SKYLIGHT: A S ca l able H undred-Channel 3D Photonic I n-Memory T ensor Core Architecture for Real-time AI Inference
ACM Transactions on Design Automation of Electronic Systems · 31.03.2027 · 32 (2) · 1-33
DOI: 10.1145/3831711 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM Transactions on Design Automation of Electronic Systems, online erschienen am 13.11.2026. Verfasst von Meng Zhang und 7 weiteren Autor:innen. Die Arbeit nennt 108 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The growing computational demands of artificial intelligence (AI) are challenging conventional electronics, making photonic computing a promising alternative. However, existing photonic architectures face fundamental scalability and reliability barriers. This article introduces SKYLIGHT , a scalable 3D photonic in-memory tensor core architecture designed for real-time AI inference. By co-designing its topology, wavelength routing, accumulation, and programming in a 3D stack, SKYLIGHT overcomes key limitations. Its innovations include a low-loss 3D Si/SiN crossbar topology, a thermally robust non-micro-ring resonator (MRR)-based wavelength-division multiplexing (WDM) component, a hierarchical signal accumulation using a multi-port photodetector (PD), and optically programmed non-volatile phase-change material (PCM) weights. Importantly, SKYLIGHT enables in-situ weight updates that support label-free, layer-local learning (e.g., forward-forward local updates) in addition to inference. With comprehensive system-level modeling, we show that a single 144× 256 SKYLIGHT core achieves a peak compute capability of 342.1 TOPS at 23.7 TOPS/W under a high-reuse tensor-core operating point. For end-to-end ResNet-50 inference, where the same physical core is reprogrammed across layers and tiles, SKYLIGHT reaches 1212 FPS with 60.9 mJ/image after explicitly accounting for PCM programming energy. System-level evaluations on four representative machine learning tasks, including unsupervised local self-learning, demonstrate SKYLIGHT ’s robustness to realistic hardware non-idealities (low-bit quantization and signal-proportional analog noise capturing modulation, PCM programming, and readout variations). With noise-aware training, SKYLIGHT maintains high task accuracy, validating its potential as a comprehensive solution for energy-efficient, large-scale photonic AI accelerators.
Zitieren
APA
Zhang, M., Yin, Z., Gangi, N., Chen, A., Bamfo, B., Xu, T., Gu, J., & Huang, R. (2027). SKYLIGHT: A S ca l able H undred-Channel 3D Photonic I n-Memory T ensor Core Architecture for Real-time AI Inference. ACM Transactions on Design Automation of Electronic Systems, 32(2), 1-33. https://doi.org/10.1145/3831711
Vancouver
Zhang M, Yin Z, Gangi N, Chen A, Bamfo B, Xu T, et al. SKYLIGHT: A S ca l able H undred-Channel 3D Photonic I n-Memory T ensor Core Architecture for Real-time AI Inference. ACM Transactions on Design Automation of Electronic Systems. 2027;32(2):1-33. doi:10.1145/3831711
BibTeX
@article{Zhang2027,
author = {Zhang, Meng and Yin, Ziang and Gangi, Nicholas and Chen, Alexander and Bamfo, Brett and Xu, Tianle and Gu, Jiaqi and Huang, Rena},
title = {SKYLIGHT: A S ca l able H undred-Channel 3D Photonic I n-Memory T ensor Core Architecture for Real-time AI Inference},
journal = {ACM Transactions on Design Automation of Electronic Systems},
year = {2027},
volume = {32},
number = {2},
pages = {1--33},
doi = {10.1145/3831711}
}
EUVIMED