Hardware-based systems for partial sorting of streaming data

Fast data sorting is an essential component of many high-performance computing systems. High-throughput and highly parallel sorting algorithms are very appropriate for devices which provide massive parallelism like FPGAs and APSoCs. One of the major drawbacks of these platforms is amount of available resources which is a serious obstacle for design of hardware sorters. However, for many practical applications complete sorting is not important and only partial sorting for extraction of maximum and minimum subsets of the data is required. In this paper we investigate the maximum and minimum extraction problem, present a hardware-based extracting circuit based on pipelined periodic sorting networks and compare it with known alternatives.