Computer solution for sparse systems of linear equations by symbolic generation

This paper presents a method of computer solution for sparse systems of linear equations of arbitrary structure by symbolicgeneration technique. A computer program, SOURCE, by symbolic processing, generates another program, RESULT, which represents the reduced algorithm. The symbolic generation method used were based on Crout and Gauss-Seidel methods, only non-zero elements are stored and operated on. Because of the increasing use of large order sparse matrices and the tendency to attempt to solve larger order problems, great attention must be focused on core allocation and execution time, as these are the limiting factors that most often dictate the practicality of solving a given problem. Computer programs were written in FORTRAN V language to implement these methods.