FUTURE COMPILATION REQUIREMENTS FOR EMERGING DRIVING GENERAL PURPOSE APPLICATIONS BY

As industry moves from single processor systems to chip multiprocessors in the general purpose community, it is becoming increasingly important for research to help enable this transition by developing tools that assist programmers in developing applications for these sys-tems. Compilers will play an important role in this transition. There has been a wealth of past research developing compilation tools to enable high-performance computing, which has utilized multiprocessor systems for years. However, there is a gap between these applications and the general purpose “driving ” applications of the future. This thesis will provide an evalu-ation of four representative benchmarks and provide insights into what new research needs to be completed in order to extend past work to these future applications. Much of the research in parallelizing compilers has focused on scientific applications that are rich in scalable inner loop parallelism. Many of these applications are written using Fortran and perform repetitive calculations on large arrays of data. While C implementations exist, these applications do not take advantage of many of the features available in C. They are similar in structure to their Fortran counterparts and therefore do not suffer the same complications a