© 1975 by British Computer Society
| ||||||||||||||||||||||||||||||||||||||||||||||||||
A quantitative measure of precision
York University, Department of Chemistry, 4700 Keele Street, Downsview M3J 1P3, Ontario, Canada
The precision
b of a real number is defined quantitatively in terms of the fractional error in the number, and the base of the arithmetic in which it is represented. The definition is an extension of the traditional rough measure of precision as the number of significant digits in the number. In binary arithmetic the integral part of
b is the number of binary digits required to store the number. Conversion of the precision from one base to another (such as binary/decimal) is discussed, and applied to consideration of the intrinsic precision of input/output routines and floating point arithmetic.
Received April 1974.
* York University, Department of Chemistry, 4700 Keele Street, Downsview M3J 1P3, Ontario, Canada