Program Interferometry

Modern microprocessors have many microarchitectural features. Quantifying the performance impact of one feature such as dynamic branch prediction can be difficult. On one hand, a timing simulator can predict the difference in performance given two different implementations of the technique, but simulators can be quite inaccurate. On the other hand, real systems are very accurate representations of themselves, but often cannot be modified to study the impact of a new technique. We demonstrate how to develop a performance model for branch prediction using real systems. The technique perturbs benchmark executables to yield a wide variety of performance points without changing program semantics or other important execution characteristics such as the number of retired instructions. By observing the behavior of the benchmarks over a range of branch prediction accuracies, we can estimate the impact of a new branch predictor by simulating only the predictor and not the rest of the microarchitecture. We call this technique Program Interferometry based on its similarity to astronomical optical interferometry. Using measurements of the Intel Xeon E5440 Processor, we quantify the impact of branch prediction on a set of benchmarks, developing regression models that estimate the performance given by changes in the branch predictor. We incorporate these models into a simulator allowing us to estimate the impact of several branch predictors. This first study in program interferometry points the way to future work on estimating the impact of other microarchitectural structures. We demonstrate the potential for interferometry to estimate the impact of L1 and L2 caches by perturbing data layouts.