2024-03-29T06:18:02Zhttps://ipsj.ixsq.nii.ac.jp/ej/?action=repository_oaipmhoai:ipsj.ixsq.nii.ac.jp:000229992023-04-27T10:00:04Z01164:01579:01585:01590
配列整数演算の精度、性能評価Evaluation of accuracy and performance of Array-integer operationjpnhttp://id.nii.ac.jp/1001/00022999/Technical Reporthttps://ipsj.ixsq.nii.ac.jp/ej/?action=repository_action_common_download&item_id=22999&item_no=1&attribute_id=1&file_no=1Copyright (c) 2007 by the Information Processing Society of Japan(株)日立製作所 ソフトウエア事業部濱口, 信行数値計算で使用される浮動小数点演算では、演算誤差を伴う為に、低い精度の計算結果をもたらす事がある。この問題に対し、4倍精度以上の演算を行う多倍長ライブラリを試作し、精度、性能評価を行った結果、浮動小数点加減算を整数演算に置き換えた配列整数演算は非常に有効で、応用範囲が広い事が、判明した。Floating point operations used in numeric calculation may result in lower precision because of errors in operations. We have developed the Multi-precision library that calculates floating point data of quadruple or higher precision using the Array-inteter operation, which replaces addition and subtraction of floating point data to arithmetic operation of integer data. We have evaluated the accuracy of computed values and the performance, and come to the conclusion that Array-integer operation is very effective and is applicable for a wide area.AN10096105情報処理学会研究報告計算機アーキテクチャ(ARC)200717(2007-ARC-172)1331382007-03-012009-06-30