摘要
研究计算机内部二进制浮点数IEEE754存储规则及相加过程,给出数值计算中两同号规范化数相加发生"大数吃小数"现象的严格理论界限,为实际数值计算中避免此类现象提供理论依据,并利用所得理论对数值试验中的现象及结论进行解释.
By analyzing about storage pattern and addition operation of format binary floating-points based on IEEE754 standard, this paper found strict theoretical bounds for avoiding the phenomenon that a large number annihilating a small number in numerical computations when both of these two numbers are positive or negative. The bounds can be applied in practical numerical computations for any machine precision and any programming language. Moreover, the conclusions in this article give theoretical explanations for the phenomenons in numerical experiments.
出处
《天津师范大学学报（自然科学版）》
CAS
2014年第4期25-27,共3页
Journal of Tianjin Normal University：Natural Science Edition
基金
国家自然科学基金资助项目(41375110)
中国科学院大气物理研究所大气科学和地球流体力学数值模拟国家重点实验室(LASG)2012年度开放课题
关键词
“大数吃小数”
数值计算
IEEE754标准
规范化二进制浮点数
a large number annihilating a small number
numerical computations
IEEE754 standard
format binary floating-points
