Section: New Software and Platforms

Oqla, Qpalm

Participants : Jean Charles Gilbert, Émilie Joannopoulos.

Oqla and Qpalm aim at minimizing a large scale convex quadratic function on a polyhedron by an augmented Lagrangian method. The original features of the approach are its capacity to solve the problem without factorization, which makes them adapted to large scale problems, and to deal with unbounded and infeasible problems. In case the problem is infeasible, the codes solve the closest feasible problem with a global linear rate of convergence [3] . In case the problem is unbounded, the solvers build a feasible direction of unboundedness for the closest feasible problem. The solvers Oqla and Qpalm only differ by their programming language; they are documented in [16] , [14] , [15] .

• Versions: 0.6 (Oqla ), 0.5 (Qpalm )

• Programming languages: C++ (Oqla ), Matlab (Qpalm )