The Rise of Low Precision Floating Point

DOI10.48550/ARXIV.2611.00301MaRDI QIDQ7414968arXivFDO

Authors Michael L. Overton, Guy Lemieux

Publication date 30 October 2026

Published in arXiv

Tldr We survey the emerging trend of low-precision floating-point computing and its impact on deep learning applications and hardware design; this trend leads to a new generation of hardware architectures that are more energy-efficient than their 32-bit predecessors.

Full work available at URL https://doi.org/10.48550/ARXIV.2611.00301

arXiv classification

cs.ARcs.LGcs.MSmath.NA

Cites work

• Emulation of a FMA and Correctly Rounded Sums: Proved Algorithms Using Rounding to Odd
• ImageNet classification with deep convolutional neural networks
• Mixed-Precision Computing for Scientific Discovery: Formats, Co-Design, and Responsible Approximation
• Numerical computing with IEEE floating point arithmetic. Including one theorem, one rule of thumb, and one hundred and six exercises
• Stochastic Rounding: Algorithms and Hardware Accelerator
• Tesla

This page was built for publication: The Rise of Low Precision Floating Point

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7414968)