• Requirements:
  • The software is written in Fortran-2003.
  • It uses QPAL, as a convex quadratic problem (QP) solver. This one combines an augmented Lagrangian approach, active set, and conjugate gradient iterations
  
  
• Implemented features:
  • the local algorithms (equality and inequality constraints),
  • use a BFGS or an l-BFGS (direct or inverse) approximation of the Hessian of the Lagrangian (its positive definiteness is maintained by the Wolfe line-search when there is no constraint and by the Powell correction otherwise),
  • the constraint Jacobian matrices can be stored in dense or sparse structures,
  • globalization by linesearch, using a merit function.