Programming Vision Pipelines on AMD’s AI Engines

AMD’s latest generation of Adaptive Compute Acceleration Platforms (ACAP), Xilinx Versal AI Core and Xilinx Versal AI Edge, include an array of powerful AI Engines alongside other computation components, such as programmable logic and ARM cores. This array of AI Engines has high computational capability to address the workloads of diverse applications, including automotive solutions. This presentation introduces the properties and capabilities of these AI Engines for image, video and vision processing. We’ll begin with a top-down look at how video data makes its way to the AI Engines. Then we’ll delve into a detailed discussion of the compute properties of the VLIW vector architecture of the AI Engines and illustrate how it efficiently executes vision processing kernels. Next, we’ll introduce the Xilinx Vitis Vision Library and give an overview of its data movement and kernel processing capabilities. We’ll conclude by showing how AMD’s Xilinx Vitis tools support building a vision pipeline and analyzing its performance.