dylp is an implementation of the dynamic simplex algorithm. Briefly, dynamic simplex attempts to work with an active constraint system which is a subset of the full constraint system. It alternates between primal and dual simplex phases. Between simplex phases, it deactivates variables and constraints which are not currently useful, and scans the full constraint system to activate variables and constraints which have become useful. The dylp documentation describes dylp’s algorithms in detail. dylp is intended as a development code. The implementation puts more weight on clarity than efficiency and is heavily instrumented and commented. It can be built with extensive statistics, informational printing, and consistency checks. It can also be built without any of these. Fully optimised, it’s decently fast, with run times averaging three to four times that of clp, depending on the problem and the runtime environment. dylp is a C code, and can be used in a pure C environment completely independent of other COIN software. It also has a COIN OSI interface, OsiDylp, which takes advantage of capabilities provided by COIN (e.g., enhanced input/output and constraint system preprocessing) and is recommended if you’re working in a C++ environment.
Keywords for this software
References in zbMATH (referenced in 1 article )
Showing result 1 of 1.
- Gassmann, Horand; Ma, Jun; Martin, Kipp: Communication protocols for options and results in a distributed optimization environment (2016)