Software

Freely available optimization software

Iterative solvers for sparse Ax = b:
SYMMLQ, MINRES, MINRES-QLP, cgLanczos, CRAIG

Iterative solvers for sparse least-squares problems:
LSQR, LSMR, LSLQ, LNLQ, CGLS, covLSQR, LSRN

Sparse and dense LU factorization (direct methods):
LUSOL, LUMOD

Nullspace of sparse matrix via SPQR or LUSOL:
lusolZ

Sparse optimization:
ASP

Optimization with convex objective and linear constraints:
PDCO (including sparse optimization)

Convex optimization in composite form:
PNOPT

Fortran 90 quad-precision dotproduct of double-precision vectors
qdotdd

Licensed optimization software

See MINOS, LSSOL, NPSOL, QPOPT, SQOPT, SNOPT for user's guides to each of those solvers.

These packages are freely available to faculty and students at Stanford University and UC San Diego.
For MINOS, please email Michael Saunders at saunders@stanford.edu.
For SNOPT, DNOPT, NPSOL, SQOPT, SQIC, please see UCSD/Stanford Optimization Software and email Philip Gill and Elizabeth Wong at optimizers@ccom.ucsd.edu.

For non-Stanford and non-UCSD personnel, please contact SBSI regarding three types of license for the above solvers:

Tier 1:	Academic (and Nonprofit)	SBSI
Tier 2:	Nonacademic individual and site	SBSI
Tier 3:	Commercial redistribution	OTL (Evan Elder)

MINOS 5.6 is available in both Double-precision and Quad-precision versions. If you need to solve linear or nonlinear optimization problems in Quad precision, please email Michael Saunders at saunders@stanford.edu. Input can be MPS files (which have limited data precision), or you can call subroutine minoss from an f90 program with floating-point variables declared real(16).
MINOS 5.6 Double and Quad is available for large-scale linear programs within the GAMS modeling system. Please see the DQQ (Double-Quad-Quad) procedure within the GAMS library: dqq.gms : Warm-starting quad-precision MINOS.

TOMLAB

TOMLAB is a MATLAB environment for optimization. It provides an interface to MINOS, LSSOL, NPSOL, SQOPT, SNOPT, and many other solvers for many different problem types.

TOMLAB at Stanford

TOMLAB is installed on the Stanford Linux cluster (March 31, 2011).
To gain access, login to the cluster (e.g. ssh corn) and proceed as follows:

    module load matlab
    matlab
    cd /afs/ir/software/matlab-2010b/tomlab
    startup
    cd examples
    democon

The following solvers are licensed: Tomlab/Base, SOL, CGO, CPLEX, MAD