An open-source implementation of the Dantzig-Wolfe decomposition method, a powerful and useful technique for solving linear programming problems of block-angular form, is described in this concise and well-written paper.

The author implemented a parallel version of the algorithm using the POSIX threads (pthreads) library, and performed comparisons with CPLEX for randomly generated instances and an air traffic management problem. The results demonstrate an improvement of two to nine times (depending upon the number of subproblems) using the proposed parallel implementation. The source code is freely available under a GPLv3 license [1].