Mixed-Precision For Energy Efficient Computations

2026-06-26

AI Summary

This study addresses the substantial time and energy costs associated with high-precision scientific computing. The authors propose a strategy that integrates mixed-precision computation with computer arithmetic optimizations to significantly improve energy efficiency while rigorously preserving numerical accuracy. Experimental evaluation on two representative scientific computing benchmarks—ReactorSimulator and LULESH—demonstrates the effectiveness of the approach: on ReactorSimulator, both execution time and energy consumption are reduced by 30%, while on LULESH, execution time decreases by 30% and energy usage drops by 25%. These results validate the method’s practicality and efficacy for enabling high-fidelity, energy-efficient simulations in scientific computing.

Abstract

As simulations grow more realistic, the pursuit of higher accuracy results in extended computation times and substantial power consumption. This study explores mixed-precision computing as a promising strategy to address these challenges, leveraging computer arithmetic tools to optimize performance. Using Reactor Simulator and LULESH benchmarks as case studies, we evaluated the potential of mixed-precision strategies to reduce both time-to-solution and energy-to-solution. For Reactor Simulator, we achieved a 30% reduction in both metrics without compromising accuracy. Similarly, for LULESH, results demonstrated up to a 30% improvement in time-to-solution and a 25% reduction in energy-to-solution.

Problem

• mixed-precision
• energy efficiency
• computational simulation
• accuracy
• power consumption

Innovation

• mixed-precision computing
• energy efficiency
• scientific simulations
• performance optimization
• computer arithmetic