Home

This documentation is based on the user manual for MrHyDE that can be found under MrHyDE/doc.

MrHyDE is a Trilinos-based framework design for solving Multi-resolution Hybridized Differential Equations, thus the name MrHyDE. This state-of-the-art software framework combines lightweight interfaces to select Trilinos second-generation packages and a set custom performance portable managers to enable the solution of transient nonlinear strongly coupled multiphysics/multiscale problems with an emphasis on beyond forward simulations (BFS) capabilities e.g., large-scale PDE-constrained optimization, and basic sampling-based capabilities for uncertainty quantification, and measure-theoretic stochastic inversion. While this document is primarily aimed at new users of Trilinos/MrHyDE, there are several detailed explanations of certain components of MrHyDE that should be valuable to all users/developers.

MrHyDE does not come with a graphical user interface (GUI) and it is expected that the user/developer will be sufficiently fluent in certain high-performance computing (HPC) tools, such as CMake and C++ compilers, to compile both Trilinos and MrHyDE. A python interface is currently under development, but the functionality will be limited. MrHyDE can be packaged into a Docker or Singularity container for rapid deployment to novice or external users. The target audience for MrHyDE is computational scientists looking for a scalable simulation framework that is modular, easy to modify, portable from laptops to exascale, and automatically enables BFS and multiscale capabilities.

Key Features:

• Enables rapid prototyping of complex multiphysics and multiscale systems;
• Easy to scale up problems and provides scalable performance on modern heterogeneous computational architectures;
• Robust user interface that allows a user to modify an existing set of equations directly from the input file in arbitrarily complex ways;
• Automated adjoint and multiscale capabilities;
• Interface to the Rapid Optimization Library (ROL) for large-scale PDE-constrained optimization;
• A basic data-consistent inversion capability that provides an introduction to stochastic inversion;
• A data integration capability for incorporating field and scalar data;
• In situ data compression to optimize memory usage;

To learn more and to start using MrHyDE, please see the Getting Started page.