Multi FPGA

Multi-FPGA systems are emerging as powerful accelerators for computationally intensive tasks, particularly in deep learning and related fields. Current research focuses on optimizing the implementation of various neural network architectures, including transformers (like LLMs), convolutional neural networks (CNNs), and graph neural networks (GNNs), across multiple FPGAs to achieve high throughput and energy efficiency. This approach offers significant advantages in resource-constrained environments like edge devices and embedded systems, impacting areas such as robotics, autonomous vehicles, and high-energy physics by enabling real-time processing of complex data streams. The development of efficient communication strategies and hardware-software co-design methodologies are key challenges driving ongoing research.

Papers