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
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
|
/*============================================================================
This source file is an extension to the SoftFloat IEC/IEEE Floating-point
Arithmetic Package, Release 2b, written for Bochs (x86 achitecture simulator)
floating point emulation.
THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort has
been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT TIMES
RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO PERSONS
AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ALL LOSSES,
COSTS, OR OTHER PROBLEMS THEY INCUR DUE TO THE SOFTWARE, AND WHO FURTHERMORE
EFFECTIVELY INDEMNIFY JOHN HAUSER AND THE INTERNATIONAL COMPUTER SCIENCE
INSTITUTE (possibly via similar legal warning) AGAINST ALL LOSSES, COSTS, OR
OTHER PROBLEMS INCURRED BY THEIR CUSTOMERS AND CLIENTS DUE TO THE SOFTWARE.
Derivative works are acceptable, even for commercial purposes, so long as
(1) the source code for the derivative work includes prominent notice that
the work is derivative, and (2) the source code includes prominent notice with
these four paragraphs for those parts of this code that are retained.
=============================================================================*/
/*============================================================================
* Written for Bochs (x86 achitecture simulator) by
* Stanislav Shwartsman [sshwarts at sourceforge net]
* ==========================================================================*/
#include "softfloatx80.h"
#include "softfloat-round-pack.h"
#include "softfloat-macros.h"
/*----------------------------------------------------------------------------
| Returns the result of converting the extended double-precision floating-
| point value `a' to the 16-bit two's complement integer format. The
| conversion is performed according to the IEC/IEEE Standard for Binary
| Floating-Point Arithmetic - which means in particular that the conversion
| is rounded according to the current rounding mode. If `a' is a NaN or the
| conversion overflows, the integer indefinite value is returned.
*----------------------------------------------------------------------------*/
Bit16s floatx80_to_int16(floatx80 a, float_status_t &status)
{
if (floatx80_is_unsupported(a)) {
float_raise(status, float_flag_invalid);
return int16_indefinite;
}
Bit32s v32 = floatx80_to_int32(a, status);
if ((v32 > 32767) || (v32 < -32768)) {
status.float_exception_flags = float_flag_invalid; // throw away other flags
return int16_indefinite;
}
return (Bit16s) v32;
}
/*----------------------------------------------------------------------------
| Returns the result of converting the extended double-precision floating-
| point value `a' to the 16-bit two's complement integer format. The
| conversion is performed according to the IEC/IEEE Standard for Binary
| Floating-Point Arithmetic, except that the conversion is always rounded
| toward zero. If `a' is a NaN or the conversion overflows, the integer
| indefinite value is returned.
*----------------------------------------------------------------------------*/
Bit16s floatx80_to_int16_round_to_zero(floatx80 a, float_status_t &status)
{
if (floatx80_is_unsupported(a)) {
float_raise(status, float_flag_invalid);
return int16_indefinite;
}
Bit32s v32 = floatx80_to_int32_round_to_zero(a, status);
if ((v32 > 32767) || (v32 < -32768)) {
status.float_exception_flags = float_flag_invalid; // throw away other flags
return int16_indefinite;
}
return (Bit16s) v32;
}
/*----------------------------------------------------------------------------
| Separate the source extended double-precision floating point value `a'
| into its exponent and significand, store the significant back to the
| 'a' and return the exponent. The operation performed is a superset of
| the IEC/IEEE recommended logb(x) function.
*----------------------------------------------------------------------------*/
floatx80 floatx80_extract(floatx80 &a, float_status_t &status)
{
Bit64u aSig = extractFloatx80Frac(a);
Bit32s aExp = extractFloatx80Exp(a);
int aSign = extractFloatx80Sign(a);
if (floatx80_is_unsupported(a))
{
float_raise(status, float_flag_invalid);
a = floatx80_default_nan;
return a;
}
if (aExp == 0x7FFF) {
if ((Bit64u) (aSig<<1))
{
a = propagateFloatx80NaN(a, status);
return a;
}
return packFloatx80(0, 0x7FFF, BX_CONST64(0x8000000000000000));
}
if (aExp == 0)
{
if (aSig == 0) {
float_raise(status, float_flag_divbyzero);
a = packFloatx80(aSign, 0, 0);
return packFloatx80(1, 0x7FFF, BX_CONST64(0x8000000000000000));
}
float_raise(status, float_flag_denormal);
normalizeFloatx80Subnormal(aSig, &aExp, &aSig);
}
a.exp = (aSign << 15) + 0x3FFF;
a.fraction = aSig;
return int32_to_floatx80(aExp - 0x3FFF);
}
/*----------------------------------------------------------------------------
| Scales extended double-precision floating-point value in operand `a' by
| value `b'. The function truncates the value in the second operand 'b' to
| an integral value and adds that value to the exponent of the operand 'a'.
| The operation performed according to the IEC/IEEE Standard for Binary
| Floating-Point Arithmetic.
