Lessons Learned on the Path to Guaranteeing the Error Bound in Lossy Quantizers

TL;DR

This paper addresses the challenge of guaranteeing error bounds in lossy quantizers, presenting solutions within the LC framework that ensure error bounds are always maintained without sacrificing compression efficiency or throughput.

Contribution

It introduces novel techniques in the LC framework to reliably guarantee error bounds across various quantizers, improving reliability in lossy compression.

Findings

Error bounds are sometimes violated in existing compressors.

The proposed solutions guarantee error bounds for all supported quantizers.

High compression ratios and throughput are maintained with the new solutions.

Abstract

Rapidly increasing data sizes in scientific computing are the driving force behind the need for lossy compression. The main drawback of lossy data compression is the introduction of error. This paper explains why many error-bounded compressors occasionally violate the error bound and presents the solutions we use in LC, a CPU/GPU compatible lossy compression framework, to guarantee the error bound for all supported types of quantizers. We show that our solutions maintain high compression ratios and cause no appreciable change in throughput.