Pith. sign in

REVIEW 8 minor 1 cited by

Relativity code ported to GPU runs 10x faster, matching original

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · glm-5.2

2026-07-10 01:55 UTC pith:TJ566S4R

load-bearing objection Validated Kokkos port of SACRA-MPI to GPU/APU with ~10x speedup; honest about scaling limits

arxiv 2607.08743 v1 pith:TJ566S4R submitted 2026-07-09 astro-ph.HE astro-ph.IMgr-qc

SACRA-K: A Performance-Portable Numerical Relativity Code with Kokkos

Ming-Zhe Han , Kenta Kiuchi , Masaru Shibata This is my paper
classification astro-ph.HE astro-ph.IMgr-qc PACS 04.25.D-04.30.-w04.40.Dg95.30.Lz
keywords numerical relativityperformance portabilityKokkosgravitational wavesbinary neutron starsadaptive mesh refinementGPU accelerationBSSN formulation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents SACRA-K, a rewrite of the Fortran numerical-relativity code SACRA-MPI into C++ using the Kokkos library, which allows a single source code to run on CPUs, GPUs, and APUs from different vendors. The authors validate that the port reproduces the original code's gravitational-wave waveforms across binary black hole, black hole–neutron star, and binary neutron star configurations to accuracy at or below the distinguishability threshold of current gravitational-wave detectors, preserves pi-symmetry at the bitwise level, and achieves second-order convergence for binary neutron star mergers. In the smallest test configuration, SACRA-K runs roughly an order of magnitude faster on NVIDIA A100 or AMD MI300A accelerators than the Fortran code on a 72-core CPU node, and the authors measure weak scaling up to 256 accelerator devices with efficiencies of 77–83% relative to a 32-device baseline.

Core claim

The central claim is that an established, physics-rich numerical-relativity code (SACRA-MPI) can be ported to a performance-portable C++/Kokkos framework (SACRA-K) without sacrificing physical fidelity, gaining approximately an order of magnitude in throughput on modern accelerators. The correctness of the port is established by three independent checks: waveform agreement with the parent code at or below detector-distinguishability thresholds, exact bitwise preservation of pi-symmetry, and second-order convergence in binary neutron star merger simulations. The performance gain is demonstrated per-device in the smallest configuration and the scaling behaviour is characterised across up to 26

What carries the argument

Kokkos performance-portability library (abstracts CUDA, HIP, OpenMP backends); BSSN formulation with Z4c constraint damping for spacetime evolution; box-in-box Berger-Oliger adaptive mesh refinement with time subcycling; finite-difference (4th-order centered/upwind) for geometry and finite-volume PPM/HLLC for hydrodynamics; ghost-buffer communication via GPU-aware MPI; team-scratch memory optimisation on AMD APU

Load-bearing premise

The headline 10x speedup is established only for the smallest configuration (4 accelerators), while production-scale simulations requiring many more devices face a strong-scaling efficiency that drops to about 10% at 256 devices, driven by the growing fraction of ghost-buffer cells relative to active cells as subdomains shrink. The claim that GPU acceleration enables large parameter-space surveys depends on weak-scaling behaviour that, while reasonable, is measured on a less

What would settle it

If the waveform discrepancies between SACRA-K and SACRA-MPI were found to exceed the distinguishability threshold of current gravitational-wave detectors during the inspiral phase, or if the pi-symmetry were broken at the bitwise level, or if the BNS convergence order fell below second order, the port would be deemed physically invalid. On the performance side, if the per-device throughput advantage over CPU did not hold at production-relevant resolutions or if weak scaling efficiency collapsed below 50% at moderate device counts, the practical utility claim would be undermined.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If the port fidelity holds at production resolutions, gravitational-wave waveform libraries for binary neutron star and black hole–neutron star systems can be generated at roughly one-tenth the previous computational cost, enabling broader parameter-space surveys.
  • The Kokkos-based approach demonstrated here provides a template for porting other established Fortran numerical-relativity codes to heterogeneous architectures without rewriting the underlying physics.
  • The observed strong-scaling degradation (10% efficiency at 256 devices) signals that future production runs will need either larger per-device workloads or algorithmic changes to ghost-buffer communication to maintain throughput.
  • The post-merger phase divergence between SACRA-K and SACRA-MPI suggests that high-precision post-merger waveforms may be intrinsically sensitive to floating-point-level differences, a limitation that affects all GPU-ported NR codes, not just this one.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The weak scaling results (77–83% at 256 devices relative to a 32-device baseline) suggest that production-scale simulations using hundreds of devices will achieve effective throughput well below the headline 10x speedup, though still substantially above CPU-only performance.
  • The strong-scaling bottleneck from the fixed 6-cell ghost buffer implies that reducing buffer width or adopting asynchronous communication overlap could yield disproportionate gains at high device counts, a direction the authors acknowledge but have not yet implemented.
  • If the second-order convergence ceiling is set by the approximate Riemann solver and stellar surfaces, then higher-order hydrodynamics schemes would be needed to improve convergence beyond what is reported, independent of the hardware port.
  • The exact bitwise pi-symmetry preservation under Kokkos parallel reductions (with -ffp-contract=off) indicates that reproducibility of symmetry properties on GPUs is achievable but requires compiler-flag-level control, which may not transfer automatically to other codes.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

