| 1 |
|
|---|
| 2 | <HTML>
|
|---|
| 3 |
|
|---|
| 4 | <HEAD>
|
|---|
| 5 | <TITLE>Berkeley TestFloat History</TITLE>
|
|---|
| 6 | </HEAD>
|
|---|
| 7 |
|
|---|
| 8 | <BODY>
|
|---|
| 9 |
|
|---|
| 10 | <H1>History of Berkeley TestFloat, to Release 3e</H1>
|
|---|
| 11 |
|
|---|
| 12 | <P>
|
|---|
| 13 | John R. Hauser<BR>
|
|---|
| 14 | 2018 January 20<BR>
|
|---|
| 15 | </P>
|
|---|
| 16 |
|
|---|
| 17 |
|
|---|
| 18 | <P>
|
|---|
| 19 | Releases of Berkeley TestFloat normally parallel those of Berkeley SoftFloat,
|
|---|
| 20 | on which TestFloat is based.
|
|---|
| 21 | Each TestFloat release necessarily incorporates all bug fixes from the
|
|---|
| 22 | corresponding release of SoftFloat.
|
|---|
| 23 | </P>
|
|---|
| 24 |
|
|---|
| 25 |
|
|---|
| 26 | <H3>Release 3e (2018 January)</H3>
|
|---|
| 27 |
|
|---|
| 28 | <UL>
|
|---|
| 29 |
|
|---|
| 30 | <LI>
|
|---|
| 31 | Fixed a problem with the all-in-one <CODE>testfloat</CODE> program whereby
|
|---|
| 32 | function set <CODE>-all1</CODE> incorrectly also tested the three-operand fused
|
|---|
| 33 | multiply-add operations.
|
|---|
| 34 |
|
|---|
| 35 | <LI>
|
|---|
| 36 | Modified the expected behavior of rounding mode <CODE>odd</CODE> (jamming) when
|
|---|
| 37 | rounding to an integer value (either conversion to an integer format or a
|
|---|
| 38 | ‘<CODE>roundToInt</CODE>’ function).
|
|---|
| 39 | Previously, for those cases only, rounding mode <CODE>odd</CODE> was expected
|
|---|
| 40 | to act the same as rounding to minimum magnitude.
|
|---|
| 41 | Now, when rounding to an integer value, the nearest odd integer is expected,
|
|---|
| 42 | consistent with the round-to-odd result of other operations.
|
|---|
| 43 |
|
|---|
| 44 | <LI>
|
|---|
| 45 | Added options <CODE>-checkInvInts</CODE> and <CODE>-checkAll</CODE> to programs
|
|---|
| 46 | <CODE>testfloat_ver</CODE> and <CODE>testfloat</CODE>.
|
|---|
| 47 |
|
|---|
| 48 | <LI>
|
|---|
| 49 | Improved the checking of integer results of invalid operations.
|
|---|
| 50 |
|
|---|
| 51 | </UL>
|
|---|
| 52 |
|
|---|
| 53 |
|
|---|
| 54 | <H3>Release 3d (2017 August)</H3>
|
|---|
| 55 |
|
|---|
| 56 | <UL>
|
|---|
| 57 |
|
|---|
| 58 | <LI>
|
|---|
| 59 | When the all-in-one <CODE>testfloat</CODE> program is compiled to test the C
|
|---|
| 60 | language’s arithmetic, added the ability to test library functions
|
|---|
| 61 | <CODE>sqrtf</CODE>, <CODE>sqrtl</CODE>, <CODE>fmaf</CODE>, <CODE>fma</CODE>,
|
|---|
| 62 | and <CODE>fmal</CODE>, which were added to the <NOBR>C Standard</NOBR> in 1999.
|
|---|
| 63 |
|
|---|
| 64 | </UL>
|
|---|
| 65 |
|
|---|
| 66 |
|
|---|
| 67 | <H3>Release 3c (2017 February)</H3>
|
|---|
| 68 |
|
|---|
| 69 | <UL>
|
|---|
| 70 |
|
|---|
| 71 | <LI>
|
|---|
| 72 | Added support for testing rounding mode <CODE>odd</CODE> (jamming).
|
|---|
| 73 |
|
|---|
| 74 | <LI>
|
|---|
| 75 | Made support for testing <NOBR>64-bit</NOBR> double-precistion floating-point
|
|---|
| 76 | be subject to macro <CODE>FLOAT64</CODE> (akin to macros <CODE>FLOAT16</CODE>,
|
|---|
| 77 | <CODE>EXTFLOAT80</CODE>, and <CODE>FLOAT128</CODE> from before).
|
|---|
| 78 |
|
|---|
| 79 | <LI>
|
|---|
| 80 | Fixed some bugs that caused compilation to fail with certain combinations of
|
|---|
| 81 | option macro settings.
|
|---|
| 82 |
|
|---|
| 83 | <LI>
|
|---|
| 84 | Corrected the types of two internal variables to be <CODE>sig_atomic_t</CODE>
|
|---|
| 85 | instead of <CODE>bool</CODE>.
