pith. sign in

arxiv: 2606.08515 · v1 · pith:5ELRU53Tnew · submitted 2026-06-07 · 💻 cs.DC

Unifying von-Neumann HPC and Neuromorphic Acceleration via the EBRAINS Research Infrastructure: A Framework for High-Performance Workflows

Krishna Kant Singh , Charl Linssen , Eric M\"uller , Eleni Mathioulaki , Wouter Klijn , Lena Oden This is my paper

Pith reviewed 2026-06-27 18:02 UTC · model grok-4.3

classification 💻 cs.DC
keywords computational neuroscienceneuromorphic computingHPC workflowscontainerizationdomain-specific languagesreproducibilityspiking neural networks
0
0 comments X

The pith

A single EBRAINS workflow lets the same spiking neural network model run on both supercomputers and neuromorphic hardware.

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

The paper presents a cloud-based system that dispatches the identical computational neuroscience model to traditional HPC sites and to a neuromorphic platform without rewriting code or maintaining separate pipelines. It relies on federated identity, PyUNICORE for HPC job submission, a dedicated interface for neuromorphic hardware, Apptainer containers that are pulled on demand, and the NESTML language that compiles custom neuron models to either backend. The approach is demonstrated on a balanced random network and is intended to eliminate version drift and site-specific maintenance for cross-architecture reproducibility.

Core claim

The central claim is that containerization together with a domain-specific language for neuron models creates a practical, end-to-end path for hardware-agnostic execution of spiking neural networks: the same source model is compiled and run on both von-Neumann HPC resources and the SpiNNaker-1 neuromorphic system through a unified JupyterLab interface on EBRAINS.

What carries the argument

The unified EBRAINS JupyterLab workflow that routes jobs via PyUNICORE to HPC sites and via the Neuromorphic Computing Platform Interface to SpiNNaker, combined with dynamically pulled PMIx-aware Apptainer containers and NESTML compilation to NEST or sPyNNaker backends.

If this is right

  • Researchers maintain only one version of a model instead of site-specific pipelines.
  • Software version drift between HPC sites is reduced by on-demand container pulls.
  • Custom neuron models written in NESTML become usable on both conventional and neuromorphic hardware without manual translation.
  • A single federated login suffices to launch jobs on both classes of architecture.

Where Pith is reading between the lines

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

  • The same container-plus-DSL pattern could be tested on other scientific domains that already span CPU and accelerator hardware.
  • If the container images become publicly archived, independent groups could replicate the exact environment without access to EBRAINS.
  • Performance comparisons between the two hardware classes become more direct because the model source remains identical.

Load-bearing premise

The chosen combination of PyUNICORE, the neuromorphic interface, Apptainer containers, and NESTML will deliver seamless portability and reproducibility across the tested platforms without hidden performance losses or integration failures.

What would settle it

Run the identical NESTML-defined balanced random network on both an HPC site and the SpiNNaker system and check whether the spike statistics and run times remain consistent within expected numerical tolerance; divergence or failure on one platform would falsify the portability claim.

read the original abstract

Modern scientific workflows increasingly span diverse computing architectures, yet executing a single computational model across disparate systems often forces researchers to maintain fragmented, site-specific pipelines. In this paper, we address this challenge within the domain of computational neuroscience by presenting a unified, cloud-based workflow orchestrated via EBRAINS JupyterLab. This workflow enables users to transparently execute spiking neural networks on both von-Neumann supercomputers and neuromorphic hardware. Using a single federated identity, the system dispatches jobs to HPC sites (JUSUF, Galileo100) via PyUNICORE and to the SpiNNaker-1 neuromorphic system via the Neuromorphic Computing Platform Interface. To guarantee cross-site reproducibility and mitigate software version drift, we utilize a zero-installation execution mode that dynamically pulls PMIx-aware Apptainer containers to HPC compute nodes. Furthermore, we demonstrate genuine model-level portability using the NESTML domain-specific language, allowing custom neuron models to be written once and automatically compiled for either the NEST (C++) or sPyNNaker backends. Validated with a balanced random network case study, this work illustrates a practical, end-to-end path for hardware-agnostic workflows while highlighting the critical role of containerization and domain-specific languages in achieving true cross-platform reproducibility.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The paper describes a cloud-based workflow orchestrated via EBRAINS JupyterLab that enables execution of spiking neural network models on both von-Neumann HPC systems (JUSUF, Galileo100 via PyUNICORE) and neuromorphic hardware (SpiNNaker-1 via the Neuromorphic Computing Platform Interface). It claims to achieve cross-site reproducibility through PMIx-aware Apptainer containers and model-level portability via the NESTML domain-specific language, which compiles to either NEST or sPyNNaker backends. The approach is validated with a balanced random network case study.

