A Data-flow Framework for Analyzing Assembly Code

Peter Braun, Andrej Rosenheinrich, Wolfram Amme, Eberhard Zehendner

This paper presents a data-flow framework for analyzing assembly code. The framework provides information needed for various optimizations, e.g. reaching definition, symbolic value sets, and alias information. Our framework is built on top of SALTO, which is a tool for assembly code transformation for various types of processors. Some applications of our technique are discussed. Special emphasis is laid on the determination of aliases between memory references. We show first results in the field of redundant load and store elimination.