is also presented that reduces overhead further, resulting in an average total overhead of ~3%.