Practical Compile-Time Analysis Algorithms: (FAW)

点击翻译按钮获取中文摘要

Compile-time analysis techniques to aid the development of software in sequential and parallel programming environments will be the focus of this research. There are two trusts: (i) the maturation of work in designing incremental data flow algorithms for semantic impact analysis of source code changes in software systems, and (ii) the investigation of how compile-time analysis can aid the compilation and debugging of programs that have implicit and/or explicit parallelism. In debugging of concurrent systems, compile-time analysis is used to help programmers debug software written with explicit parallelism. Typical problems include races in shared memory machines, indeterminately ordered accesses to the same memory location where at least one access is a write, and deadlocks, non-terminating computations which cannot continue because of improper synchronizations. For detection on shared memory machines, use of interprocedural data flow information and good internal representations of parallel primitives are being explored, as well as a case study analyzing Ada rendezvous mechanisms. It is intended to combine compile-time analysis with run-time testing for these problems, using the former to guide the latter. This work will culminate in development of prototype software tools demonstrating the techniques on actual Ada and Fortran programs. Desired extensions include studying the applicability to a broad range of synchronization primitives and classifying their properties with respect to these problems.