0 major / 8 minor

Summary. The manuscript presents SACRA-K, a C++ port of the Fortran numerical relativity code SACRA-MPI, built on the Kokkos performance-portability library. The port retains the physics modules of the parent code (BSSN with Z4c constraint damping, Berger-Oliger AMR, HRSC hydrodynamics from NANASI) while enabling execution on CPUs, NVIDIA GPUs, and AMD APUs from a single source. The authors validate the port through code-to-code waveform comparisons (BBH, BHNS, BNS), a bitwise pi-symmetry test, and a four-resolution self-convergence test for BNS. Performance is benchmarked on A100 and MI300A hardware, showing roughly an order-of-magnitude speedup over the Fortran code on a 72-core CPU node in the smallest configuration, with strong and weak scaling measured up to 256 devices.

Significance. Performance-portable numerical relativity codes are of growing importance to the field as computing architectures diversify and next-generation detectors demand higher-fidelity waveforms. SACRA-MPI is an established code with a strong track record in BNS and BHNS science; porting it to GPU/APU architectures via Kokkos is a valuable contribution. The validation suite is thorough and the acceptance criteria are well-motivated, grounded in inter-code scatter, resolution-dependent variation, and detector distinguishability thresholds. The honest reporting of scaling limitations (strong scaling efficiency dropping to ~10% at 256 devices due to fixed ghost-buffer overhead) is commendable. The pi-symmetry test, showing exact bitwise preservation to the roundoff floor, is a particularly strong correctness check on the Kokkos parallelization. The work is a solid engineering contribution that enables future science campaigns on heterogeneous architectures.

minor comments (8)
  1. Abstract: The statement that waveform discrepancies 'remain at or below the distinguishability threshold of current gravitational-wave detectors' is not qualified for the BNS post-merger phase, where phase differences grow to O(1) rad (Section 3.1.3, Figure 4). The body text is transparent about this and correctly argues that post-merger signal is not the primary target for current detector waveform modeling, but the abstract should be qualified to avoid overgeneralization.
  2. Section 3.1.1: The BBH extraction radius r_0 = 24 M_sun is acknowledged as insufficiently large for scientific runs. The authors argue that the finite-radius bias is 'largely common mode' between the two codes. This is reasonable, but a sentence noting that the measured discrepancy may underestimate the total error with respect to the asymptotic waveform would strengthen the discussion.
  3. Section 3.1.2: The BHNS consistency test meets the acceptance criteria only marginally, with |ΔA|/A riding the 10^-3 line in late inspiral and |Δφ_GW| touching 10^-2 rad just before merger. The authors attribute this to the sensitivity of finite-volume reconstruction at the stellar surface to floating-point operation ordering. This is plausible, but a brief quantitative comparison to the inter-resolution spread at similar configurations would help contextualize whether this marginality is a concern for production runs at higher N.
  4. Section 4.1, Table 1: The zone-cycles-per-second metric is defined in the text but the exact zone-cycle count formula (number of level evolutions times active cells per level) could be stated more explicitly, perhaps as an equation, to aid reproducibility of the benchmark.
  5. Section 4.2: The strong scaling degradation to ~10% efficiency at 256 devices is attributed to the growing buffer-to-active-cell ratio. The 6-cell ghost buffer width is set by the stencil requirements (fourth-order finite differencing plus KO dissipation). It would be useful to note whether a smaller buffer is feasible at lower stencil order or whether this is a fundamental constraint of the current numerical scheme.
  6. Section 2.4: The statement that the entire evolution is carried out on the device through Kokkos kernels is clear, but the specific optimization for the AMD APU (team scratch for 8^3 tiles) is described only briefly. A sentence or two on whether this optimization is also applicable to NVIDIA GPUs (or why it is APU-specific) would help readers understand the portability of the optimization strategy.
  7. Figure 2: The y-axis label for the bottom panel uses |Δφ_GW| while the text uses both |Δφ_GW| and |Δφ|. Standardizing the notation (the text uses |Δϕ_GW| with a phi symbol in some places and |Δφ_GW| with a varphi in others) would improve readability.
  8. Section 2.2: The sentence describing the fallback to HLLE solver 'wherever the intermediate state would be unphysical' could briefly clarify how frequently this fallback occurs in practice for the test configurations, as it affects the effective order of the hydrodynamics scheme (not stated explicitly).

