AUTOMATIC PROGRAM ANALYSIS.
CALIFORNIA UNIV LOS ANGELES SCHOOL OF ENGINEERING AND APPLIED SCIENCE
Pagination or Media Count:
The report presents a graph-model of computer programs and then uses this model to develop tools for analysis of programs executing on single and multiprocessors. The tools for analysis include a new method of simplifying program structures through triangularization of the Boolean connection matrix representing the program, integration of execution-time measurement statistics, and post execution display of detailed analysis of critical program segments those which constitute the major portion of the program both in terms of frequency of execution and execution time. Once the program has been modeled for a single processor, analysis is performed to detect parallelism inherent in the program. A graphical display of potential parallelism, together with detected bottlenecks is presented for further analysis by the programmer. The entire system is designed to be interactive and to act as a guide for the programmer in optimizing his program, whether on a single or multi-processor. Examples of several significant programs including numerical weather prediction, circuit design and Boolean minimization, are completely analyzed, starting from a working program for each written in the FORTRAN language. Author
- Computer Programming and Software