*----------------------------------------------------------------------------*/
floatx80 floatx80_scale(floatx80 a, floatx80 b, float_status_t &status)
{
Bit32s aExp, bExp;
Bit64u aSig, bSig;
// handle unsupported extended double-precision floating encodings
if (floatx80_is_unsupported(a) || floatx80_is_unsupported(b))
{
float_raise(status, float_flag_invalid);
return floatx80_default_nan;
}
aSig = extractFloatx80Frac(a);
aExp = extractFloatx80Exp(a);
int aSign = extractFloatx80Sign(a);
bSig = extractFloatx80Frac(b);
bExp = extractFloatx80Exp(b);
int bSign = extractFloatx80Sign(b);
if (aExp == 0x7FFF) {
if ((Bit64u) (aSig<<1) || ((bExp == 0x7FFF) && (Bit64u) (bSig<<1)))
{
return propagateFloatx80NaN(a, b, status);
}
if ((bExp == 0x7FFF) && bSign) {
float_raise(status, float_flag_invalid);
return floatx80_default_nan;
}
if (bSig && (bExp == 0)) float_raise(status, float_flag_denormal);
return a;
}
if (bExp == 0x7FFF) {
if ((Bit64u) (bSig<<1)) return propagateFloatx80NaN(a, b, status);
if ((aExp | aSig) == 0) {
if (! bSign) {
float_raise(status, float_flag_invalid);
return floatx80_default_nan;
}
return a;
}
if (aSig && (aExp == 0)) float_raise(status, float_flag_denormal);
if (bSign) return packFloatx80(aSign, 0, 0);
return packFloatx80(aSign, 0x7FFF, BX_CONST64(0x8000000000000000));
}
if (aExp == 0) {
if (bSig && (bExp == 0)) float_raise(status, float_flag_denormal);
if (aSig == 0) return a;
float_raise(status, float_flag_denormal);
normalizeFloatx80Subnormal(aSig, &aExp, &aSig);
if (bExp < 0x3FFF)
return normalizeRoundAndPackFloatx80(80, aSign, aExp, aSig, 0, status);
}
if (bExp == 0) {
if (bSig == 0) return a;
float_raise(status, float_flag_denormal);
normalizeFloatx80Subnormal(bSig, &bExp, &bSig);
}
if (bExp > 0x400E) {
/* generate appropriate overflow/underflow */
return roundAndPackFloatx80(80, aSign,
bSign ? -0x3FFF : 0x7FFF, aSig, 0, status);
}
if (bExp < 0x3FFF) return a;
int shiftCount = 0x403E - bExp;
bSig >>= shiftCount;
Bit32s scale = (Bit32s) bSig;
if (bSign) scale = -scale; /* -32768..32767 */
return
roundAndPackFloatx80(80, aSign, aExp+scale, aSig, 0, status);
}
/*----------------------------------------------------------------------------
| Determine extended-precision floating-point number class.
*----------------------------------------------------------------------------*/
float_class_t floatx80_class(floatx80 a)
{
Bit32s aExp = extractFloatx80Exp(a);
Bit64u aSig = extractFloatx80Frac(a);
if(aExp == 0) {
if (aSig == 0)
return float_zero;
/* denormal or pseudo-denormal */
return float_denormal;
}
/* valid numbers have the MS bit set */
if (!(aSig & BX_CONST64(0x8000000000000000)))
return float_SNaN; /* report unsupported as SNaNs */
if(aExp == 0x7fff) {
int aSign = extractFloatx80Sign(a);
if (((Bit64u) (aSig<< 1)) == 0)
return (aSign) ? float_negative_inf : float_positive_inf;
return (aSig & BX_CONST64(0x4000000000000000)) ? float_QNaN : float_SNaN;
}
return float_normalized;
}
/*----------------------------------------------------------------------------
| Compare between two extended precision floating point numbers. Returns
| 'float_relation_equal' if the operands are equal, 'float_relation_less' if
| the value 'a' is less than the corresponding value `b',
| 'float_relation_greater' if the value 'a' is greater than the corresponding
| value `b', or 'float_relation_unordered' otherwise.
