Edge AI Acceleration Using OpenCL Framework on NXP i.MX 8M Nano SoM

Edge computing brings computational resources closer to the data source by enhancing processing capabilities and reducing the complexity and latency of data transmission. This approach reduces traffic volume and bandwidth requirements for IoT applications while improving response times for real-time operations. Modern edge computing must balance the need for energy efficiency with the demand for high computational performance, particularly in battery-powered devices. Advances in artificial intelligence have intensified the need for efficient inference on resource-constrained devices. Heterogeneous computing systems, which can offload tasks to optimized hardware components, offer a partial solution, though they add complexity for developers. The OpenCL framework helps mitigate this complexity by providing a standardized way to utilize heterogeneous systems. This paper explores the feasibility of performing AI inference on the NXP i.MX 8M Nano processor using OpenCL, addressing the challenges and potential benefits of deploying machine learning models on small” energy-efficient systems.