Decimal Encoding Specification, version 1.01
Copyright (c) IBM Corporation, 2009. All rights reserved. ©
| 7 Apr 2009 |
| [previous | contents]
|
This appendix documents changes since the first circulation of the
‘Strawman 3’ proposal (7 November 2002). It is not part of the
specification.
-
It is pointed out that, although infinities and NaNs should be
recognized as such based on only the first byte of the number,
arithmetic may well be required to fully define all bits of an infinite
or NaN result (for example, by clearing the remaining bytes to 0).
-
A note has been added to Appendix A, pointing out that the coefficient
can be considered to have an internal decimal point if an appropriate
change to the bias is made.
-
The term Umax has been renamed Elimit for
consistency with other documents.
-
The exponent range of all three formats has been expanded by making the
largest normal number be the largest representable number. This also
simplifies the calculations of Emax and bias. There are now
no ‘supernormal’ values, and one value of encoded exponent is
always unused (and is therefore available for future expansion of the
encoding scheme).
-
The names compact,
single precision, and
double precision have been replaced by the names
decimal4,
decimal8, and
decimal16 respectively, to avoid confusion with the binary
floating-point formats.
-
The distinction between signaling NaN and quiet NaN, using the top bit
of the exponent continuation field, has been made part of the
specification.
-
For an Infinity, the top bits of the exponent continuation field have
been defined to be 0 (instead of 1). This places Infinity logically
adjacent to the finite numbers, and allows Infinities to have all bits
zero after the combination field.
During the IEEE 754r committee meeting in January 2003, the
following changes were agreed.
-
Signaling NaN has been removed.
-
Redundant Densely Packed Decimal encodings are permitted in operands.
-
The value of Emax is raised by one for all encodings
(leaving Emin unchanged), hence using all exponent values
available.
These changes mean that all possible bit patterns in the formats are
valid.
During the IEEE 754r committee meeting in February 2003, the
following changes were agreed.
-
Signaling NaN has been restored.
-
The names
decimal4,
decimal8, and
decimal16
have been replaced by the names
decimal32,
decimal64, and
decimal128
respectively.
The decimal encodings are now included in the IEEE 754 standard
approved in June 2008, so this document is no longer a proposal
and has been updated to reflect that. There are no technical
changes from the previous version (Strawman 4d).
Also, all references to the General Decimal Arithmetic website have
been updated to
http://speleotrove.com/decimal
(its new location).
The document is now formatted using OpenOffice (generated from GML),
for improved PDF files with bookmarks, hot links, etc. There are no
technical changes.
[previous | contents]