Design and Implementation of a Primal Simplex Network Optimizer in C.

摘要

This thesis documents the design and implementation of an efficient primal simplex capacitated transshipment network optimizer, SNET, written in the C programming language. It describes a general symbolic network algorithm, discusses fundamental decisions regarding data structures and essential functions and their relationship to the network algorithm, and then details SNET's development. Development tools used in this project, including standard test problems, profilers, timing routines, external drivers, and debuggers, are also covered. The resulting solver, SNET, is quite fast on standard NETGEN test problems, approximately twice as fast as a primal simplex network solver written in FORTRAN. The effect of tuning parameters on SNET's performance is minimal.