

LSSOL
(invented by Philip Gill, Walter
Murray, Michael Saunders and
Margaret Wright) is a software
package for solving constrained
linear leastsquares problems
and convex quadratic programs
(definite or semi definite),
including linear programs. Dense
matrices are assumed throughout
LSSOL is recommended for QP
problems whose objective










includes
a term of the form x'A'Ax
for some matrix A (which may
be rectangular, square or
triangular).
Linear
constraints and bounds on
the variables are treated
separately by an activeset
method. If the problem has
no feasible solution, LSSOL
minimizes the sum of the constraint
and bound violations.








A twophase (primal)
quadratic programming
method is used, with
features to exploit
the convexity of the
objective function.
LSSOL may also be used
for linear programming
and to find a feasible
point with respect to
a set of linear equality
and inequality constraints.
LSSOL treats all matrices
as dense, and hence
is not intended for
large sparse problems.










LSSOL
requires the quadratic
to be positive definite
or semi definite
On the other hand QPOPT
can be used for general
QP problems (but may
find just a local minimum).
LSSOL is contained in
the nonlinear programming
package NPSOL.






