Research Statement

Even the simplest kernel has a diverse structure of execution behavior. This is a consequence of the structured loops and repetitive traversal of call graphs universal to all code. Understanding this time varying behavior has ramifications across several different computer science applications: from new optimizations in compiler design to streamlining parallel programs and computer architecture. My research has focused on the development of analysis techniques to characterize time varying program behavior using Phase Analysis and Variational Path Profiling. Phase Analysis groups together similar sections of the execution into phases and has been used to expedite computer architecture simulation. Variational Path Profiling finds variations in program behavior and communicates it back to a programmer to help find ‘hot-spots ’ for optimization. 1 Phase Analysis Understanding the cycle-level behavior of a processor running an application is crucial to modern computer architecture research. To gain this understanding, architects typically employ detailed cycle-level simulators. Unfortunately, this level of detail comes at the cost of simulation speed, and simulating the full execution of an industry standard benchmark on even th