HARD: A Performance Portable Radiation Hydrodynamics Code based on FleCSI Framework Journal: SoftwareX
Julien Loiseau, Hyun Lim, Andrés Yagüe López, Mammadbaghir Baghirzade, Shihab Shahriar Khan, Yoonsoo Kim, Sudarshan Neopane, Alexander Strack, Farhana Taiyebah, Ben Bergen · 2025
Read HARD: A Performance Portable Radiation Hydrodynamics Code based on FleCSI Framework Journal: SoftwareX by Julien Loiseau, Hyun Lim, Andrés Yagüe López, Mammadbaghir Baghirzade, Shihab Shahriar Khan, Yoonsoo Kim, Sudarshan Neopane, Alexander Strack, Farhana Taiyebah, Ben Bergen (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-08971.txt.
Introduction
Julien Loiseau and 9 co-authors present HARD, a performance portable radiation hydrodynamics code, in a 2025 paper. The code simulates compressible fluid flow coupled with radiation diffusion. It is built on the FleCSI framework for high performance computing.
Work is expressed as tasks that can run under several back ends, including Legion, MPI, and HPX. Node level parallelism uses Kokkos, so one code base runs from laptops to supercomputers. A regression suite checks classic shock tube and blast wave problems against known answers. The project is open source with build scripts and continuous integration, aiming at sustainable community use.
Catalog record
| CREATOR | Julien Loiseau, Hyun Lim, Andrés Yagüe López, Mammadbaghir Baghirzade, Shihab Shahriar Khan, Yoonsoo Kim, Sudarshan Neopane, Alexander Strack, Farhana Taiyebah, Ben Bergen |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 26 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-08971
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
HARD: A Performance Portable Radiation Hydrodynamics Code based on FleCSI Framework Journal: SoftwareX Julien Loiseau, Hyun Lim, Andrés Yagüe López, Mammadbaghir Baghirzade, Shihab Shahriar Khan, Yoonsoo Kim, Sudarshan Neopane, Alexander Strack, Farhana Taiyebah, Ben Bergen Abstract Hydrodynamics And Radiation Diffusion ( HARD ) is an open-source application for high-performance simulations of compressible hydrodynamics with radiation-diffusion coupling. Built on the FleCSI [ 1 ] (Flexible Computational Science Infrastructure) framework, HARD expresses its computational units as tasks whose execution can be orchestrated by multiple back-end runtimes, including Legion [ 2 ] , MPI [ 3 ] , and HPX [ 4 ] . Node-level parallelism is delegated to Kokkos [ 5 ] , providing a single, portable code base that runs efficiently on laptops, small homogeneous clusters, and the largest heterogeneous supercomputers currently available. To ensure scientific reliability, HARD includes a regression-test suite that automatically reproduces canonical verification problems such as the Sod and LeBlanc shock tubes, and the Sedov blast wave, comparing numerical solutions against known analytical results. The project is distributed under an OSI-approved license, hosted on GitHub, and accompanied by reproducible build scripts and continuous integration workflows. This combination of performance portability, verification infrastructure, and community-focused development makes HARD a sustainable platform for advancing radiation hydrodynamics research across multiple domains. Metadata Nr. Code metadata description Metadata C1 Current code version V1.0 C2 Permanent link to code/repository used for this code version https://github.com/lanl/hard C3 Permanent link to Reproducible Capsule C4 Legal Code License BSD 3-Clause License C5 Code versioning system used git C6 Software code languages, tools, and services used C++, MPI C7 Compilation requirements, operating environments & dependencies Kokkos, Singularity-EOS C8 If available Link to developer documentation/manual https://lanl.github.io/HARD/ C9 Support email for questions [email protected] Table 1: Code metadata (mandatory) 1 Motivation and significance Radiation hydrodynamics (RHD) governs systems where radiation–matter coupling shapes the global dynamics, including high-energy-density physics (HEDP) problems such as inertial confinement fusion (ICF), stellar evolution and explosions, accretion flows, and radiative shocks. Accurately capturing these phenomena requires solving tightly coupled, nonlinear partial differential equations for hydrodynamics and radiation diffusion across disparate spatial and temporal scales, often in the presence of stiff source terms and strong gradients. While several mature RHD and radiation–Magnetohydrodynamics (MHD) codes exist [ 6 , 7 , 8 , 9 , 10 , 11 ] , many legacy implementations face practical challenges on modern heterogeneous platforms such as limited performance portability, rigid data models that impede algorithmic experimentation, and difficulty integrating with contemporary tooling for CI, packaging, and reproducibility. As experimental capabilities and computing architectures evolve, a software platform that cleanly separates physics from execution back ends, and that scales from laptops to leadership-class systems, becomes essential. HARD is designed to meet this need. Built on the task-based FleCSI runtime layer, it decouples application physics from parallel orchestration. This design enables the same source to target multiple distributed-memory