Recycled Error Bits: Energy-Efficient Architectural Support for Floating Point Accuracy

Conference Paper

In this work, we provide energy-efficient architectural support for floating point accuracy. For each floating point addition performed, we "recycle" that operation's rounding error. We make this error architecturally visible such that it can be used, whenever desired, by software. We also design a compiler pass that allows software to automatically use this feature. Experimental results on physical hardware show that software that exploits architecturally recycled error bits can (a) achieve accuracy comparable to a 64-bit FPU with performance and energy that are comparable to a 32-bit FPU, and (b) achieve accuracy comparable to an all-software scheme for 128-bit accuracy with far better performance and energy usage.

Full Text

Duke Authors

Cited Authors

  • Nathan, R; Anthonio, B; Lu, SL; Naeimi, H; Sorin, DJ; Sun, X

Published Date

  • January 16, 2014

Published In

Volume / Issue

  • 2015-January / January

Start / End Page

  • 117 - 127

Electronic International Standard Serial Number (EISSN)

  • 2167-4337

International Standard Serial Number (ISSN)

  • 2167-4329

Digital Object Identifier (DOI)

  • 10.1109/SC.2014.15

Citation Source

  • Scopus