primal-dual interior-point code for linear programming
PCx is a freely available primal-dual interior-point code for linear
programming. It implements Mehrotra's predictor-corrector algorithm,
the algorithm that forms the basis of most existing interior-point
codes for general linear programming. The major computational
operation--solution of a linear system with a large, sparse positive
definite coefficient matrix--is performed with the sparse Cholesky
package of Ng and Peyton (Oak Ridge National Laboratory), with minor
modifications to handle small pivot elements. PCx does not solve
integer programming problems.