Large-Scale LP/MIP Solver DLL

• Solves LP Problems with Thousands of Variables
• Sparse Methods Save Megabytes of Memory
• Matrix Factorization for Speed and Accuracy
• Automatic Scaling, Basis Restart, Crashing
• Callable from Visual C++, Visual Basic, Access Basic

Frontline's Large-Scale LP/MIP Solver DLL is a state of the art implementation of linear programming and integer programming solution algorithms, offered in the convenient form of a Dynamic Link Library usable by any application under Microsoft Windows. It is available as a 16-bit DLL (for 16-bit applications run under Windows 3.1, Windows 95 or Windows NT) and as a 32-bit DLL for native Win32 applications. For even larger problems, users can easily move their Win32 applications (written in supported languages such as C/C++) to the CPLEX Callable Library on PCs or UNIX workstations, which also uses compatible calling conventions.

The Large-Scale LP/MIP Solver efficiently stores and processes LP models in sparse matrix form. It uses modern matrix factorization methods to control numerical stability during the solution of large LP models, where the cumulative effect of the small errors inherent in finite precision computer arithmetic would otherwise compromise the solution process. These same methods often result in solution speeds far beyond earlier generations of LP solvers. Thanks to these methods, the Large-Scale LP/MIP Solver can readily handle all of the problems in the well-known NETLIB test suite -- virtually all with the default tolerance settings "out of the box."

The Large-Scale LP/MIP Solver is also a "well-behaved" Microsoft Windows Dynamic Link Library. It respects Windows conventions for loading and sharing program code, and for sharing the processor with other applications through Windows multitasking (including non-preemptive multitasking in Windows 3.1). It solves the problems of Windows 3.1 memory management through use of a "two-level" storage allocator, which draws upon Windows global memory resources in an efficient manner. These issues are handled in a way which is transparent to the application program. Callback functions (for supported languages such as C/C++) allow you to display progress information and accept user actions during the solution process.

Algorithmic Features

• Simplex method with two-sided bounds on both variables and constraints
• Memory-efficient Branch & Bound method for mixed-integer problems
• Matrix factorization using the LU decomposition, with the Bartels-Golub update
• Refactorization using dynamic Markowitz methods for speed and stability
• Steepest-edge techniques often substantially reduce the number of pivots
• Sophisticated "crash-like" methods for finding an effective initial basis
• Basis restart for faster solution of related LP problems (including MIP subproblems)
• Automatic row and column scaling, plus control of various algorithm tolerances

The Large-Scale LP/MIP Solver DLL is priced at $1,495. Volume discounts are available. For more information please call Frontline Systems at (702) 831-0300, fax to (702) 831-0314, or send email to sales@frontsys.com.

Copyright © 1996 Frontline Systems Inc.
Last modified: December 01, 1996