An Analyzer for Program Parallelization and Optimization

Abstract

The article describes new facilities for program optimization and parallelization, work-in-progress modifications of previously implemented program transformations and compiler libraries, and future development of Optimizing parallelizing system (OPS) including opening its source code on GitHub. These new facilities, such as dialog-based optimization and parallelization, user-friendly program dependency visualization (which is needed for high-quality analyzers), parallel code generation for accelerators (GPUs, DSPs, FPGAs, or high performance clusters), are made possible by the fact, that OPS uses high-level intermediate representation as opposed to low-level intermediate representation used in popular compilers.