Circularity Check

0 steps flagged

No circularity found: SACRA-K is a code port validated against external benchmarks, independent codes, and self-convergence tests with no self-definitional or fitted-input-as-prediction patterns.

full rationale

SACRA-K is a port of SACRA-MPI from Fortran to C++/Kokkos. Its central claims are validated through three independent strategies: (1) code-to-code comparison against SACRA-MPI, which is an independently published code (T. Yamamoto et al. 2008) with its own validation history; (2) bitwise pi-symmetry preservation, which is a direct numerical test with no fitted parameters; and (3) a self-convergence test at four resolutions (N=92,108,124,188) using the standard ansatz (Eq. 28) to extract convergence order, yielding p_conv ~ 2 as expected from the method-of-lines with third-order PPM reconstruction and an approximate Riemann solver. The acceptance thresholds (|Delta phi_GW| <= 10^-2 rad, |Delta A|/A <= 10^-3) are derived from external cross-code comparison studies (Hannam et al. 2009; Hinder et al. 2014; Lovelace et al. 2016) and detector distinguishability criteria (Lindblom et al. 2008; Flanagan & Hughes 1998), not from the authors' own prior results. The convergence test configuration matches K. Kiuchi (2025), but this is used for cross-validation against an independently reported convergence order, not as a self-definitional input. The performance claims (speedup, scaling) are direct measurements with no fitted parameters or ansatz. No step in the derivation chain reduces to its own inputs by construction.

Axiom & Free-Parameter Ledger

8 free parameters · 6 axioms · 0 invented entities

The paper introduces no new physical entities, particles, forces, or mathematical objects. All evolved variables (W, gamma_tilde, K, A_tilde, Gamma_tilde, Theta, alpha, beta, B) are standard BSSN+Z4c fields. The free parameters are standard numerical choices in NR codes, not fitted to produce a specific result. The Kokkos library and FUKA initial data solver are external tools.

free parameters (8)
  • sigma_KO = 0.5
    Kreiss-Oliger dissipation coefficient, set by hand (Section 2.2). Standard value in BSSN codes.
  • Courant factor = 0.5
    Time step constraint per refinement level (Section 2.2). Standard choice.
  • kappa_1 = 0.005 M_sun^-1
    Z4c damping parameter (Section 3). Chosen for constraint propagation control.
  • kappa_2 = 0
    Z4c damping parameter (Section 3). Set to zero.
  • eta = 0.36 M_sun^-1 (default), 0.03 M_sun^-1 (convergence test)
    Shift damping parameter (Section 3). The convergence test uses a different value to match K. Kiuchi (2025).
  • b (PPM limiter) = 3
    Steep min-mod limiter parameter for PPM reconstruction (Section 2.2).
  • Gamma_th = 1.8 (DD2 tests), 5/3 (H4 convergence test)
    Thermal adiabatic index for the hybrid EOS (Section 2.1.2, Section 3.3).
  • f_cut = 0.8 m Omega_0 / (2 pi)
    Cutoff frequency for fixed frequency integration of Psi4 (Section 2.3).
axioms (6)
  • domain assumption BSSN formulation with Z4c constraint propagation is a valid and stable formulation of the Einstein equations for compact binary evolution.
    Section 2.1.1. This is the foundational evolution scheme, inherited from SACRA-MPI and widely used in the NR community.
  • domain assumption The moving puncture gauge (1+log slicing, hyperbolic Gamma driver) is appropriate for evolving puncture black holes.
    Section 2.1.1, Equations (6)-(8). Standard gauge choice in NR.
  • domain assumption A perfect fluid without magnetic fields is a sufficient matter model for the validation configurations studied.
    Section 2.1.2. The paper explicitly states it evolves matter as a perfect fluid without magnetic fields, deferring MHD to future work.
  • domain assumption The box-in-box Berger-Oliger AMR with 2:1 refinement ratio and time subcycling correctly handles the multiscale nature of compact binary mergers.
    Section 2.2.1. Inherited from SACRA-MPI (T. Yamamoto et al. 2008).
  • domain assumption Kokkos provides performance portability across CUDA, HIP, and OpenMP backends without altering numerical results.
    Section 2.4. The paper compiles with -fno-fast-math to mitigate floating-point non-determinism, but the fundamental assumption is that Kokkos dispatches identical kernels across backends.
  • domain assumption The FUKA initial data solver produces physically valid initial data for BBH, BHNS, and BNS configurations.
    Section 3. External dependency (L. J. Papenfort et al. 2021).

pith-pipeline@v1.1.0-glm · 25382 in / 3613 out tokens · 354328 ms · 2026-07-10T01:55:55.426735+00:00 · methodology

0 comments
read the original abstract

We present SACRA-K (SimulAtor for Compact objects in Relativistic Astrophysics with Kokkos), a performance-portable numerical relativity code ported from the Fortran code SACRA-MPI to C++ with the Kokkos library. SACRA-K retains the physics and numerical methods of SACRA-MPI, namely a Baumgarte-Shapiro-Shibata-Nakamura (BSSN) spacetime evolution with Z4c constraint propagation and a box-in-box Berger-Oliger adaptive mesh refinement, together with the high resolution shock capturing scheme for the general relativistic hydrodynamics from NANASI, while gaining cross vendor portability. We validate the port against SACRA-MPI across binary black hole, black hole neutron star, and binary neutron star configurations. Across these tests, the waveform discrepancies are well below both the scatter among independent numerical relativity codes and the resolution dependent variation within a given code, and remain at or below the distinguishability threshold of current gravitational-wave detectors; the $\pi$ symmetry is preserved exactly at the bitwise level; and the gravitational wave phase of the binary neutron star merger exhibits second order convergence. In the smallest test configuration, SACRA-K runs about an order of magnitude faster on the NVIDIA A100 GPU cluster or the AMD MI300A APU cluster than the Fortran SACRA-MPI on the CPU cluster, and we measure its scaling up to 256 accelerator devices.