runtimes (Legion, MPI, and HPX) while delegating node-level parallelism to Kokkos through an abstraction layer, yielding a single, portable implementation for CPUs and GPUs. Beyond portability, HARD aims to serve as a testbed for new methods in RHD by providing clear interfaces for numerics (reconstruction, Riemann solvers, time integrators) and physics (equations of state, radiation closures). To promote scientific reliability and re-use, HARD includes: 1. a regression test suite that automatically reproduces canonical verification problems (e.g., Sod and LeBlanc shock tubes, Sedov blast) and compares against analytic or reference solutions; 2. containerized and scriptable builds to ensure bitwise-reproducible environments across developer machines and CI; 3. open-source governance and contribution workflows to lower the barrier for community extensions (e.g., MHD, multigroup radiation, advanced microphysics). In summary, HARD addresses two complementary gaps: (i) a performance-portable RHD code that maps cleanly to heterogeneous architectures, and (ii) a modular, reproducible platform that accelerates method development and cross-code comparison for the RHD community. 2 Software description HARD is implemented in modern C++ for UNIX and high-performance computing (HPC) platforms, taking advantage of the Standard Template Library (STL). Spack [ 12 ] is used to configure both build- and run-time environments. Simulation data is written in CSV and Catalyst formats, enabling post-processing with visualization tools such as ParaView [ 13 ] and VisIt [ 14 ] . 2.1 Software architecture Figure 1: HARD within the FleCSI ecosystem. The software stack is shown in Figure 1 . HARD uses a FleCSI specialization layer that employs the N-Array topology (a multidimensional array with exclusive, shared, and ghost entities). The HARD base code defines its specialization in spec/ , and the application that uses the specialization in app/ . The current control model consists of initialization, time-stepping (advection, diffusion, and source updates), and output. 2.2 Software functionalities HARD provides a set of core capabilities for solving RHD problems on modern HPC systems. At a high level, these include: 1. implementation of the coupled hydrodynamics and radiation–diffusion equations, 2. modular numerical algorithms (finite-volume solvers, reconstruction methods, Riemann solvers, and implicit diffusion solvers), 3. verification problems and regression tests for scientific reliability, and 4. scalable distributed parallel execution across CPUs and GPUs via FleCSI back ends. The following subsections describe the mathematical models, numerical algorithms, and example problems that illustrate these capabilities. 2.2.1 Basic Equations We adopt the co-moving (Eulerian) frame formulation of the RHD equations in the diffusion limit [ 15 ] . This formulation is particularly well-suited for regimes where the mean free path of photons is small compared to the characteristic hydrodynamic length scales, allowing the radiative flux to be described by a diffusion approximation. In this framework, the governing equations can be compactly expressed in Einstein notation as \partial_{t}\mathbf{U}+\partial_{j}\mathbf{F}^{j}=\mathbf{S} (1) with the evolved variables \mathbf{U}=\left[\rho,\rho v^{i},e+\rho v^{2}/2,E\right]^{T}, (2) fluxes \mathbf{F}^{j}=\left[\begin{matrix}\rho v^{j}\\ \rho v^{i}v^{j}+p\delta_{ij}\\ \left(e+\rho v^{2}/2+p\right)v^{j}\\ Ev^{j}\\ \end{matrix}\right], (3) and source terms \mathbf{S}=\left[\begin{matrix}0\\ f^{i}\\ f^{i}v_{i}+\dot{q}\\ -\partial_{i}F_{r}^{i}-P_{r}^{ij}\nabla_{i}v_{j}-\dot{q}\\ \end{matrix}\right], (4) where \rho is fluid mass density, v^{i} is the component i of the fluid velocity vector, e is the fluid internal energy density, p is the fluid pressure, E is radiation energy density, F_{r}^{i} is the component i of the radiation energy flux vector, P_{r}^{ij} is the i , j element of the radiation pressure tensor, f^{i} is the component i of the radiation force density vector f^{i}=\frac{\kappa\rho}{c}F_{r}^{i}, (5) and \dot{q} is the radiative heating rate \dot{q}=\frac{dq}{dt}=c\kappa\rho(E-aT^{4}), (6) where \kappa is the mean opacity, c is the speed of light, T is the fluid temperature, and a is the radiation defined as a\equiv 4\sigma/c , where \sigma is the Stefan-Boltzmann constant. To close the evolution system for the fluid, it is necessary to specify the functional form of the fluid pressure p , which can be obtained from an equation of state (EoS). In this work, we adopt Singularity-EOS [ 16 ] , which provides a flexible EoS framework that enables the modeling of different materials. The radiative transfer equation—the final component of the evolution system given in Equation 1 —further requires closure relations for both the radiation energy flux density F^{i} and the radiation pressure tensor P_{r}^{ij} . In the optically-thick diffusion limit, they are given as F^{i}=-\frac{c}{3\kappa\rho}\nabla E, (7) P_{r}^{ij}=\frac{1}{3}E\delta_{ij}, (8) provided the mean free path of photons are much shorter than the length scale of interest. The evolution equations presented above are valid in the diffusion (optically thick) limit. However, it is