An Entropy Coding Based on Binary Encoding for Mixed-Radix Digits

TL;DR

This paper introduces a binary encoding method for mixed-radix digits and a variant of rANS coding supporting parallel decoding, achieving higher throughput without sacrificing compression efficiency.

Contribution

It presents a novel binary encoding for mixed-radix digits and a parallel decoding rANS variant, improving speed over existing methods.

Findings

Higher throughput in serial mode (about 2X speed-up)

No loss in compression ratio

Outperforms existing 2-way interleaving implementations

Abstract

The necessity of radix conversion of numeric data is an indispensable component in any complete analysis of digital computation. In this paper, we propose a binary encoding for mixed-radix digits. Second, a variant of rANS coding based on this conversion is given, which supports parallel decoding. The simulations show that the proposed coding in serial mode has a higher throughput than the baseline (with the speed-up factor about 2X) without loss of compression ratio, and it outperforms the existing 2-way interleaving implementation.