Large-Scale Program Behavior Analysis for Adaptation and Parallelization

Motivated by the relentless quest for program performance and energy savings, program execution environments (e.g. computer architecture and operating systems) are becoming reconfigurable and adaptive. But most programs are not: despite dramatic differences in inputs, machine configurations, and the workload of the underlying operating systems, most programs always have the same code running with the same data structure. The resulting mismatch between program and environment often leads to execution slowdown and resource under-utilization. The problem is exacerbated as chip multi-processors are becoming commonplace and most user programs are still sequential, increasingly composed with library code and running with interpreters and virtual machines. The ultimate goal of my research is an intelligent programming system, which injects into a program the ability to automatically adapt and evolve its code and data and configure its running environment in order to achieve a better match between the (improved) program, its input, and the environment. Program adaptation is not possible without accurately forecasting a program’s behavior. However, traditional modular program design and analysis are ill-fitted for finding large-scale composite patterns in increasingly complicated code, dynamically allocated data, and multi-layered execution environments (e.g. interpreters, virtual machines, operating systems and computer architecture.) My research views a program as a composition of large-scale behavior patterns, each of which may span a large number of loops and procedures statically and billions of instructions dynamically. I apply statistical technology to automatically recognize the patterns, build models of program