|
|---|
| 86 |
|
|---|
| 87 | <LI>
|
|---|
| 88 | Improved the formatting of some error reports (concerning where lines are
|
|---|
| 89 | broken when they exceed 79 characters in length).
|
|---|
| 90 |
|
|---|
| 91 | </UL>
|
|---|
| 92 |
|
|---|
| 93 |
|
|---|
| 94 | <H3>Release 3b (2016 July)</H3>
|
|---|
| 95 |
|
|---|
| 96 | <UL>
|
|---|
| 97 |
|
|---|
| 98 | <LI>
|
|---|
| 99 | Added the ability to test the common <NOBR>16-bit</NOBR>
|
|---|
| 100 | “half-precision” floating-point format.
|
|---|
| 101 |
|
|---|
| 102 | <LI>
|
|---|
| 103 | Added a <CODE>-seed</CODE> option to programs <CODE>testfloat_gen</CODE>,
|
|---|
| 104 | <CODE>testfloat</CODE>, and <CODE>testsoftfloat</CODE> for setting the seed for
|
|---|
| 105 | the pseudo-random number generator used to generate test cases.
|
|---|
| 106 |
|
|---|
| 107 | <LI>
|
|---|
| 108 | Where a specific choice is needed for how tininess is detected on underflow,
|
|---|
| 109 | changed the default to be the detection of tininess after rounding
|
|---|
| 110 | (<CODE>-tininessafter</CODE>) instead of before rounding
|
|---|
| 111 | (<CODE>-tininessbefore</CODE>).
|
|---|
| 112 |
|
|---|
| 113 | <LI>
|
|---|
| 114 | Modified the format of reported discrepancies to show the signs of
|
|---|
| 115 | floating-point values using <CODE>+</CODE> and <CODE>-</CODE> characters.
|
|---|
| 116 |
|
|---|
| 117 | <LI>
|
|---|
| 118 | Documented the use of the <CODE>INLINE</CODE> macro, and fixed the sources for
|
|---|
| 119 | the case that function inlining is not supported by the C compiler.
|
|---|
| 120 |
|
|---|
| 121 | <LI>
|
|---|
| 122 | Documented the possible need to define macro <CODE>THREAD_LOCAL</CODE> to match
|
|---|
| 123 | how the SoftFloat library was built.
|
|---|
| 124 |
|
|---|
| 125 | <LI>
|
|---|
| 126 | Modified the provided Makefiles to allow some options to be overridden from the
|
|---|
| 127 | <CODE>make</CODE> command.
|
|---|
| 128 |
|
|---|
| 129 | </UL>
|
|---|
| 130 |
|
|---|
| 131 |
|
|---|
| 132 | <H3>Release 3a (2015 October)</H3>
|
|---|
| 133 |
|
|---|
| 134 | <UL>
|
|---|
| 135 |
|
|---|
| 136 | <LI>
|
|---|
| 137 | Replaced the license text supplied by the University of California, Berkeley,
|
|---|
| 138 | and fixed some minor build problems.
|
|---|
| 139 |
|
|---|
| 140 | </UL>
|
|---|
| 141 |
|
|---|
| 142 |
|
|---|
| 143 | <H3>Release 3 (2015 February)</H3>
|
|---|
| 144 |
|
|---|
| 145 | <UL>
|
|---|
| 146 |
|
|---|
| 147 | <LI>
|
|---|
| 148 | Complete rewrite, funded by the University of California, Berkeley, and
|
|---|
| 149 | consequently having a different use license than earlier releases.
|
|---|
| 150 | Visible changes included different names for testable functions and command
|
|---|
| 151 | options.
|
|---|
| 152 |
|
|---|
| 153 | <LI>
|
|---|
| 154 | Reinstated separate programs for generating test cases
|
|---|
| 155 | (<CODE>testfloat_ver</CODE>) and verifying test results
|
|---|
| 156 | (<CODE>testfloat_gen</CODE>), as alternatives to the all-in-one
|
|---|
| 157 | <CODE>testfloat</CODE> program (which remained supported).
|
|---|
| 158 |
|
|---|
| 159 | <LI>
|
|---|
| 160 | Added support for testing conversions between floating-point and unsigned
|
|---|
| 161 | integers, both <NOBR>32-bit</NOBR> and <NOBR>64-bit</NOBR>.
|
|---|
| 162 |
|
|---|
| 163 | <LI>
|
|---|
| 164 | Added support for testing a fused multiply-add operation, for all testable
|
|---|
| 165 | floating-point formats except <NOBR>80-bit</NOBR> double-extended-precision.
|
|---|
| 166 |
|
|---|
| 167 | <LI>
|
|---|
| 168 | Added support for testing a fifth rounding mode, <CODE>near_maxMag</CODE>
|
|---|
| 169 | (round to nearest, with ties to maximum magnitude, away from zero).
|
|---|
| 170 |
|
|---|
| 171 | <LI>
|
|---|
| 172 | Added <CODE>timesoftfloat</CODE> (previously found in the Berkeley SoftFloat
|
|---|
| 173 | package).
