Discovering the Runtime Structure of Software with Probabilistic Generative Models Anonymous Author(s) Affiliation

Address email Modern computer systems have become so complex that understanding and predicting the performance of programs is a significant challenge. We address the problem of estimating the efficiency of a program on a hypothetical hardware architecture, typically measured by cycles per instruction executed (CPI). Simulating the entire program at the hardware level is extremely computation intensive. Consequently, researchers have resorted to picking portions of the program execution that are considered typical, called simulation points, and extrapolating from detailed simulation of these points to estimate CPI of the program as a whole. Simulation points are chosen by running the program at the software level and collecting statistics such as counts of basic code blocks executed, resulting in a time-varying sequence of basic-block vectors. The state-of-the-art algorithm for selecting simulation points uses k-means clustering of basic-block vectors. W