Figures

Figures reproduced from arXiv: 2607.08743 by Kenta Kiuchi, Masaru Shibata, Ming-Zhe Han.

Figure 1
Figure 1. Figure 1: Schematic figures of the AMR scheme in SACRA-K. Panel (a): hierarchical computational domain structure. For 0 ≤ ℓ ≤ ℓfix each level consists of a single cubic patch centered on the coordinate origin (fixed levels, black). For ℓfix < ℓ ≤ ℓmax each level consists of two patches that follow the two compact objects (moving levels, red). Successive levels are related by a 2:1 refinement ratio, so the grid spaci… view at source ↗
Figure 2
Figure 2. Figure 2: Waveform comparison between SACRA-K and SACRA-MPI Fortran for the BBH case. Top: real part of the l = m = 2 mode of h l,m extracted on a sphere, for SACRA-MPI Fortran (blue solid) and SACRA-K (orange dashed). Middle: relative amplitude difference |∆A|/A. Bottom: absolute GW phase difference |∆ϕGW|. The horizontal black dotted lines denote the acceptance criteria, 10−3 for the relative ampli￾tude difference… view at source ↗
Figure 3
Figure 3. Figure 3: Same as [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Same as [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: π symmetry preservation diagnostic for an equal mass non spinning BNS evolution at resolution N = 102. The momentum components |Px| and |Py| are shown in the upper and lower panels; in each panel the ADM and fluid linear momenta are drawn in blue and orange, respectively. During the evolution we monitor the symmetry di￾rectly. For every symmetric pair of points, we com￾pute the difference between each fiel… view at source ↗
Figure 6
Figure 6. Figure 6: Self convergence of the SACRA-K BNS merger simulation at four resolutions, N = 92 (blue solid), 108 (or￾ange dashed), 124 (green dash dotted), and 188 (red dot￾ted), shown against the retarded time tret. Top: dominant strain mode D0h 22 + /m0. Second: gravitational wave phase difference |∆ϕGW| of the three lower resolutions relative to N = 188. Third: locally fitted convergence order pconv, with the shaded… view at source ↗
Figure 7
Figure 7. Figure 7: Strong scaling of SACRA-K on the Viper APU cluster at a fixed total problem size (N = 144). Stars show the strong scaling efficiency, the achieved speedup divided by the ideal linear speedup, against the number of APU devices, from the four device 2 × 2 × 1 baseline to 256 devices. The dotted line marks ideal scaling. N = 144 configuration and distribute it over a growing number of devices. Specifically, w… view at source ↗
Figure 8
Figure 8. Figure 8: Weak scaling of SACRA-K on the Viper APU clus￾ter at a fixed subdomain size per device. The weak scaling ef￾ficiency tbase/tn is shown against the number of APU devices for two per device loads, N = 50 (blue) and N = 100 (or￾ange), each normalized to the four device baseline (solid) and to the 32 device baseline (dashed). The dotted line marks ideal weak scaling. 4.3. Weak scaling We finally test the weak … view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. GRACE: An Open-Source Framework for GPU-Accelerated Numerical Relativity

    gr-qc 2026-07 accept novelty 6.0

    GRACE is a validated, open-source, Kokkos+p4est GPU-portable framework that evolves ideal GRMHD with constrained transport self-consistently coupled to Z4c Einstein equations on fixed or adaptive meshes.

Reference graph

Works this paper leans on

79 extracted references · 79 canonical work pages · cited by 1 Pith paper · 10 internal anchors

  1. [1]

    G., et al

    Abac, A. G., et al. 2025, Phys. Rev. Lett., 135, 111403, doi: 10.1103/kw5g-d732

    work page doi:10.1103/kw5g-d732 2025

  2. [2]

    P., et al

    Abbott, B. P., et al. 2016, Phys. Rev. Lett., 116, 061102, doi: 10.1103/PhysRevLett.116.061102

    work page doi:10.1103/physrevlett.116.061102 2016

  3. [3]

    Physical Review Letters , archivePrefix = "arXiv", eprint =

    Abbott, B. P., et al. 2017a, Phys. Rev. Lett., 119, 161101, doi: 10.1103/PhysRevLett.119.161101

    work page doi:10.1103/physrevlett.119.161101

  4. [4]

    P., et al

    Abbott, B. P., et al. 2017b, Astrophys. J. Lett., 848, L12, doi: 10.3847/2041-8213/aa91c9

    work page doi:10.3847/2041-8213/aa91c9 2041

  5. [5]

    P., et al

    Abbott, B. P., et al. 2018, Phys. Rev. Lett., 121, 161101, doi: 10.1103/PhysRevLett.121.161101

    work page doi:10.1103/physrevlett.121.161101 2018

  6. [6]

