Eyck Jentzsch
dcaf5467e8
(http://www.jhauser.us/arithmetic/SoftFloat.html) with RISCV specialization and cmake build
259 lines
6.2 KiB
HTML
259 lines
6.2 KiB
HTML
|
|
<HTML>
|
|
|
|
<HEAD>
|
|
<TITLE>Berkeley SoftFloat History</TITLE>
|
|
</HEAD>
|
|
|
|
<BODY>
|
|
|
|
<H1>History of Berkeley SoftFloat, to Release 3e</H1>
|
|
|
|
<P>
|
|
John R. Hauser<BR>
|
|
2018 January 20<BR>
|
|
</P>
|
|
|
|
|
|
<H3>Release 3e (2018 January)</H3>
|
|
|
|
<UL>
|
|
|
|
<LI>
|
|
Changed the default numeric code for optional rounding mode <CODE>odd</CODE>
|
|
(round to odd, also known as <EM>jamming</EM>) from 5 to 6.
|
|
|
|
<LI>
|
|
Modified the behavior of rounding mode <CODE>odd</CODE> when rounding to an
|
|
integer value (either conversion to an integer format or a
|
|
‘<CODE>roundToInt</CODE>’ function).
|
|
Previously, for those cases only, rounding mode <CODE>odd</CODE> acted the same
|
|
as rounding to minimum magnitude.
|
|
Now all operations are rounded consistently.
|
|
|
|
<LI>
|
|
Fixed some errors in the specialization code modeling Intel x86 floating-point,
|
|
specifically the integers returned on invalid operations and the propagation of
|
|
NaN payloads in a few rare cases.
|
|
|
|
<LI>
|
|
Added specialization code modeling ARM floating-point, conforming to VFPv2 or
|
|
later.
|
|
|
|
<LI>
|
|
Added an example target for ARM processors.
|
|
|
|
<LI>
|
|
Fixed a minor bug whereby function <CODE>f16_to_ui64</CODE> might return a
|
|
different integer than expected in the case that the floating-point operand is
|
|
negative.
|
|
|
|
<LI>
|
|
Added example target-specific optimization for GCC, employing GCC instrinsics
|
|
and support for <NOBR>128-bit</NOBR> integer arithmetic.
|
|
|
|
<LI>
|
|
Made other minor improvements.
|
|
|
|
</UL>
|
|
|
|
|
|
<H3>Release 3d (2017 August)</H3>
|
|
|
|
<UL>
|
|
|
|
<LI>
|
|
Fixed bugs in the square root functions for <NOBR>64-bit</NOBR>
|
|
double-precision, <NOBR>80-bit</NOBR> double-extended-precision, and
|
|
<NOBR>128-bit</NOBR> quadruple-precision.
|
|
For <NOBR>64-bit</NOBR> double-precision (<CODE>f64_sqrt</CODE>), the result
|
|
could sometimes be off by <NOBR>1 unit</NOBR> in the last place
|
|
(<NOBR>1 ulp</NOBR>) from what it should be.
|
|
For the larger formats, the square root could be wrong in a large portion of
|
|
the less-significant bits.
|
|
(A bug in <CODE>f128_sqrt</CODE> was first reported by Alexei Sibidanov.)
|
|
|
|
</UL>
|
|
|
|
|
|
<H3>Release 3c (2017 February)</H3>
|
|
|
|
<UL>
|
|
|
|
<LI>
|
|
Added optional rounding mode <CODE>odd</CODE> (round to odd, also known as
|
|
<EM>jamming</EM>).
|
|
|
|
<LI>
|
|
Corrected the documentation concerning non-canonical representations in
|
|
<NOBR>80-bit</NOBR> double-extended-precision.
|
|
|
|
</UL>
|
|
|
|
|
|
<H3>Release 3b (2016 July)</H3>
|
|
|
|
<UL>
|
|
|
|
<LI>
|
|
Implemented the common <NOBR>16-bit</NOBR> “half-precision”
|
|
floating-point format (<CODE>float16_t</CODE>).
|
|
|
|
<LI>
|
|
Made the integer values returned on invalid conversions to integer formats
|
|
be determined by the port-specific specialization instead of being the same for
|
|
all ports.
|
|
|
|
<LI>
|
|
Added preprocessor macro <CODE>THREAD_LOCAL</CODE> to allow the floating-point
|
|
state (modes and exception flags) to be made per-thread.
|
|
|
|
<LI>
|
|
Modified the provided Makefiles to allow some options to be overridden from the
|
|
<CODE>make</CODE> command.
|
|
|
|
<LI>
|
|
Made other minor improvements.
|
|
|
|
</UL>
|
|
|
|
|
|
<H3>Release 3a (2015 October)</H3>
|
|
|
|
<UL>
|
|
|
|
<LI>
|
|
Replaced the license text supplied by the University of California, Berkeley.
