CrossLight: A Cross-Layer Optimized Silicon Photonic Neural Network Accelerator
This work addresses the need for more efficient neural network accelerators for AI applications, representing an incremental advancement in photonic accelerator design.
The paper tackled the problem of improving energy efficiency and performance in neural network accelerators by proposing CrossLight, a cross-layer optimized silicon photonic accelerator, which achieved 9.5x lower energy-per-bit and 15.9x higher performance-per-watt at 16-bit resolution compared to state-of-the-art photonic deep learning accelerators.
Domain-specific neural network accelerators have seen growing interest in recent years due to their improved energy efficiency and inference performance compared to CPUs and GPUs. In this paper, we propose a novel cross-layer optimized neural network accelerator called CrossLight that leverages silicon photonics. CrossLight includes device-level engineering for resilience to process variations and thermal crosstalk, circuit-level tuning enhancements for inference latency reduction, and architecture-level optimization to enable higher resolution, better energy-efficiency, and improved throughput. On average, CrossLight offers 9.5x lower energy-per-bit and 15.9x higher performance-per-watt at 16-bit resolution than state-of-the-art photonic deep learning accelerators.