|
|---|
| 174 |
|
|---|
| 175 | </UL>
|
|---|
| 176 |
|
|---|
| 177 |
|
|---|
| 178 | <H3>Release 2c (2015 January)</H3>
|
|---|
| 179 |
|
|---|
| 180 | <UL>
|
|---|
| 181 |
|
|---|
| 182 | <LI>
|
|---|
| 183 | Fixed mistakes affecting some <NOBR>64-bit</NOBR> processors.
|
|---|
| 184 |
|
|---|
| 185 | <LI>
|
|---|
| 186 | Made minor updates to the documentation, including improved wording for the
|
|---|
| 187 | legal restrictions on using TestFloat releases <NOBR>through 2c</NOBR> (not
|
|---|
| 188 | applicable to <NOBR>Release 3</NOBR> or later).
|
|---|
| 189 |
|
|---|
| 190 | </UL>
|
|---|
| 191 |
|
|---|
| 192 |
|
|---|
| 193 | <P>
|
|---|
| 194 | There was never a <NOBR>Release 2b</NOBR>.
|
|---|
| 195 | </P>
|
|---|
| 196 |
|
|---|
| 197 |
|
|---|
| 198 | <H3>Release 2a (1998 December)</H3>
|
|---|
| 199 |
|
|---|
| 200 | <UL>
|
|---|
| 201 |
|
|---|
| 202 | <LI>
|
|---|
| 203 | Added support for testing conversions between floating-point and
|
|---|
| 204 | <NOBR>64-bit</NOBR> signed integers.
|
|---|
| 205 |
|
|---|
| 206 | <LI>
|
|---|
| 207 | Improved the Makefiles.
|
|---|
| 208 |
|
|---|
| 209 | </UL>
|
|---|
| 210 |
|
|---|
| 211 |
|
|---|
| 212 | <H3>Release 2 (1997 June)</H3>
|
|---|
| 213 |
|
|---|
| 214 | <UL>
|
|---|
| 215 |
|
|---|
| 216 | <LI>
|
|---|
| 217 | Integrated the generation of test cases and the checking of system results into
|
|---|
| 218 | a single program.
|
|---|
| 219 | (Before they were separate programs, normally joined by explicit command-line
|
|---|
| 220 | pipes.)
|
|---|
| 221 |
|
|---|
| 222 | <LI>
|
|---|
| 223 | Improved the sequence of test cases.
|
|---|
| 224 |
|
|---|
| 225 | <LI>
|
|---|
| 226 | Added support for testing <NOBR>80-bit</NOBR> double-extended-precision and
|
|---|
| 227 | <NOBR>128-bit</NOBR> quadruple precision.
|
|---|
| 228 |
|
|---|
| 229 | <LI>
|
|---|
| 230 | Made program output more readable, and added new command arguments.
|
|---|
| 231 |
|
|---|
| 232 | <LI>
|
|---|
| 233 | Reduced dependence on the quality of the standard <CODE>rand</CODE> function
|
|---|
| 234 | for generating test cases.
|
|---|
| 235 | (Previously naively expected <CODE>rand</CODE> to be able to generate good
|
|---|
| 236 | random bits for the entire machine word width.)
|
|---|
| 237 |
|
|---|
| 238 | <LI>
|
|---|
| 239 | Created <CODE>testsoftfloat</CODE>, with its own simpler complete software
|
|---|
| 240 | floating-point (“slowfloat”) for comparison purposes.
|
|---|
| 241 |
|
|---|
| 242 | <LI>
|
|---|
| 243 | Made some changes to the source file structure, including renaming
|
|---|
| 244 | <CODE>environment.h</CODE> to <CODE>milieu.h</CODE> (to avoid confusion with
|
|---|
| 245 | environment variables).
|
|---|
| 246 |
|
|---|
| 247 | </UL>
|
|---|
| 248 |
|
|---|
| 249 |
|
|---|
| 250 | <H3>Release 1a (1996 July)</H3>
|
|---|
| 251 |
|
|---|
| 252 | <UL>
|
|---|
| 253 |
|
|---|
| 254 | <LI>
|
|---|
| 255 | Added the <CODE>-tininessbefore</CODE> and <CODE>-tininessafter</CODE> options
|
|---|
| 256 | to control whether tininess should be detected before or after rounding.
|
|---|
| 257 |
|
|---|
| 258 | </UL>
|
|---|
| 259 |
|
|---|
| 260 |
|
|---|
| 261 | <H3>Release 1 (1996 July)</H3>
|
|---|
| 262 |
|
|---|
| 263 | <UL>
|
|---|
| 264 |
|
|---|
| 265 | <LI>
|
|---|
| 266 | Original release.
|
|---|
| 267 |
|
|---|
| 268 | </UL>
|
|---|
| 269 |
|
|---|
| 270 |
|
|---|
| 271 | </BODY>
|
|---|
| 272 |
|
|---|