Significance. If the integration claims hold with supporting performance data, the work would demonstrate a practical engineering path for hardware-agnostic neuroscience workflows, underscoring the value of containerization and DSLs for reproducibility across disparate architectures. The implemented system with federated identity and zero-installation execution is a concrete contribution, though its impact depends on evidence of seamless operation.

major comments (1)
  1. [Abstract] Abstract (and implied validation section): the manuscript states that the framework was 'Validated with a balanced random network case study' yet provides no quantitative results, tables, or figures reporting wall-clock times, scaling behavior, spike-count agreement, container overhead, or performance parity between HPC and SpiNNaker-1 runs. This absence leaves the central claim of seamless hardware-agnostic execution and true cross-platform reproducibility without empirical support.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback and the recommendation for major revision. We agree that quantitative empirical support is necessary to substantiate the claims of seamless cross-platform execution and reproducibility. We address the single major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract (and implied validation section): the manuscript states that the framework was 'Validated with a balanced random network case study' yet provides no quantitative results, tables, or figures reporting wall-clock times, scaling behavior, spike-count agreement, container overhead, or performance parity between HPC and SpiNNaker-1 runs. This absence leaves the central claim of seamless hardware-agnostic execution and true cross-platform reproducibility without empirical support.

    Authors: We acknowledge the validity of this observation. While the manuscript describes the end-to-end workflow, the NESTML portability mechanism, and the balanced random network case study, it does not currently include the requested quantitative metrics or comparative figures. To provide the necessary empirical support for hardware-agnostic execution and cross-platform reproducibility, we will revise the validation section (and update the abstract accordingly) to incorporate tables and figures reporting wall-clock times, scaling behavior, spike-count agreement between NEST and sPyNNaker backends, container overhead on the HPC sites, and direct performance parity measurements across JUSUF, Galileo100, and SpiNNaker-1. revision: yes

Circularity Check

0 steps flagged

No circularity: engineering workflow description with no derivations or fitted predictions

full rationale

The paper describes an implemented system for cross-platform execution using PyUNICORE, Apptainer containers, NESTML, and the Neuromorphic Computing Platform Interface. No equations, parameters, or mathematical derivations appear in the abstract or described content. The single case study is referenced only at a high level without any fitted inputs, predictions, or reductions to prior definitions. Claims rest on engineering integration and reproducibility via containers/DSLs rather than any self-referential logic or self-citation chains that bear the central result. This is a standard non-finding for an applied systems paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an engineering framework paper describing a practical workflow implementation rather than a theoretical or empirical scientific result with derivations; no free parameters, axioms, or invented entities are introduced.

pith-pipeline@v0.9.1-grok · 5786 in / 1309 out tokens · 40076 ms · 2026-06-27T18:02:04.851821+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

17 extracted references · 10 canonical work pages · 1 internal anchor

  1. [1]

    Frontiers in Neuroscience12, 291 (2018)

    van Albada, S.J., et al.: Performance comparison of the digital neuromorphic hardware SpiNNaker and the neural network simulation software NEST for a full-scale cortical microcircuit model. Frontiers in Neuroscience12, 291 (2018). https://doi.org/10.3389/fnins.2018.00291

    work page doi:10.3389/fnins.2018.00291 2018

  2. [2]

    Imaging Neuroscience 2, 1–35 (2024).https://doi.org/10.1162/imag_a_00137

    Amunts, K., et al.: The coming decade of digital brain research: A vision for neuroscience at the intersection of technology and computing. Imaging Neuroscience 2, 1–35 (2024).https://doi.org/10.1162/imag_a_00137

    work page doi:10.1162/imag_a_00137 2024

  3. [3]

    In: International Conference on High Performance Computing & Simulation (HPCS)

    Benedyczak, K., et al.: Unicore 7—middleware services for distributed and feder- ated computing. In: International Conference on High Performance Computing & Simulation (HPCS). pp. 613–620. IEEE (2016)

    2016

  4. [4]

    Biological Cybernetics 104(4–5), 263–296 (2011).https://doi.org/10.1007/s00422-011-0435-9

