Abstract

Fourier-Motzkin elimination is a computationally expensive but powerful method to solve a system of linear inequalities for real and integer solution spaces. Because it yields an explicit representation of the solution set, in contrast to other methods such as Simplex, one may, in some cases, take its longer run time into account. We show in this paper that it is possible to considerably speed up Fourier-Motzkin elimination by massively parallel processing. We present a parallel implementation for a shared memory parallel computer, and sketch several variants for distributed memory parallelization.

Citations