BEST POSSIBLE FLOATING POINT ARITHMETIC.
Abstract
The report presents an algorithm for floating point arithmetic, using single-length arithmetic registers, which yields the most accurate approximation which can be expressed in the given floating point format, the greatest lower bound, or the least upper bound for the result of the operation, whichever is specified by the programmer. This makes the algorithm particularly useful for implementation of interval arithmetic. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1970
- Accession Number
- AD0713699
Entities
People
- J. Michael Yohe
Organizations
- University of Wisconsin–Madison