Authors: Brett Saiki, Oliver Flatt, Chandrakana Nandi, Pavel Panchekha, Zachary Tatlock

Venue: IEEE Symposium on Computer Arithmetic (ARITH) 2021

Links: Paper | Talk

Abstract

Precision tuning and rewriting can improve both the accuracy and speed of floating point expressions, yet these techniques are typically applied separately. This paper explores how finer-grained interleaving of precision tuning and rewriting can help automatically generate a richer set of Pareto-optimal accuracy versus speed trade-offs.

We introduce Pherbie (pareto Herbie), a tool providing both precision tuning and rewriting, and evaluate interleaving these two strategies at different granularities. Our results demonstrate that finer-grained interleavings improve both the Pareto curve of candidate implementations and overall optimization time. On a popular set of tests from the FPBench suite, Pherbie finds both implementations that are significantly more accurate for a given cost and significantly faster for a given accuracy bound compared to baselines using precision tuning and rewriting alone or in sequence.

@INPROCEEDINGS{9603367,
  author={Saiki, Brett and Flatt, Oliver and Nandi, Chandrakana and Panchekha, Pavel and Tatlock, Zachary},
  booktitle={2021 IEEE 28th Symposium on Computer Arithmetic (ARITH)}, 
  title={Combining Precision Tuning and Rewriting}, 
  year={2021},
  volume={},
  number={},
  pages={1-8},
  keywords={Costs;Tools;Digital arithmetic;Tuning;Optimization;precision tuning;term rewriting;optimization},
  doi={10.1109/ARITH51176.2021.00013}}