A Multi-Core FPGA-based 2D-Clustering Algorithm for High-Throughput Data Intensive Applications

A multi-core FPGA-based clustering algorithm for high-throughput data intensive applications is presented. The algorithm is optimized for data with two dimensional organization (e.g. image processing, tracking devices for high energy physics detectors etc.) and uses a moving window of generic size to adjust to the application’s processing requirements (the cluster sizes and shapes that appear in the input data sets). One or more windows (cores) can be used to identify clusters in parallel, allowing for versatility to increase performance or reduce the amount of used resources. In addition to the inherent parallelism the algorithm is executed in a pipeline, thus allowing for readout to be performed without imposing extra combinatorial delays.