The GLPK package is intended for solving linear
programming (LP) and mixed integer linear
programming (MIP) problems. It is a set of
routines organized in the form of a library and
written in the ANSI C programming language.
Information regarding Project Releases and Project Resources. Note that the information here is a quote from Freecode.com page, and the downloads themselves may not be hosted on OSDN.
A new edition of the GLPK's user guide was included, describing all additional API routines which were recently added. The package structure was re-organized in order to make its maintenance easier (these package structure changes are transparent to the user).
A new, more efficient implementation of the two-phase primal simplex method. The new LP solver is faster and numerically more stable than the old one. The new LP solver is used by default to solve an initial LP problem in the branch-and-bound routine glp_call_bbm1 instead of rsm1_driver.
The relative tolerance used on the first pass of the two-pass ratio test was replaced by the absolute tolerance. The absolute tolerance passed to the two-pass ratio test routine was decreased (for both primal and dual simplex). Both changes were made in order to improve numerical stability of the simplex method. The default form of the inverse was changed from RFI to AFI.
EFI was renamed to PFI. Routines implementing AFI (Advanced Form of the Inverse) based on FHV-Factorization were added to the package. A new version of the basis maintaining routines was implemented. These perform the main operations on the basis matrix faster, at the expense of some numerical accuracy.
New API routines for scaling, writing problems in MPS format, a driver in the simplex method, and new routines for basis maintaining, inclusion of the "Modeling language GLPK/L" document into the distribution in texinfo format, and a new edition of the User's Guide. GLPSOL is able to process model descriptions in the GLPK/L language and to solve pure LP problems using the interior point method (hence, GLPIPM was removed). MIP was reprogrammed to improve its performance and robustness.
