Merged Processing Element for Polar Code Decoder

KRASSER, F.; ARNONE, L. J.; COPPOLILLO, L.; LIBERATORI, M. C.; CASTIÑEIRA MOREIRA, J.

Bahia Blanca

Congreso; Information Processing and Control Workshop; 2019

UNS

Polar codes are a class of forward error correctioncodes that can provably achieve the capacity of discretememoryless channels. In some polar decoding schemes, successivecancellation decoders carry out estimations of the originalmessages, using update rules from two functions that involvearithmetic operations. Processing elements are basic blocks ofthe decoder that implement these functions. The interconnectionof a great number of these elements forms a network whichoperates on the log-likelihood ratios from the channel to finallyachieve the best possible estimation of the original messagebits, hence the importance of a thorough design of them. Anysimplification or resource saving in the design of these elementsdirectly impacts the hardware complexity of the decoder. Also,any improvement in their throughput should have a similareffect in the performance of the decoder. In this work wepresent two generic, fully combinational designs for processingelements. They take into consideration the possibility of overflowof the quantized values, and use pre-computation to reducelatency. These designs can be used in different hardware decoderarchitectures to improve their throughput and area requirement.