DocumentCode :
2312473
Title :
On a tapered floating point system
Author :
Azmi, Aqil M. ; Lombardi, Fabrizio
Author_Institution :
Dept. of Electr. & Comput. Eng., Colorado Univ., Boulder, CO, USA
fYear :
1989
fDate :
6-8 Sep 1989
Firstpage :
2
Lastpage :
9
Abstract :
R. Morris (see IEEE Trans. Comput., vol.TC-20, p.1578-9, 1971), suggested adding an extra field to the fixed floating point system, so that exponents can be stored more efficiently. The exponents are stored in the smallest possible space, passing the extra bits to the mantissa. The extra field is used to monitor the current length of the exponent. The gain in precision and/or exponent range outweighs the overhead of the extra field and the processing speed. The authors provide implementation details, error analysis, and some future research ideas. Simulation results are provided for comparison purposes
Keywords :
digital arithmetic; error analysis; exponents; tapered floating point system; Accuracy; Arithmetic; Error analysis; Gaussian distribution; Monitoring; Performance analysis;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Computer Arithmetic, 1989., Proceedings of 9th Symposium on
Conference_Location :
Santa Monica, CA
Print_ISBN :
0-8186-8963-3
Type :
conf
DOI :
10.1109/ARITH.1989.72803
Filename :
72803
Link To Document :
بازگشت