Interleaved entropy coders
Fabian Giesen

TL;DR
This paper explores the properties of ANS entropy coders that allow for symmetric encoding and decoding, enabling interleaving of multiple streams without extra metadata, which improves efficiency on modern hardware.
Contribution
It demonstrates how ANS coders' symmetry facilitates interleaving of multiple streams without extra metadata and extends this approach to other entropy coders with increased encoder complexity.
Findings
ANS coders allow perfect symmetry between encoder and decoder.
Multiple streams can be interleaved without additional metadata.
Interleaving can be generalized to other entropy coders with some complexity trade-offs.
Abstract
The ANS family of arithmetic coders developed by Jarek Duda has the unique property that encoder and decoder are completely symmetric in the sense that a decoder reading bits will be in the exact same state that the encoder was in when writing those bits---all "buffering" of information is explicitly part of the coder state and identical between encoder and decoder. As a consequence, the output from multiple ABS/ANS coders can be interleaved into the same bitstream without any additional metadata. This allows for very efficient encoding and decoding on CPUs supporting superscalar execution or SIMD instructions, as well as GPU implementations. We also show how interleaving without additional metadata can be implemented for any entropy coder, at some increase in encoder complexity.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Code & Models
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsComputability, Logic, AI Algorithms · Algorithms and Data Compression
