Architecture-aware Precision Tuning with Multiple Number Representation Systems

Cattaneo, Daniele; Chiari, Michele; Fossati, Nicola; Cherubin, Stefano; Agosta, Giovanni

Abstract

Precision tuning trades accuracy for speed and energy savings, usually by reducing the data width, or by switching from floating point to fixed point representations. However, comparing the precision across different representations is a difficult task. We present a metric that enables this comparison, and employ it to build a methodology based on Integer Linear Programming for tuning the data type selection. We apply the proposed metric and methodology to a range of processors, demonstrating an improvement in performance (up to 9×) with a very limited precision loss (<2.8% for 90% of the benchmarks) on the PolyBench benchmark suite.

Citation

Cattaneo, D., Chiari, M., Fossati, N., Cherubin, S., & Agosta, G. (2021, December). Architecture-aware Precision Tuning with Multiple Number Representation Systems. Presented at 2021 58th ACM/IEEE Design Automation Conference (DAC), San Francisco, CA, USA

• Presentation Conference Type: Conference Paper (Published)
• Conference Name: 2021 58th ACM/IEEE Design Automation Conference (DAC)
• Start Date: Dec 5, 2021
• End Date: Dec 9, 2021
• Acceptance Date: Feb 12, 2021
• Online Publication Date: Nov 13, 2021
• Publication Date: 2021
• Deposit Date: Apr 1, 2022
• Publicly Available Date: Apr 1, 2022
• Publisher: Institute of Electrical and Electronics Engineers
• Series ISSN: 0738-100X
• Book Title: 2021 58th ACM/IEEE Design Automation Conference (DAC)
• ISBN: 978-1-6654-3274-0
• DOI: https://doi.org/10.1109/dac18074.2021.9586303
• Public URL: http://researchrepository.napier.ac.uk/Output/2859689