A Parallel Radix-Sort-Based VLSI Architecture for Finding the First Maximum/Minimum Values

Very-large-scale integration (VLSI) architectures for finding the first W (W>2) maximum (or minimum) values are required in the implementation of several applications such as nonbinary low-density-parity-check decoders, K-best multiple-input-multiple-output (MIMO) detectors, and turbo product codes. In this brief, a parallel radix-sort-based VLSI architecture for finding the first W maximum (or minimum) values is proposed. The described architecture, called Bit-Wise-And (BWA) architecture, relies on analyzing input data from the most significant bit to the least significant one, with very simple logic circuits. One key feature in the BWA architecture is its high level of scalability, which enables the adoption of this solution in a large spectrum of applications, corresponding to large ranges for both W and the size of the input data set. Experimental results, achieved by implementing the proposed architecture on a high-speed 90-nm CMOS standard-cell technology, show that BWA architecture requires significantly less area than other solutions available in the literature, i.e., less than or about 50% in all the considered cases and about 50% in the worst case. Moreover, the BWA architecture exhibits the lowest area-delay product among almost all considered cases.