This report investigates alternative methods of compactly storing the coefficient matrix of a linear programming problem. Details of the data to be stored, including column headers, are given together with descriptions of sparse and super-sparse schemes of storage. Particular emphasis is given to the concept of machine independence in the implementation of these schemes with a view to the development of all-in-core LP systems. Ten representative problems of widely varying sizes and a selection of computers with different word and character lengths are used to compare the efficiency of the storage schemes. It is shown that the methods based on super-sparsity usually lead to large savings in storage requirements.
ACM SIGMAP Bulletin – Association for Computing Machinery
Published: Apr 1, 1983