A Source Code Analyzer to Predict Compilation Time for Avionics Software Using Software Science Measures. Volume 1. Main Report.
AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH SCHOOL OF ENGINEERING
Pagination or Media Count:
This thesis describes the construction of an Ada source code analyzer SCA which produces values for the Software Science measures 1c1, 1c2, N1, and N2. The measures are used to evaluate a mathematical model designed to predict the compile time of Ada modules. The primary goal of this effort was to provide a software tool to metrics researchers which could automatically compute Software Science measures for Ada modules. A secondary goal was to produce a convenient method for Ada compiler researchers to predict the amount of time consumed during compilation of given avionics software modules. As the SCA was built, we incorporated the rules of a new Ada token counting strategy designed to yield meaningful results for entire Ada programs, not just executable code. Once satisfied the SCA implemented the rules correctly and produced accurate counts for the Software Science measures, we added the compile time model to the SCA. To test the validity of the compile time model, over 200 modules were selected at random from among the Common Ada Missile Packages CAMP software library. For each module chosen, both the compile time as predicted by the SCA and the actual compile time using the Verdix Ada compiler were recorded. Finally, the prediction error values predicted compile time minus actual compile time were recorded and analyzed.
- Computer Programming and Software