High Speed VLSI Architecture for Finding the First W Maximum/Minimum Values

VLSI architectures for finding the first W maximum/minimum values are highly demanded in the fields of K-best MIMO detector, non-binary LDPC decoder and product-code decoder. In this paper, a VLSI architecture based on parallel comparing scheme is explored for finding the first W maximum/minimum values from M inputs. The place and route results using a TSMC 90-nm CMOS technology show that, despite some hardware cost, it achieves on average a 3.6x faster speed performance compared to the existing partial sorting architectures.