    P., et al

    Abbott, B. P., et al. 2019, Phys. Rev. X, 9, 011001, doi: 10.1103/PhysRevX.9.011001

    work page doi:10.1103/physrevx.9.011001 2019

  7. [7]

    2006, Phys

    Meter, J. 2006, Phys. Rev. Lett., 96, 111102, doi: 10.1103/PhysRevLett.96.111102

    work page doi:10.1103/physrevlett.96.111102 2006

  8. [8]

    Phoebus: Performance Portable GRRMHD for Relativistic Astrophysics

    Barker, B., Gogilashvili, M., Rodriguez-Bueno, J., et al. 2024, https://arxiv.org/abs/2410.09146

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  9. [9]

    W., & Shapiro, S

    Baumgarte, T. W., & Shapiro, S. L. 1998, Phys. Rev. D, 59, 024007, doi: 10.1103/PhysRevD.59.024007

    work page doi:10.1103/physrevd.59.024007 1998

  10. [10]

    J., & Colella, P

    Berger, M. J., & Colella, P. 1989, Journal of computational Physics, 82, 64

    work page 1989

  11. [11]

    2015, Phys

    Bernuzzi, S., Nagar, A., Dietrich, T., & Damour, T. 2015, Phys. Rev. Lett., 114, 161103, doi: 10.1103/PhysRevLett.114.161103

    work page doi:10.1103/physrevlett.114.161103 2015

  12. [12]

    2014, Living Rev

    Blanchet, L. 2014, Living Rev. Rel., 17, 2, doi: 10.12942/lrr-2014-2

    work page doi:10.12942/lrr-2014-2 2014

  13. [13]

    B., & Pretorius, F

    Buonanno, A., Cook, G. B., & Pretorius, F. 2007, Phys. Rev. D, 75, 124018, doi: 10.1103/PhysRevD.75.124018

    work page doi:10.1103/physrevd.75.124018 2007

  14. [14]

    1999, Phys

    Buonanno, A., & Damour, T. 1999, Phys. Rev. D, 59, 084006, doi: 10.1103/PhysRevD.59.084006

    work page doi:10.1103/physrevd.59.084006 1999

  15. [15]

    2006, Phys

    Zlochower, Y. 2006, Phys. Rev. Lett., 96, 111101, doi: 10.1103/PhysRevLett.96.111101 Carter Edwards, H., Trott, C. R., & Sunderland, D. 2014, Journal of Parallel and Distributed Computing, 74, 3202, doi: https://doi.org/10.1016/j.jpdc.2014.07.003

    work page doi:10.1103/physrevlett.96.111101 2006

  16. [16]

    2017, Phys

    Chatziioannou, K., Klein, A., Yunes, N., & Cornish, N. 2017, Phys. Rev. D, 95, 104004, doi: 10.1103/PhysRevD.95.104004

    work page doi:10.1103/physrevd.95.104004 2017

  17. [17]

    Colella, P., & Woodward, P. R. 1984, J. Comput. Phys., 54, 174, doi: 10.1016/0021-9991(84)90143-8

    work page doi:10.1016/0021-9991(84)90143-8 1984

  18. [18]

    Taking GPU Programming Models to Task for Performance Portability

    Davis, J. H., Sivaraman, P., Kitson, J., et al. 2024, arXiv e-prints, arXiv:2402.08950, doi: 10.48550/arXiv.2402.08950 Del Zanna, L., Bucciantini, N., & Londrillo, P. 2003, Astron. Astrophys., 400, 397, doi: 10.1051/0004-6361:20021641

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2402.08950 2024

  19. [19]

    2017, Phys

    Dietrich, T., Bernuzzi, S., & Tichy, W. 2017, Phys. Rev. D, 96, 121501, doi: 10.1103/PhysRevD.96.121501

    work page doi:10.1103/physrevd.96.121501 2017

  20. [20]

    2021, Gen

    Dietrich, T., Hinderer, T., & Samajdar, A. 2021, Gen. Rel. Grav., 53, 27, doi: 10.1007/s10714-020-02751-6

    work page doi:10.1007/s10714-020-02751-6 2021

  21. [21]

    A Horizon Study for Cosmic Explorer: Science, Observatories, and Community

    Evans, M., et al. 2021, https://arxiv.org/abs/2109.09882

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  22. [22]

    2025, Phys

    Ferguson, D., et al. 2025, Phys. Rev. D, 112, 044043, doi: 10.1103/gk7x-9cds

    work page doi:10.1103/gk7x-9cds 2025

  23. [23]

    W., & Sundar, H

    Fernando, M., Neilsen, D., Zlochower, Y., Hirschmann, E. W., & Sundar, H. 2023, Phys. Rev. D, 107, 064035, doi: 10.1103/PhysRevD.107.064035

    work page doi:10.1103/physrevd.107.064035 2023

  24. [24]

    2025, Astrophys

    Fields, J., Zhu, H., Radice, D., et al. 2025, Astrophys. J. Suppl., 276, 35, doi: 10.3847/1538-4365/ad9687

    work page doi:10.3847/1538-4365/ad9687 2025

  25. [25]

