Cactus Framework

Cactus Framework
Name	Cactus Framework
Developer	Einstein Toolkit Consortium
Initial release	1995
Latest release	2020s
Programming language	C, Fortran, C++
Operating system	Unix-like
License	GPL-compatible

Cactus Framework The Cactus Framework is a modular, component-based software framework primarily used for high-performance scientific computing, particularly in numerical relativity and computational astrophysics. It serves as a glue layer connecting simulation components from projects such as the Einstein Toolkit, Carpet mesh-refinement, Whisky hydrodynamics, and analysis tools developed by teams at institutions like CITA, Max Planck Institute for Gravitational Physics, Perimeter Institute, and Caltech. The framework integrates with supercomputing centers including Oak Ridge National Laboratory, Argonne National Laboratory, and NERSC to run large-scale simulations that inform experiments at facilities like LIGO and observations by telescopes such as Hubble Space Telescope and Vera C. Rubin Observatory.

Overview

Cactus provides a component model that separates core infrastructure from scientific modules, enabling collaboration across groups at University of Illinois Urbana-Champaign, University of Southampton, University of Texas at Austin, University of Glasgow, and University of Pisa. It interoperates with libraries like MPI, HDF5, PETSc, Hypre, and BLAS to leverage hardware at centers including Texas Advanced Computing Center and National Energy Research Scientific Computing Center. Funding and coordination often involve agencies such as the National Science Foundation, European Research Council, Deutsche Forschungsgemeinschaft, and national labs like Lawrence Berkeley National Laboratory.

History

Cactus originated in the mid-1990s from collaborations involving researchers associated with Albert Einstein Institute, AEI Potsdam, University of Southampton, and Louisiana State University as part of efforts to simulate compact-object mergers relevant to the Laser Interferometer Gravitational-Wave Observatory search. Early adopters included groups working on projects connected to Binary Black Hole Grand Challenge, Numerical Relativity Network, and researchers from University of Illinois. Over time Cactus integrated work from initiatives like the Simulating eXtreme Spacetimes collaboration and incorporated contributions from institutions such as Syracuse University, University of Maryland, University of Nottingham, and University of Wisconsin–Milwaukee. The framework evolved alongside community software such as FLASH, Enzo, Gadget-2, and AMReX while influencing projects at National Center for Supercomputing Applications and Lawrence Livermore National Laboratory.

Architecture

Cactus uses a core called the "flesh" and plug-in modules called "thorns", following an architecture similar in modular intent to Apache HTTP Server modules and GNU Compiler Collection backends. Thorns implement physics solvers, I/O, boundary conditions, and infrastructure linking to packages like Carpet for adaptive mesh refinement, CactusEinstein components for relativistic formulations, and McLachlan for finite-difference operators. The build and configuration system integrates with tools such as CMake, Autoconf, and supports parallel execution through OpenMP and MPI. Workflow and provenance are often managed alongside systems developed at Sandia National Laboratories, Los Alamos National Laboratory, and European Space Agency collaborations.

Development and Community

Development occurs through distributed version control systems used by teams at GitHub, Bitbucket, and institutional repositories at Max Planck Society and Cambridge University. Community governance draws contributors from Louisiana State University, University of Southampton, Rutherford Appleton Laboratory, Cardiff University, and University of Birmingham. Training and workshops have been held in partnership with organizations such as PRACE, XSEDE, ISC High Performance, and academic conferences including Supercomputing Conference, International Conference for High Performance Computing, Networking, Storage and Analysis, and Gravitational Wave Astronomy Workshops. Collaborative papers appear in journals like Physical Review D, Classical and Quantum Gravity, Astrophysical Journal, and Computer Physics Communications.

Use Cases and Applications

Cactus is widely used for simulations supporting observations and experiments by LIGO, VIRGO, KAGRA, and theoretical studies informing missions like James Webb Space Telescope and surveys by Sloan Digital Sky Survey. Typical applications include binary black hole mergers studied with codes related to SpEC and BAM, neutron star mergers interfacing with microphysics modules used by researchers at JINA-CEE and Max Planck Institute for Astrophysics, core-collapse supernova modeling alongside communities from Oak Ridge National Laboratory and University of Tennessee, and magnetohydrodynamics simulations relevant to European Southern Observatory instrument teams. The framework also supports methodological work in high-order methods comparable to efforts in ExaHyPE and comparisons with solvers like FLASH and PLUTO.

Performance and Scalability

Cactus is designed to scale on petascale and emerging exascale systems, having been benchmarked on machines such as Titan (supercomputer), Summit (supercomputer), SuperMUC-NG, and Fugaku. The Carpet driver enables scalable block-structured adaptive mesh refinement comparable to Chombo and AMReX, while coupling to PETSc and Hypre supports scalable linear solvers used by laboratories including Lawrence Livermore National Laboratory and Sandia National Laboratories. Performance engineering engages hardware vendors like NVIDIA, Intel, AMD, and partnerships with centers such as Cray Inc. and IBM to optimize for GPU-accelerated and many-core architectures.

Licensing and Availability

Cactus and many associated thorns are distributed under licenses compatible with GNU General Public License terms through public repositories used by communities at Einstein Toolkit Consortium, GitHub, and institutional mirrors at Max Planck Institute for Gravitational Physics and Perimeter Institute. Source code releases, documentation, and tutorials are maintained by contributors from University of Illinois, Syracuse University, University of Southampton, and allied research groups, and are available to researchers affiliated with consortia such as Einstein Toolkit and computational facilities at National Institute for Computational Sciences.

Category:Scientific computing software