In this talk, I describe a deep problem in computer performance analysis and our solution, Shim, which improves fidelity by three or more orders of magnitude.   Developers and architects spend a lot of time trying to understand and eliminate performance problems. Unfortunately, the root causes of many problems occur at a fine granularity that existing continuous profiling and direct measurement approaches cannot observe.  

I will describe Shim, a continuous profiler that samples at resolutions as fine as 15 cycles; three to five orders of magnitude finer than current continuous profilers. Shim's fine-grain measurements reveal new behaviors, such as variations in instructions per cycle (IPC) within the execution of a single function.   This work was led by my student Xi Yang and is joint work with my collaborator Kathryn McKinley, who is at Google.