    E., & Hughes, S

    Flanagan, E. E., & Hughes, S. A. 1998, Phys. Rev. D, 57, 4566, doi: 10.1103/PhysRevD.57.4566

    work page doi:10.1103/physrevd.57.4566 1998

  26. [26]

    Snowmass2021 Cosmic Frontier White Paper: Numerical relativity for next-generation gravitational-wave probes of fundamental physics

    Foucart, F., Laguna, P., Lovelace, G., Radice, D., & Witek, H. 2022, https://arxiv.org/abs/2203.08139

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  27. [27]

    2023, Astrophys

    Fujibayashi, S., Kiuchi, K., Wanajo, S., et al. 2023, Astrophys. J., 942, 39, doi: 10.3847/1538-4357/ac9ce0

    work page doi:10.3847/1538-4357/ac9ce0 2023

  28. [28]

    Subsolar-mass binary mergers of strange stars and neutron stars: gravitational waves and ejecta

    Gao, Y., Han, M.-Z., Kiuchi, K., et al. 2026a, https://arxiv.org/abs/2607.07668

    work page internal anchor Pith review Pith/arXiv arXiv

  29. [29]

    2026b, Phys

    Gao, Y., Hayashi, K., Kiuchi, K., et al. 2026b, Phys. Rev. D, 113, 023011, doi: 10.1103/w8gq-3d54

    work page doi:10.1103/w8gq-3d54

  30. [30]

    K., & Moszkowski, S

    Glendenning, N. K., & Moszkowski, S. A. 1991, Phys. Rev. Lett., 67, 2414, doi: 10.1103/PhysRevLett.67.2414

    work page doi:10.1103/physrevlett.67.2414 1991

  31. [31]

    W., & O’Shea, B

    Grete, P., Glines, F. W., & O’Shea, B. W. 2021, IEEE Transactions on Parallel and Distributed Systems, 32, 85, doi: 10.1109/TPDS.2020.3010016

    work page doi:10.1109/tpds.2020.3010016 2021

  32. [32]

    Parthenon -- a performance portable block-structured adaptive mesh refinement framework

    Grete, P., Dolence, J. C., Miller, J. M., et al. 2022, arXiv e-prints, arXiv:2202.12309, doi: 10.48550/arXiv.2202.12309

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2202.12309 2022

  33. [33]

    2026, Phys

    Habib, S., et al. 2026, Phys. Rev. D, 113, 104062, doi: 10.1103/5sxn-knwz

    work page doi:10.1103/5sxn-knwz 2026

  34. [34]

    Hamilton, M. C. B., & Messman, W. A. 2025, Class. Quant. Grav., 42, 095012, doi: 10.1088/1361-6382/adccb1

    work page doi:10.1088/1361-6382/adccb1 2025

  35. [35]

    2009, Phys

    Hannam, M., et al. 2009, Phys. Rev. D, 79, 084025, doi: 10.1103/PhysRevD.79.084025

    work page doi:10.1103/physrevd.79.084025 2009

  36. [36]

    2010, Nucl

    Hempel, M., & Schaffner-Bielich, J. 2010, Nucl. Phys. A, 837, 210, doi: 10.1016/j.nuclphysa.2010.02.010

    work page doi:10.1016/j.nuclphysa.2010.02.010 2010

  37. [37]

    2013, Phys

    Hilditch, D., Bernuzzi, S., Thierfelder, M., et al. 2013, Phys. Rev. D, 88, 084057, doi: 10.1103/PhysRevD.88.084057 15

    work page doi:10.1103/physrevd.88.084057 2013

  38. [38]

    2014, Class

    Hinder, I., et al. 2014, Class. Quant. Grav., 31, 025012, doi: 10.1088/0264-9381/31/2/025012

    work page doi:10.1088/0264-9381/31/2/025012 2014

  39. [39]

    2025, Phys

    Ji, L., Haas, R., Zlochower, Y., et al. 2025, Phys. Rev. D, 112, 024049, doi: 10.1103/bqpq-5cp9

    work page doi:10.1103/bqpq-5cp9 2025

  40. [40]

    1996, Journal of Computational Physics, 126, 202, doi: 10.1006/jcph.1996.0130

    Jiang, G.-S., & Shu, C.-W. 1996, Journal of Computational Physics, 126, 202, doi: 10.1006/jcph.1996.0130

    work page doi:10.1006/jcph.1996.0130 1996

  41. [41]

    V., et al

    Kalinani, J. V., et al. 2025, Class. Quant. Grav., 42, 025016, doi: 10.1088/1361-6382/ad9c11

    work page doi:10.1088/1361-6382/ad9c11 2025

  42. [42]

    2017, Nature, 551, 80, doi: 10.1038/nature24453

    Ramirez-Ruiz, E. 2017, Nature, 551, 80, doi: 10.1038/nature24453

    work page doi:10.1038/nature24453 2017

  43. [43]

