English

UniSpike: Accelerating Spiking Neural Networks on Neuromorphic Systems via Eliminating Address Redundancy

Neural and Evolutionary Computing 2026-05-25 v1 Hardware Architecture

Abstract

Many-core neuromorphic systems accelerate Spiking Neural Networks (SNNs), yet their packet-based spike communication can spend substantial traffic and energy repeatedly transmitting destination addresses. This overhead is amplified by the small payload of spike packets: in representative workloads, duplicate address transmissions account for up to 49% of the total traffic. This paper presents UniSpike, a hardware-software co-design that removes address redundancy by aggregating spikes destined for the same core into compact packets. UniSpike combines destination-centric spike scheduling, lightweight runtime packet assembly hardware, and destination-aware SNN partitioning. Across diverse SNN workloads, UniSpike reduces traffic by 1.93×\times on average, delivering 1.77×\times speedup and 1.50×\times energy efficiency improvement over state-of-the-art designs.

Keywords

Cite

@article{arxiv.2605.23796,
  title  = {UniSpike: Accelerating Spiking Neural Networks on Neuromorphic Systems via Eliminating Address Redundancy},
  author = {Qinghui Xing and Zhuo Chen and Xin Du and Ouwen Jin and Ming Zhang and Pan Lv and Ying Li and Shuiguang Deng and Gang Pan},
  journal= {arXiv preprint arXiv:2605.23796},
  year   = {2026}
}

Comments

Accepted to the 63rd Design Automation Conference (DAC 2026)