• Version 0.6: in development
    1. Cosmetic changes.

  • Version 0.5: May 22, 2009
    1. Elementary truncated version: the QP's can be solved approximately with a precision depending on the one reached on the nonlinear problem at the current iteration.
    2. The following options have been added: qp_max_alit, qp_max_cgit, qp_max_hvpd, qp_max_avpd, qp_norm, qp_precision_id, qp_tol_glan, qp_tol_feas, qp_forcing_factor.
    3. Use of the QP solver QPAL (version 0.6.1).

  • Version 0.4: May 13, 2009
    1. Sparse version: Hessian of the Lagrangian and constraint Jacobian matrices can be stored as sparse (or dense) matrices.
    2. The code has been largely restructured.
    3. Use of the QP solver QPAL (version 0.6).

    These modifications imply many changes for the user, who should read again the documentation.

  • Version 0.3: October 2008
    1. For unconstrained problems and l-BFGS updates, SQPpro now generates an approximation of the inverse of the Hessian, which makes it much more efficient wrt the computing time. See the performance profiles.
    2. Default options are now set in sqppro_init, so that the user has only to set options with non-default values, before calling sqppro_solve. The options can now also be given by a specification file, named sqppro.spc in the working directory, which provides other tunings than those in the options.
    3. The argument hessian_id of the subroutine sqppro_init has been replaced by htype, with a different meaning.
    4. Use of of the QP solver QPAL (version 0.5).
    5. Bug corrected, cleaning.

    Changes for the user

    1. The arguments of sqppro_init have been modified in quantity and nature.
    2. The options set after having called sqppro_init whose values coincide with the default values need no longer be set. There is now also the possibility to set the options through a specification file, named sqppro.spc in the working directory; the values specified in that file prevail on those hard-coded in the solver and the calling procedure.

  • Version 0.2: August 2008
    1. Possibility to update the Hessian of the Lagrangian by the l-BFGS technique.
    2. The data is no longer stored in a single structure, which used to be largely fixed at compile time in a module (previously in the la_mod module). It is now a structure passed in argument of SQPpro. As a result, the solver can now work on different problems (data and sizes) in the same program. The data structure is allocated at run time.
    3. Use of of the QP solver QPAL (version 0.4).

  • Version 0.1: September 2005
    1. Initial version.