    2018, Phys

    Kawaguchi, K., Kiuchi, K., Kyutoku, K., et al. 2018, Phys. Rev. D, 97, 044044, doi: 10.1103/PhysRevD.97.044044

    work page doi:10.1103/physrevd.97.044044 2018

  44. [44]

    2025, Phys

    Kiuchi, K. 2025, Phys. Rev. D, 112, 084008, doi: 10.1103/zmdc-xkcm

    work page doi:10.1103/zmdc-xkcm 2025

  45. [45]

    2015, Phys

    Shibata, M. 2015, Phys. Rev. D, 92, 124034, doi: 10.1103/PhysRevD.92.124034

    work page doi:10.1103/physrevd.92.124034 2015

  46. [46]

    2023, Phys

    Kiuchi, K., Fujibayashi, S., Hayashi, K., et al. 2023, Phys. Rev. Lett., 131, 011401, doi: 10.1103/PhysRevLett.131.011401

    work page doi:10.1103/physrevlett.131.011401 2023

  47. [47]

    E., Sekiguchi, Y., & Shibata, M

    Kiuchi, K., Held, L. E., Sekiguchi, Y., & Shibata, M. 2022, Phys. Rev. D, 106, 124041, doi: 10.1103/PhysRevD.106.124041

    work page doi:10.1103/physrevd.106.124041 2022

  48. [48]

    2020, Phys

    Shibata, M. 2020, Phys. Rev. D, 101, 084006, doi: 10.1103/PhysRevD.101.084006

    work page doi:10.1103/physrevd.101.084006 2020

  49. [49]

    2017, Phys

    Kiuchi, K., Kawaguchi, K., Kyutoku, K., et al. 2017, Phys. Rev. D, 96, 084060, doi: 10.1103/PhysRevD.96.084060

    work page doi:10.1103/physrevd.96.084060 2017

  50. [50]

    A magnetar formation in binary neutron star merger

    Kiuchi, K., Reboul-Salze, A., Sekiguchi, Y., & Shibata, M. 2026, https://arxiv.org/abs/2606.11299

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  51. [51]

    2024, Nature Astron., 8, 298, doi: 10.1038/s41550-024-02194-y

    Kiuchi, K., Reboul-Salze, A., Shibata, M., & Sekiguchi, Y. 2024, Nature Astron., 8, 298, doi: 10.1038/s41550-024-02194-y

    work page doi:10.1038/s41550-024-02194-y 2024

  52. [52]

    1973, Methods for the approximate solution of time dependent problems No

    Kreiss, H., & Oliger, J. 1973, Methods for the approximate solution of time dependent problems No. 10 (International Council of Scientific Unions, World Meteorological Organization)

    work page 1973

  53. [53]

    2025a, Phys

    Kuan, H.-J., Kiuchi, K., & Shibata, M. 2025a, Phys. Rev. Lett., 135, 141403, doi: 10.1103/j3zk-z17h

    work page doi:10.1103/j3zk-z17h

  54. [54]

    The error budget of binary neutron star merger simulations for configurations with high spin

    Kuan, H.-J., Markin, I., Ujevic, M., et al. 2025b, https://arxiv.org/abs/2506.02115

    work page internal anchor Pith review Pith/arXiv arXiv

  55. [55]

    1998, Astrophys

    Li, L.-X., & Paczynski, B. 1998, Astrophys. J. Lett., 507, L59, doi: 10.1086/311680

    work page doi:10.1086/311680 1998

  56. [56]

    J., & Brown, D

    Lindblom, L., Owen, B. J., & Brown, D. A. 2008, Phys. Rev. D, 78, 124020, doi: 10.1103/PhysRevD.78.124020

    work page doi:10.1103/physrevd.78.124020 2008

  57. [57]

    2016, Class

    Lovelace, G., et al. 2016, Class. Quant. Grav., 33, 244002, doi: 10.1088/0264-9381/33/24/244002

    work page doi:10.1088/0264-9381/33/24/244002 2016

  58. [58]

    2020, JCAP, 03, 050, doi: 10.1088/1475-7516/2020/03/050

    Maggiore, M., et al. 2020, JCAP, 03, 050, doi: 10.1088/1475-7516/2020/03/050

    work page doi:10.1088/1475-7516/2020/03/050 2020

  59. [59]

    Metzger, B. D. 2020, Living Rev. Rel., 23, 1, doi: 10.1007/s41114-019-0024-0

    work page doi:10.1007/s41114-019-0024-0 2020

  60. [60]

    MNRAS , author =

    Metzger, B. D., Martinez-Pinedo, G., Darbha, S., et al. 2010, Mon. Not. Roy. Astron. Soc., 406, 2650, doi: 10.1111/j.1365-2966.2010.16864.x

    work page doi:10.1111/j.1365-2966.2010.16864.x 2010

  61. [61]

    2005, title The three-point function in large-scale structure: redshift distortions and galaxy bias , , 361, 824, 10.1111/j.1365-2966.2005.09234.x

    Mignone, A., & Bodo, G. 2005, Mon. Not. Roy. Astron. Soc., 364, 126, doi: 10.1111/j.1365-2966.2005.09546.x

    work page doi:10.1111/j.1365-2966.2005.09546.x 2005

  62. [62]

