(700c) Model Generation: Massive Parallel Computations for the Next Generation of Biomedical Engineering Applications

Distributed systems models occur in life science applications such as drug delivery, intracranial dynamics or whole-body pharmacokinetics. These models lead to very large computational meshes due to the geometrical complexity of the problem domain typically obtained from medical images as well as the need for dense computational domains to proof mesh-independence in the spatio-temporal discretization scheme which involves multiple time and length scales.

Due to the huge problem sizes, a massive parallel approach for the solution of the patial differential equation systems is often required. In this presentation, we emphasize model generation as an approach to seamlessly port distributed application between various hardware solution architectures. Instead of insisting on solving partial differential equations in a single monolithic software tool, model generation synthesizes a set of algebraic equations in desired target languages or syntax. Especially large scale parallel processing require low level user-supplied residual functions and derivatives; thus the novel model generation paradigm serves as an important bridge enabling consistent specification, and solution of large scale application required in biomedical engineering. The presentation will demonstrate the state-of-the-art model generation tools for the solutions of mass and momentum transfer problems in drug delivery and simulation of intracranial dynamics.