*----------------------------------------------------------------------------*/
int floatx80_compare(floatx80 a, floatx80 b, float_status_t &status)
{
float_class_t aClass = floatx80_class(a);
float_class_t bClass = floatx80_class(b);
if (aClass == float_SNaN || aClass == float_QNaN || bClass == float_SNaN || bClass == float_QNaN)
{
float_raise(status, float_flag_invalid);
return float_relation_unordered;
}
if (aClass == float_denormal || bClass == float_denormal) {
float_raise(status, float_flag_denormal);
}
int aSign = extractFloatx80Sign(a);
int bSign = extractFloatx80Sign(b);
if (aClass == float_zero) {
if (bClass == float_zero) return float_relation_equal;
return bSign ? float_relation_greater : float_relation_less;
}
if (bClass == float_zero || aSign != bSign) {
return aSign ? float_relation_less : float_relation_greater;
}
Bit64u aSig = extractFloatx80Frac(a);
Bit32s aExp = extractFloatx80Exp(a);
Bit64u bSig = extractFloatx80Frac(b);
Bit32s bExp = extractFloatx80Exp(b);
if (aClass == float_denormal)
normalizeFloatx80Subnormal(aSig, &aExp, &aSig);
if (bClass == float_denormal)
normalizeFloatx80Subnormal(bSig, &bExp, &bSig);
if (aExp == bExp && aSig == bSig)
return float_relation_equal;
int less_than =
aSign ? ((bExp < aExp) || ((bExp == aExp) && (bSig < aSig)))
: ((aExp < bExp) || ((aExp == bExp) && (aSig < bSig)));
if (less_than) return float_relation_less;
return float_relation_greater;
}
/*----------------------------------------------------------------------------
| Compare between two extended precision floating point numbers. Returns
| 'float_relation_equal' if the operands are equal, 'float_relation_less' if
| the value 'a' is less than the corresponding value `b',
| 'float_relation_greater' if the value 'a' is greater than the corresponding
| value `b', or 'float_relation_unordered' otherwise. Quiet NaNs do not cause
| an exception.
*----------------------------------------------------------------------------*/
int floatx80_compare_quiet(floatx80 a, floatx80 b, float_status_t &status)
{
float_class_t aClass = floatx80_class(a);
float_class_t bClass = floatx80_class(b);
if (aClass == float_SNaN || bClass == float_SNaN)
{
/* unsupported reported as SNaN */
float_raise(status, float_flag_invalid);
return float_relation_unordered;
}
if (aClass == float_QNaN || bClass == float_QNaN) {
return float_relation_unordered;
}
if (aClass == float_denormal || bClass == float_denormal) {
float_raise(status, float_flag_denormal);
}
int aSign = extractFloatx80Sign(a);
int bSign = extractFloatx80Sign(b);
if (aClass == float_zero) {
if (bClass == float_zero) return float_relation_equal;
return bSign ? float_relation_greater : float_relation_less;
}
if (bClass == float_zero || aSign != bSign) {
return aSign ? float_relation_less : float_relation_greater;
}
Bit64u aSig = extractFloatx80Frac(a);
Bit32s aExp = extractFloatx80Exp(a);
Bit64u bSig = extractFloatx80Frac(b);
Bit32s bExp = extractFloatx80Exp(b);
if (aClass == float_denormal)
normalizeFloatx80Subnormal(aSig, &aExp, &aSig);
if (bClass == float_denormal)
normalizeFloatx80Subnormal(bSig, &bExp, &bSig);
if (aExp == bExp && aSig == bSig)
return float_relation_equal;
int less_than =
aSign ? ((bExp < aExp) || ((bExp == aExp) && (bSig < aSig)))
: ((aExp < bExp) || ((aExp == bExp) && (aSig < bSig)));
if (less_than) return float_relation_less;
return float_relation_greater;
}
|
