Summary: This chapter is devoted to an important requirement of successful model order reduction (MOR) application, namely, the software aspect. The most common situation is the existence of a so-called full model, i.e., a high-fidelity, high-dimensional simulation model, that needs to be accelerated by MOR techniques, optimally without reimplementing the partially complex reduction techniques, as presented in the first volume of this handbook.
Initially, as neither full simulation models nor MOR algorithms are to be reprogrammed, but ideally are reused from existing implementations, we concentrate on the aspect of the interplay of such packages. We will discriminate, discuss, and exemplify different levels of solver “intrusiveness” that allow corresponding reduction techniques to be applied. On the one hand, most effective MOR techniques require deep access into the full model’s simulation code. On the other hand, application-specific full model simulators may only offer very restricted access to internals, especially in case of commercial packages. This gap in requirements and practical accessibility motivates the discrimination into “white-box,” “gray-box,” and “black-box” simulation scenarios. In particular, we exemplify the ideal case of MOR for white-box situations on two examples, namely, parametric linear elliptic PDE and parametric nonlinear ODE systems. Depending on those access classes, different corresponding reduction techniques can be applied.
The second part of the current chapter then discusses existing MOR software. Several program packages exist which provide MOR techniques. They differ in availability, licensing, programming language, system types, physical application domains, external simulator bindings, etc. We give an overview of the most relevant of those MOR packages, such that applicants can identify potential suitable software library candidates.
For the entire collection see [Zbl 1455.93002].