Java Reference
In-Depth Information
IEEE 754 Standard
Originally, the computer implementation of floating-point mathematics
was based on proprietary data formats and processing routines developed
by each manufacturer. The results were unreliable and often erroneous,
includingaCraysupercomputerthatwasunabletoperformexactdivisions
by 2, a Honeywell computer in which the precision guard bits would disap-
pear unexpectedly, and several cases in which multiplication by 1.0 could
cause an overflow.
The first suggestion of a computer mathematics standard was an arti-
cle entitled A Proposed Standard for Binary Floating-Point Arithmetic ,
published in the SIGNUM Newsletter of the Association for Computing
Machinery (ACM) in October, 1979. This was followed by an article by
Jerome T. Coonen, titled An Implementation Guide to a Proposed Stan-
dard for Floating-Point Arithmetic. This one appeared in Computer
Magazine, January, 1980. Draft 8.0 of the proposed standard was pub-
lished in the March 1981 edition of Computer . The standard was approved
by the IEEE Standards Board on March 21, 1985, and by the American Na-
tional Standards Institute in July, 1985. The standard was reaffirmed on
December 6, 1990. A more general standard was published in 1987, under
the designation of IEEE Standard 854for Radix-Independent
Floating-Point Arithmetic .
The Foreword to IEEE 754 states that the intent of the standard is to
promote the portability of numeric software, to provide a uniform envi-
ronment for programs, and to encourage the development of better, safer,
and more sophisticated mathematical code. Among the specific refine-
ments of IEEE 754 are the diagnosis of anomalies at execution time, the
improved handling of exception conditions, and the implementation of in-
terval arithmetic. In addition, the standard provides for standard elemen-
tary functions, very high precision calculations, and the use of algebraic
symbolism in numerical operations.
A system in compliance with the IEEE 754 Standard can be imple-
mented in hardware, in software, or in both. However, conformance to
the standard is not determined by the internal properties of a system, but
by the user's perception. In other words, if a hardware product requires
additional software to comply with the provisions of the standard, it can-
not state that it conforms. The standard includes the following topics and
operations:
1. Floating-point numeric formats
Search WWH ::




Custom Search