    Brüderle, D., et al.: A comprehensive workflow for general-purpose neural modeling with highly configurable neuromorphic hardware systems. Biological Cybernetics 104(4–5), 263–296 (2011).https://doi.org/10.1007/s00422-011-0435-9

    work page doi:10.1007/s00422-011-0435-9 2011

  5. [5]

    Journal of Computational Neuroscience8(3), 183–208 (2000)

    Brunel, N.: Dynamics of sparsely connected networks of excitatory and inhibitory spiking neurons. Journal of Computational Neuroscience8(3), 183–208 (2000)

    2000

  6. [6]

    Frontiers in Neuroinformatics2, 388 (2009)

    Davison, A.P., et al.: PyNN: a common interface for neuronal network simulators. Frontiers in Neuroinformatics2, 388 (2009)

    2009

  7. [7]

    Proceedings of the IEEE102(5), 652–665 (2014)

    Furber, S.B., et al.: The SpiNNaker project. Proceedings of the IEEE102(5), 652–665 (2014)

    2014

  8. [8]

    Scholarpedia 2(4), 1430 (2007).https://doi.org/10.4249/scholarpedia.1430

    Gewaltig, M.O., Diesmann, M.: NEST (NEural Simulation Tool). Scholarpedia 2(4), 1430 (2007).https://doi.org/10.4249/scholarpedia.1430

    work page doi:10.4249/scholarpedia.1430 2007

  9. [9]

    Frontiers in Neuroinformatics19, 1544143 (2025)

    Linssen, C., et al.: NESTML: a generic modeling language and code generation tool for the simulation of spiking neural networks with advanced plasticity rules. Frontiers in Neuroinformatics19, 1544143 (2025). https://doi.org/10.3389/fninf.2025. 1544143

    work page doi:10.3389/fninf.2025 2025

  10. [10]

    Nature Communications15(1), 8122 (2024).https://doi.org/10.1038/s41467-024-52259-9

    Pedersen, J.E., et al.: Neuromorphic intermediate representation: A unified instruc- tion set for interoperable brain-inspired computing. Nature Communications15(1), 8122 (2024).https://doi.org/10.1038/s41467-024-52259-9

    work page doi:10.1038/s41467-024-52259-9 2024

  11. [11]

    Frontiers in Neuroscience16, 795876 (2022).https://doi.org/10.3389/ fnins.2022.795876

    Pehle, C., et al.: The BrainScaleS-2 accelerated neuromorphic system with hybrid plasticity. Frontiers in Neuroscience16, 795876 (2022).https://doi.org/10.3389/ fnins.2022.795876

    arXiv 2022

  12. [12]

    NESTML: a modeling language for spiking neurons

    Plotnikov, D., et al.: NESTML: a modeling language for spiking neurons. arXiv preprint (2016).https://doi.org/10.48550/arXiv.1606.02882

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1606.02882 2016

  13. [13]

    Frontiers in Neuroscience12, 816 (2018)

    Rhodes, O., et al.: sPyNNaker: a software package for running PyNN simulations on SpiNNaker. Frontiers in Neuroscience12, 816 (2018)

    2018

  14. [14]

    Neuromorph

    Schmidt, H., et al.: From clean room to machine room: commissioning of the first- generation BrainScaleS wafer-scale neuromorphic system. Neuromorph. Comput. Eng.3(3), 034013 (Sep 2023).https://doi.org/10.1088/2634-4386/acf7e4

    work page doi:10.1088/2634-4386/acf7e4 2023

  15. [15]

    In: High-Performance Scientific Computing (JHPCS 2016)

    Senk, J., et al.: A collaborative simulation-analysis workflow for computational neuroscience using HPC. In: High-Performance Scientific Computing (JHPCS 2016). Lecture Notes in Computer Science, vol. 10164, pp. 243–256. Cham (2017). https://doi.org/10.1007/978-3-319-53862-4_21

    work page doi:10.1007/978-3-319-53862-4_21 2016

  16. [16]

    Singh, K.K., et al.: HPC containers for EBRAINS: Towards portable cross-domain software environment (2026).https://doi.org/10.48550/arXiv.2603.12044, ac- cepted at HPCW’26 Unifying von-Neumann HPC and Neuromorphic Acceleration via EBRAINS 13

    work page doi:10.48550/arxiv.2603.12044 2026

  17. [17]

    Scientific Data3(1), 160018 (2016).https://doi.org/10.1038/ sdata.2016.18

    Wilkinson, M.D., et al.: The FAIR guiding principles for scientific data management and stewardship. Scientific Data3(1), 160018 (2016).https://doi.org/10.1038/ sdata.2016.18

    2016