Special note regarding convex QP models. Prior to version 5.00, QP models were automatically assumed to be convex. This is no longer true: QP's are assumed to be nonconvex (and hence are treated differently by the algorithm) unless otherwise specified. To specify that a problem is convex, just include the keyword "convex" in the loqo_options string. We strongly suggest that this convexity assertion be explicitly made for any NLP that is known to be convex.
Clicking on the appropriate platform, you will get the executable (with built-in AMPL interface) and the subroutine library.
Perform the following in the order given:
Version 5.05
Version 5.06
Version 6.00
Version 6.03
Version 6.06
Version 6.07