    2015, Class

    Nakano, H. 2015, Class. Quant. Grav., 32, 177002, doi: 10.1088/0264-9381/32/17/177002

    work page doi:10.1088/0264-9381/32/17/177002 2015

  63. [63]

    J., Tootle, S

    Papenfort, L. J., Tootle, S. D., Grandcl´ ement, P., Most, E. R., & Rezzolla, L. 2021, Phys. Rev. D, 104, 024057, doi: 10.1103/PhysRevD.104.024057

    work page doi:10.1103/physrevd.104.024057 2021

  64. [64]

    A Metric for Performance Portability

    Pennycook, S. J., Sewall, J. D., & Lee, V. W. 2016, arXiv e-prints, arXiv:1611.07409, doi: 10.48550/arXiv.1611.07409

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1611.07409 2016

  65. [65]

    Prather, B. S. 2025, KHARMA: Flexible, Portable Performance for GRMHD, doi: 10.1007/978-981-97-8522-3 5 P¨ urrer, M., & Haster, C.-J. 2020, Phys. Rev. Res., 2, 023151, doi: 10.1103/PhysRevResearch.2.023151

    work page doi:10.1007/978-981-97-8522-3 2025

  66. [66]

    2025, Class

    Radice, D., Gamba, R., Zhu, H., & Rashti, A. 2025, Class. Quant. Grav., 42, 185003, doi: 10.1088/1361-6382/adfffa

    work page doi:10.1088/1361-6382/adfffa 2025

  67. [67]

    2025, Phys

    Rashti, A., Gamba, R., Chandra, K., et al. 2025, Phys. Rev. D, 111, 104078, doi: 10.1103/n5pz-qv3x

    work page doi:10.1103/n5pz-qv3x 2025

  68. [68]

    2011, Class

    Reisswig, C., & Pollney, D. 2011, Class. Quant. Grav., 28, 195015, doi: 10.1088/0264-9381/28/19/195015

    work page doi:10.1088/0264-9381/28/19/195015 2011

  69. [69]

    A., et al

    Scheel, M. A., et al. 2025, Class. Quant. Grav., 42, 195017, doi: 10.1088/1361-6382/adfd34

    work page doi:10.1088/1361-6382/adfd34 2025

  70. [70]

    R., et al

    Shankar, S., M¨ osta, P., Brandt, S. R., et al. 2023, Class. Quant. Grav., 40, 205009, doi: 10.1088/1361-6382/acf2d9

    work page doi:10.1088/1361-6382/acf2d9 2023

  71. [71]

    2016, Numerical Relativity, doi: 10.1142/9692

    Shibata, M. 2016, Numerical Relativity, doi: 10.1142/9692

    work page doi:10.1142/9692 2016

  72. [72]

    2019, Ann

    Shibata, M., & Hotokezaka, K. 2019, Ann. Rev. Nucl. Part. Sci., 69, 41, doi: 10.1146/annurev-nucl-101918-023625

    work page doi:10.1146/annurev-nucl-101918-023625 2019

  73. [73]

    1995, Phys

    Shibata, M., & Nakamura, T. 1995, Phys. Rev. D, 52, 5428, doi: 10.1103/PhysRevD.52.5428

    work page doi:10.1103/physrevd.52.5428 1995

  74. [74]

    M., Mullen, P

    Stone, J. M., Mullen, P. D., Fielding, D., et al. 2026, ApJS, 283, 27, doi: 10.3847/1538-4365/ae3717

    work page doi:10.3847/1538-4365/ae3717 2026

  75. [75]

    R., Lebrun-Grandie, D., Arndt, D., et al

    Trott, C., Lebrun-Grandie, D., Arndt, D., et al. 2022, IEEE Transactions on Parallel and Distributed Systems, 33, 805, doi: 10.1109/TPDS.2021.3097283

    work page doi:10.1109/tpds.2021.3097283 2022

  76. [76]

    Typel, S., Ropke, G., Klahn, T., Blaschke, D., & Wolter, H. H. 2010, Phys. Rev. C, 81, 015803, doi: 10.1103/PhysRevC.81.015803

    work page doi:10.1103/physrevc.81.015803 2010

  77. [77]

    2008, Phys

    Yamamoto, T., Shibata, M., & Taniguchi, K. 2008, Phys. Rev. D, 78, 064054, doi: 10.1103/PhysRevD.78.064054

    work page doi:10.1103/physrevd.78.064054 2008

  78. [78]

    AMReX: Block-Structured Adaptive Mesh Refinement for Multiphysics Applications

    Zhang, W., Myers, A., Gott, K., Almgren, A., & Bell, J. 2020, arXiv e-prints, arXiv:2009.12009, doi: 10.48550/arXiv.2009.12009

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2009.12009 2020

  79. [79]

    2025, Astrophys

    Zhu, H., Fields, J., Zappa, F., et al. 2025, Astrophys. J. Suppl., 278, 50, doi: 10.3847/1538-4365/adcf96

    work page doi:10.3847/1538-4365/adcf96 2025