TAILIEUCHUNG - Towards a mathematical library for UNUMs, an alternative to ieee 754 floating point numbers

This paper presented the ideas leading to UNUMs, gave some hints on floating point units for UNUMs and illustrated the difficulties in developing the said mathematical library by the example of approximating zeroes of analytic functions. | How to cite this paper: Risse, Th. (2018). Towards a mathematical library for UNUMs, an alternative to IEEE 754 floating point numbers. Journal of Information and Communication Technology (JICT), 17 (1), 1-16. TOWARDS A MATHEMATICAL LIBRARY FOR UNUMs, AN ALTERNATIVE TO IEEE 754 FLOATING POINT NUMBERS Thomas Risse Institute of Informatics and Automation, Faculty EEE & CS City University of Applied Sciences Bremen, Germany risse@ ABSTRACT The 1985 IEEE 754 standard for the representation of and the arithmetic with floating point numbers has been reconsidered. On the one hand today, its technological assumptions are by no means longer valid. On the other hand, the irritating numerical phenomena which have been collected cast a doubt as to whether this much uncertainty in numerical results is fate. Fortunately, around 2015, Gustafson proposed UNUMs, a modification of the IEEE 754 standard with the potential to heal the said shortcomings. Till now, there are some attempts to implement his ideas, both in software and in hardware. With these activities well under way, the other necessity is development of a mathematical library for UNUMs when one wants UNUMs to become the new floating point standard. This paper presented the ideas leading to UNUMs, gave some hints on floating point units for UNUMs and illustrated the difficulties in developing the said mathematical library by the example of approximating zeroes of analytic functions. Keywords: Floating point numbers, IEEE 754, UNUMs, arithmetic, mathematical library. Received: 25 June 2017 Accepted: 3 August 2017 Journal of ICT, 17, No. 1 (Jan) 2018, pp: 1–16 TOWARDS A MATHEMATICAL LIBRARY FOR UNUMs, AN ALTERNATIVE TO IEEE 754 FLOATING POINT NUMBERS Thomas Risse Institute of Informatics and Automation, Faculty EEE & CS City University of Applied Sciences Bremen, Germany risse@ ABSTRACT The 1985 IEEE 754 standard for the representation of and the arithmetic with floating point numbers has .

• Toán học
• Towards a mathematical library for UNUMs
• Mathematical library for UNUMs
• Alternative to ieee 754 floating point numbers
• Ieee 754 floating point numbers
• Approximating zeroes of analytic functions