Modules to Aid the Implementation of LP Algorithms.

摘要

Implementing large scale LP algorithms is a difficult, laborious and often poorly rewarded task. This is particularly true of algorithms which exploit the structure of the LP matrix. For this reason many algorithms have been proposed in the literature, but few have been turned into good computer codes. Very little is known about the relative performance of different algorithms. In this paper we discuss some of the suggestions that have been made for alleviating this problem and describe an approach based upon a carefully defined collection of subroutines which are designed to aid the task for implementing and comparing LP algorithms. These subroutines or modules may be regarded as the 'primitives' of a language of implementing experimental LP algorithms, particularly algorithms which exploit matrix strucutre. A set of such modules is described. These have been implemented in FORTRAN, and user documentation is available. (Author)