The Multistart Coordinate Search Filter (MCSFilter) is an optimization method suitable to find all minimizers of a non convex problem, with any type of constraints. When used in industrial contexts, execution time may be critical, in order to keep production processes within safe and
expected bounds. One way to increase performance is through parallelization. In this work, a second parallel version of the MCSFilter method is presented, aiming at faster execution times than a previous parallel implementation. The new solver was tested with a set of fourteen problems, with different characteristics and behavior. The results obtained represent an improvement of the execution times over all previous MCSFilter implementations (sequential and parallel). They also allowed to identify bottlenecks to be lifted in future parallel versions.
The Multistart Coordinate Search Filter (MCSFilter) is an optimization method suitable to find all minimizers of a non convex problem, with any type of constraints. When used in industrial contexts, execution time may be critical, in order to keep production processes within safe and expected bounds. One way to increase performance is through parallelization. In this work, a second parallel version of the MCSFilter method is presented, aiming at faster execution times than a previous parallel implementation. The new solver was tested with a set of fourteen problems, with different characteristics and behavior. The results obtained represent an improvement of the execution times over all previous MCSFilter implementations (sequential and parallel). They also allowed to identify bottlenecks to be lifted in future parallel versions.
This work has been supported by FCT - Funda¸c˜ao para a Ciˆencia e Tecnologia within the Project Scope: UIDB/05757/2020.
