NSF has awarded a CNS-CSR grant on Behavior-Based Speculative Parallelization and Optimization on Desktop Multiprocessors to Prof. Xipeng Shen.
The 3-year project, collaborated with University of Rochester, is motivated by the remarkable difficulty for software to utilize the fast growing parallelism in modern chip multiprocessors. The goal is to develop adaptive and efficient software speculation techniques to help users effectively parallelize complex programs without the need of parallel programming or debugging. Unlike traditional code-based techniques, the new approach is based on program behavior analysis, allowing a user or a profiling tool to parallelize or optimize a program based on partial information about the program code and the input. It uses modern parallel processors to reduce and hide the overhead of dynamic correctness checking and error recovery.