WORHP 1.16 released ⌚ 2024-05-07
We are pleased to announce WORHP v1.16.
New Linux Packages
Ubuntu 24.04, Debian 12 and Fedora 40 were added, Fedora 36 was removed.
Changes
- New simple sparse factorization as linear solver option. Activated by setting
IP_LeqSolverorqp.ipLsMethodto10. On very sparse and not too large problems, this option can be much faster than the default MA97, but occasionally also less robust - Major performance improvements for sparse finite differences
- Improved initial Lagrange Multiplier estimation based on minimizing the gradient of the Lagrangian in least squares sense in the initial point. To revert to the old behavior based on an initial gradient step,
InitialLMestMethodmay be set to0 - Fix for scaling of sensitivities with automatic scaling modes (
ScaledQP,ScaledObj,IP_ScaleF,IP_ScaleG) and further accuracy improvements - Fix for thread-unsafe behavior in initialization procedures
- The application-wide locale setting is not set to the "C" locale anymore. I/O now uses locale-independent functions to avoid interfering with and relying on users' locale settings
- Many additional minor robustness and performance improvements
Patch 1.16-1 (2024-06-07)
- Fixed a bug in the AMPL interface that caused linear objectives to be ignored