丸め方向間違いの少ない関数近似算法(実用)
スポンサーリンク
概要
- 論文の詳細を見る
This paper proposes a function approximation algorithm with reduced rounding mistakes for single precision IEEE format, without using double or higher intermediate data. The algorithm is suited for embedded micro-controllers without double precision floating-point hardware. With the new algorithm, 3-9 extra cycles reduces the ration of rounding mistakes to around 1%.
- 日本応用数理学会の論文
- 2006-09-25