Skip to main content

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

Publication ,  Conference
Nathan, R; Anthonio, B; Lu, SL; Naeimi, H; Sorin, DJ; Sun, X
Published in: International Conference for High Performance Computing, Networking, Storage and Analysis, SC
January 16, 2014

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.

Duke Scholars

Altmetric Attention Stats
Dimensions Citation Stats

Published In

International Conference for High Performance Computing, Networking, Storage and Analysis, SC

DOI

EISSN

2167-4337

ISSN

2167-4329

Publication Date

January 16, 2014

Volume

2015-January

Issue

January

Start / End Page

117 / 127
 

Citation

APA
Chicago
ICMJE
MLA
NLM
Nathan, R., Anthonio, B., Lu, S. L., Naeimi, H., Sorin, D. J., & Sun, X. (2014). Recycled Error Bits: Energy-Efficient Architectural Support for Floating Point Accuracy. In International Conference for High Performance Computing, Networking, Storage and Analysis, SC (Vol. 2015-January, pp. 117–127). https://doi.org/10.1109/SC.2014.15
Nathan, R., B. Anthonio, S. L. Lu, H. Naeimi, D. J. Sorin, and X. Sun. “Recycled Error Bits: Energy-Efficient Architectural Support for Floating Point Accuracy.” In International Conference for High Performance Computing, Networking, Storage and Analysis, SC, 2015-January:117–27, 2014. https://doi.org/10.1109/SC.2014.15.
Nathan R, Anthonio B, Lu SL, Naeimi H, Sorin DJ, Sun X. Recycled Error Bits: Energy-Efficient Architectural Support for Floating Point Accuracy. In: International Conference for High Performance Computing, Networking, Storage and Analysis, SC. 2014. p. 117–27.
Nathan, R., et al. “Recycled Error Bits: Energy-Efficient Architectural Support for Floating Point Accuracy.” International Conference for High Performance Computing, Networking, Storage and Analysis, SC, vol. 2015-January, no. January, 2014, pp. 117–27. Scopus, doi:10.1109/SC.2014.15.
Nathan R, Anthonio B, Lu SL, Naeimi H, Sorin DJ, Sun X. Recycled Error Bits: Energy-Efficient Architectural Support for Floating Point Accuracy. International Conference for High Performance Computing, Networking, Storage and Analysis, SC. 2014. p. 117–127.

Published In

International Conference for High Performance Computing, Networking, Storage and Analysis, SC

DOI

EISSN

2167-4337

ISSN

2167-4329

Publication Date

January 16, 2014

Volume

2015-January

Issue

January

Start / End Page

117 / 127