1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
|
<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>
|
