Floating-point autotuning with customized precisions
Abstract
Reduced-precision arithmetic offers significant opportunities to improve performance, memory usage, and energy efficiency in numerical applications, provided that numerical accuracy is preserved. This work investigates automated precision tuning through customized floating-point formats with user-defined exponent and significand sizes, enabling the emulation of emerging low-precision formats and the exploration of non-standard precision configurations within a unified mixed-precision framework. The proposed methodology, implemented in the PROMISE precision autotuning tool, combines numerical validation with a systematic search to generate program variants that satisfy user-defined accuracy requirements. To address the computational cost of this exploration, a containerized benchmarking framework supports parallel execution across multiple algorithms and parameter configurations. The approach is evaluated on a suite of numerical programs, including linear solvers and applications from the Rodinia benchmark. Results show that a substantial proportion of variables can be safely reduced to lower precision while preserving accuracy, indicating that standard double precision is often over-provisioned. These findings highlight the potential of automated precision tuning to derive efficient mixed-precision configurations tailored to application-specific accuracy requirements.
Related
- The Entropy of Floating-Point Numbers
- The Rise of Low Precision Floating Point
- Performance Evaluation of an Adaptive Quadrature and a Double Exponential Formula Using Arbitrary-Precision Floating-Point Arithmetic
- Analysis and Detection of Stagnation in Floating-Point Summation
- FPScan: An Automated Constraint-Based Analyzer for Floating-Point Anomaly Detection
BibTeX
@misc{chen2026floatingpoint,
title = {{Floating-point autotuning with customized precisions}},
author = {Xinye Chen and Fabienne Jézéquel and Thibault Hilaire},
howpublished = {ACM Transactions on Mathematical Software},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2606.08339},
eprint = {2606.08339},
archivePrefix = {arXiv},
url = {https://hal.science/hal-05651002v2},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Floating-point autotuning with customized precisions
AU - Xinye Chen
AU - Fabienne Jézéquel
AU - Thibault Hilaire
PY - 2026
DA - 2026/10/29
PB - ACM Transactions on Mathematical Software
DO - 10.48550/arxiv.2606.08339
UR - https://hal.science/hal-05651002v2
AB - Reduced-precision arithmetic offers significant opportunities to improve performance, memory usage, and energy efficiency in numerical applications, provided that numerical accuracy is preserved. This work investigates automated precision tuning through customized floating-point formats with user-defined exponent and significand sizes, enabling the emulation of emerging low-precision formats and the exploration of non-standard precision configurations within a unified mixed-precision framework. The proposed methodology, implemented in the PROMISE precision autotuning tool, combines numerical validation with a systematic search to generate program variants that satisfy user-defined accuracy requirements. To address the computational cost of this exploration, a containerized benchmarking framework supports parallel execution across multiple algorithms and parameter configurations. The approach is evaluated on a suite of numerical programs, including linear solvers and applications from the Rodinia benchmark. Results show that a substantial proportion of variables can be safely reduced to lower precision while preserving accuracy, indicating that standard double precision is often over-provisioned. These findings highlight the potential of automated precision tuning to derive efficient mixed-precision configurations tailored to application-specific accuracy requirements.
ER -