Trilinos

Trilinos
Name	Trilinos
Developer	Sandia National Laboratories
Released	2000
Programming language	C++, Fortran, Python
Operating system	Linux, macOS, Windows
License	BSD-style

Trilinos is an open-source collection of software libraries for the solution of large-scale, complex multi-physics and engineering problems. It provides a modular framework for parallel computing, numerical linear algebra, and iterative methods used in high-performance computing environments such as national laboratories and research universities. The project emphasizes interoperable packages for partial differential equations, optimization, and uncertainty quantification and integrates with prominent supercomputing centers and vendor ecosystems.

History

Trilinos originated at Sandia National Laboratories in the late 1990s as part of efforts tied to national programs and collaborations with institutions like Lawrence Livermore National Laboratory, Los Alamos National Laboratory, and academic partners such as Massachusetts Institute of Technology and University of California, Berkeley. Early development was influenced by research from groups associated with the U.S. Department of Energy and initiatives including the Advanced Simulation and Computing Program and various Scientific Discovery through Advanced Computing projects. The project matured through multiple releases, responding to the growth of MPI-based supercomputers at facilities like Oak Ridge National Laboratory and Argonne National Laboratory, and expansions driven by collaborations with industrial partners including Intel and NVIDIA. Over time, Trilinos evolved alongside standards such as C++98, C++11, and integration efforts with language bindings used by teams at institutions like Stanford University and California Institute of Technology.

Architecture and Components

The Trilinos ecosystem is organized as modular packages that implement functionality across domains familiar to researchers at Princeton University, Cornell University, and University of Texas at Austin. Core packages provide abstractions for distributed linear algebra influenced by algorithms from groups at ETH Zurich and RIKEN, while solver packages expose interfaces for iterative methods developed in cooperation with researchers from University of Cambridge and Imperial College London. Notable components include solver interfaces interoperable with libraries such as PETSc and preconditioners following strategies used by teams at University of Illinois Urbana-Champaign and University of Michigan. The software supports integration with visualization and workflow tools from Kitware, scientific formats adopted by Oak Ridge National Laboratory, and build systems compatible with tools used at National Renewable Energy Laboratory.

Development and Governance

Development is coordinated through a governance model centered at Sandia National Laboratories with contribution policies similar to large open-source projects at organizations like The Linux Foundation and Apache Software Foundation. The project uses collaborative platforms and continuous integration systems akin to infrastructure employed by GitHub-hosted communities and research groups at Lawrence Berkeley National Laboratory. Technical steering and release processes involve contributors from Los Alamos National Laboratory, industrial partners such as Cray (now part of Hewlett Packard Enterprise), and academic consortia including researchers from University of California, San Diego and University of Colorado Boulder. Licensing and export controls reflect policies coordinated with U.S. Department of Energy offices and procurement practices observed at National Laboratories.

Applications and Use Cases

Trilinos is applied in multi-physics simulations used by teams engaged with projects at NASA centers, computational fluid dynamics groups at General Electric, and structural analysis groups at Boeing. Researchers at Argonne National Laboratory and Sandia National Laboratories employ Trilinos for large-scale simulations in seismology research collaborations with United States Geological Survey and in nuclear engineering studies linked to Idaho National Laboratory. The framework supports optimization workflows used in energy systems modeling with partners like Pacific Northwest National Laboratory and uncertainty quantification pipelines used in climate research by groups at National Center for Atmospheric Research. Domain scientists at MIT Lincoln Laboratory and Lawrence Livermore National Laboratory integrate Trilinos with codes addressing electromagnetics, material science, and reservoir simulation.

Performance and Scalability

Trilinos targets high scalability on architectures deployed at facilities such as Oak Ridge Leadership Computing Facility and Argonne Leadership Computing Facility. Performance engineering draws on parallel programming models used at Intel and NVIDIA and optimization strategies developed with vendors like AMD and system integrators such as HPE. Benchmarks reported by teams at Sandia National Laboratories and collaborators from University of Tennessee show strong scaling for sparse linear algebra and multigrid preconditioning on up to tens of thousands of processors, while efforts with groups at Lawrence Berkeley National Laboratory investigate GPU-accelerated kernels and hybrid MPI+OpenMP strategies. Profiling and tuning often leverage tools produced by Perforce partners and analysis frameworks used by performance engineering teams at Cray.

Reception and Adoption

Trilinos has been widely adopted by national laboratories including Sandia National Laboratories, Los Alamos National Laboratory, and Lawrence Livermore National Laboratory, as well as academic research groups at Stanford University and University of Illinois Urbana-Champaign. Industry users include teams at General Electric and defense research organizations collaborating with MITRE Corporation. The project is cited in scientific literature alongside other foundational infrastructures such as PETSc, and its interoperability has been praised in workshops run by SIAM and conferences like SC (conference) and ICCS. Community feedback from users at Oak Ridge National Laboratory and Argonne National Laboratory has shaped feature priorities and contributed to an ecosystem of tutorials and training materials maintained by contributors from Sandia National Laboratories and partner universities.

Category:Numerical software