|
|
|
|
</UL>
|
|
|
|
|
|
<H3>Release 3 (2015 February)</H3>
|
|
|
|
<UL>
|
|
|
|
<LI>
|
|
Complete rewrite, funded by the University of California, Berkeley, and
|
|
consequently having a different use license than earlier releases.
|
|
Major changes included renaming most types and functions, upgrading some
|
|
algorithms, restructuring the source files, and making SoftFloat into a true
|
|
library.
|
|
|
|
<LI>
|
|
Added functions to convert between floating-point and unsigned integers, both
|
|
<NOBR>32-bit</NOBR> and <NOBR>64-bit</NOBR> (<CODE>uint32_t</CODE> and
|
|
<CODE>uint64_t</CODE>).
|
|
|
|
<LI>
|
|
Added functions for fused multiply-add, for all supported floating-point
|
|
formats except <NOBR>80-bit</NOBR> double-extended-precision.
|
|
|
|
<LI>
|
|
Added support for a fifth rounding mode, <CODE>near_maxMag</CODE> (round to
|
|
nearest, with ties to maximum magnitude, away from zero).
|
|
|
|
<LI>
|
|
Dropped the <CODE>timesoftfloat</CODE> program (now part of the Berkeley
|
|
TestFloat package).
|
|
|
|
</UL>
|
|
|
|
|
|
<H3>Release 2c (2015 January)</H3>
|
|
|
|
<UL>
|
|
|
|
<LI>
|
|
Fixed mistakes affecting some <NOBR>64-bit</NOBR> processors.
|
|
|
|
<LI>
|
|
Further improved the documentation and the wording for the legal restrictions
|
|
on using SoftFloat releases <NOBR>through 2c</NOBR> (not applicable to
|
|
<NOBR>Release 3</NOBR> or later).
|
|
|
|
</UL>
|
|
|
|
|
|
<H3>Release 2b (2002 May)</H3>
|
|
|
|
<UL>
|
|
|
|
<LI>
|
|
Made minor updates to the documentation, including improved wording for the
|
|
legal restrictions on using SoftFloat.
|
|
|
|
</UL>
|
|
|
|
|
|
<H3>Release 2a (1998 December)</H3>
|
|
|
|
<UL>
|
|
|
|
<LI>
|
|
Added functions to convert between <NOBR>64-bit</NOBR> integers
|
|
(<CODE>int64</CODE>) and all supported floating-point formats.
|
|
|
|
<LI>
|
|
Fixed a bug in all <NOBR>64-bit</NOBR>-version square root functions except
|
|
<CODE>float32_sqrt</CODE> that caused the result sometimes to be off by
|
|
<NOBR>1 unit</NOBR> in the last place (<NOBR>1 ulp</NOBR>) from what it should
|
|
be.
|
|
(Bug discovered by Paul Donahue.)
|
|
|
|
<LI>
|
|
Improved the Makefiles.
|
|
</UL>
|
|
|
|
|
|
<H3>Release 2 (1997 June)</H3>
|
|
|
|
<UL>
|
|
|
|
<LI>
|
|
Created the <NOBR>64-bit</NOBR> (<CODE>bits64</CODE>) version, adding the
|
|
<CODE>floatx80</CODE> and <CODE>float128</CODE> formats.
|
|
|
|
<LI>
|
|
Changed the source directory structure, splitting the sources into a
|
|
<CODE>bits32</CODE> and a <CODE>bits64</CODE> version.
|
|
Renamed <CODE>environment.h</CODE> to <CODE>milieu.h</CODE> to avoid confusion
|
|
with environment variables.
|
|
|
|
<LI>
|
|
Fixed a small error that caused <CODE>float64_round_to_int</CODE> often to
|
|
round the wrong way in nearest/even mode when the operand was between
|
|
2<SUP>20</SUP> and 2<SUP>21</SUP> and halfway between two integers.
|
|
|
|
</UL>
|
|
|
|
|
|
<H3>Release 1a (1996 July)</H3>
|
|
|
|
<UL>
|
|
|
|
<LI>
|
|
Corrected a mistake that caused borderline underflow cases not to raise the
|
|
underflow flag when they should have.
|
|
(Problem reported by Doug Priest.)
|
|
|
|
<LI>
|
|
Added the <CODE>float_detect_tininess</CODE> variable to control whether
|
|
tininess is detected before or after rounding.
|
|
|
|
</UL>
|
|
|
|
|
|
<H3>Release 1 (1996 July)</H3>
|
|
|
|
<UL>
|
|
|
|
<LI>
|
|
Original release, based on work done for the International Computer Science
|
|
Institute (ICSI) in Berkeley, California.
|
|
|
|
</UL>
|
|
|
|
|
|
</BODY>
|
|
|