AbstractsComputer Science

Compilers are among the most widely-studied pieces of software; and, modularizing these valuable artifacts is a recurring theme in research. However, modularization of cross-cutting concerns in compilers is not yet well explored. Even today, implementation of one compiler concern scatters across and tangles with the implementation of several other concerns, thereby leading to a mismatch between different compiler modules and the operations they represent. Essentially, current compiler implementations fail to explicitly identify the control dependencies of different phases, and separately characterize the actions to execute during those phases. As a result, information about their program-execution path remains non-intuitive: it stays hidden within the program structure and cuts-across several phase implementations. Consequently, this makes compiler designs and artifacts difficult to comprehend, maintain and reuse. Such limitations occur primarily as a result of the inability of mainstream object-oriented languages, such as Java, to organize the cross-cutting concerns into clean modular units. This thesis demonstrates how such modularity-issues in compilers can be addressed with the help of a relatively new, yet powerful programming paradigm called aspect-oriented programming.