Automated analysis of scientific and engineering semantics

Physical and mathematical formulae and concepts are fundamental elements of scientific and engineering software. These classical equations and methods are time tested, universally accepted, and relatively unambiguous. The existence of this classical ontology suggests an ideal problem for program comprehension: automated scientific code semantic analysis. This problem is motivated by the pervasive use of scientific code and the high costs of code development and maintenance. To investigate code comprehension in this classical knowledge domain, a research prototype has bean developed. The prototype incorporates scientific domain knowledge to recognize code properties (including units formulae, physical and mathematical equations). Also, the procedure emulates program execution to propagate these symbolic properties through the code. This prototype´s ability to elucidate code and detect errors is demonstrated with state of the art scientific codes