Adjoint Code Design Patterns

Adjoint methods have become fundamental ingredients of the scientific computing toolbox over the past decades. Large-scale parameter sensitivity analysis, uncertainty quantification and nonlinear optimization as well as modern approaches to deep learning would otherwise turn out computationally infeasible. For nontrivial real-world problems the algorithmic derivation of adjoint numerical simulation programs quickly becomes a highly complex software engineering task requiring expertise in software analysis, transformation and optimization. Despite rather mature software tool support for algorithmic differentiation substantial user intervention is typically required. A large number of patterns shared by numerous application codes results in repeated duplication of development effort. The adjoint code design patterns discussed in this talk based on [8] aim to reduce this problem through improved formalization from the software engineering perspective.