Fractional Number Representation for the Digital Sintesi Simplifying Denis Ferreira Figueiredo UNIVERSIDADE VEIGA DE ALMEIDA Instituto de Ciência e Tecnologia – ICT – CINFO Rua Ibituruna No. 108 - Maracanã Rio de Janeiro - RJ - Cep: 20271-020 [email protected] Prof. Luiz Biondi Neto UNIVERSIDADE VEIGA DE ALMEIDA Instituto de Ciência e Tecnologia – ICT – CINFO Rua Ibituruna No. 108 - Maracanã Rio de Janeiro - RJ - Cep: 20271-020 [email protected] MSc. Eng. Fernando Hideo Fukuda UNIVERSIDADE FEDERAL DO RIO DE JANEIRO (UFRJ-COPPE) Núcleo de Transferência de Tecnologia (NTT) [email protected] ABSTRACT The complex task of a Floating-Point Math Unit implementation, well as their functional performance is evaluated under a proposal of a different point of view in computational methodology, representation formats and their primary functions. The use of the actual Floating-Point Standards (IEEE 754) is put under comparative view with more simple math methods, with increased performance for computational environments, as illustrated by the numeric analysis and results, by the adoption of fractional number representation of REAL numbers, against the exponential representation. That simplification in methods being applied in the implementation, in a well stable technological base, is able to represent improvements of performance, and in numerical math precision too. Under a analysis of standard math benchmark models, commonly used for math precision and speed evaluation, can qualify the precision scope of the two forms of numerical representation, well as their calculation performances and pondering the error propagation issues. References [1] IEEE 1987, IEEE STANDARD 754-1985 for Binary Floating–point Arithmetic, IEEE, Reprinted in SIGPLAN 22(2) pp925. [2] Brow, W. S. 1981. A simple but realistic model of floating-point computation. ACM trans. Math. Softw. 7,4, 445-480 [3] Courant, R. 2000, O que é matemática, Robbins Herbert. Ed Ciência Moderna, Rio de Janeiro, RJ, Brasil [4] Cody, W. J. 1988. Floating-point standards – theory and practice. In reliability in Computing: The Role of Interval Methods on Scientific Computing, Ramon E. Moore. Ed. Academic Press, Boston, Mass. Numerical Microsystems, [5] Sun Computation Guide Part no. 800-5277-10 (Rev. A, 22 Feb. 1991) [6] INTEL, Intel Pentium Family User's Manual, Volume 3: Architecture and Programming Manual (1994) Order no. 241430 [7] ANSI 1978, American National Standard Programming Language FORTRAN, ANSI Standard X3.9-1978 American National Standard, New York [8] Goldberg, D., 1995, “What Every Computer Scientist Should Know About Floating-Point Arithmetic”, ACM Computing Surveys 23, 1 (March 1991), 5-48. A version of it is reprinted in SunPro’s Numerical Computation Guide. 151