FPGA-Accelerated SpeckleNN with SNL for Real-time X-ray Single-Particle Imaging

We implement a specialized version of our SpeckleNN model for real-time speckle pattern classification in X-ray Single-Particle Imaging (SPI) using the SLAC Neural Network Library (SNL) on an FPGA. This hardware is optimized for inference near detectors in high-throughput X-ray free-electron laser (XFEL) facilities like the Linac Coherent Light Source (LCLS). To fit FPGA constraints, we optimized SpeckleNN, reducing parameters from 5.6M to 64.6K (98.8% reduction) with 90% accuracy. We also compressed the latent space from 128 to 50 dimensions. Deployed on a KCU1500 FPGA, the model used 71% of DSPs, 75% of LUTs, and 48% of FFs, with an average power consumption of 9.4W. The FPGA achieved 45.015us inference latency at 200 MHz. On an NVIDIA A100 GPU, the same inference consumed ~73W and had a 400us latency. Our FPGA version achieved an 8.9x speedup and 7.8x power reduction over the GPU. Key advancements include model specialization and dynamic weight loading through SNL, eliminating time-consuming FPGA re-synthesis for fast, continuous deployment of (re)trained models. These innovations enable real-time adaptive classification and efficient speckle pattern vetoing, making SpeckleNN ideal for XFEL facilities. This implementation accelerates SPI experiments and enhances adaptability to evolving conditions.

Analysis of Hardware Synthesis Strategies for Machine Learning in Collider Trigger and Data Acquisition

To fully exploit the physics potential of current and future high energy particle colliders, machine learning (ML) can be implemented in detector electronics for intelligent data processing and acquisition. The implementation of ML in real-time at colliders requires very low latencies that are unachievable with a software-based approach, requiring optimization and synthesis of ML algorithms for deployment on hardware. An analysis of neural network inference efficiency is presented, focusing on the application of collider trigger algorithms in field programmable gate arrays (FPGAs). Trade-offs are evaluated between two frameworks, the SLAC Neural Network Library (SNL) and hls4ml, in terms of resources and latency for different model sizes. Results highlight the strengths and limitations of each approach, offering valuable insights for optimizing real-time neural network deployments at colliders. This work aims to guide researchers and engineers in selecting the most suitable hardware and software configurations for real-time, resource-constrained environments.