-
Notifications
You must be signed in to change notification settings - Fork 13
/
Copy pathsoftfloat.c
7450 lines (6724 loc) · 263 KB
/
softfloat.c
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
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*
* QEMU float support
*
* Derived from SoftFloat.
*/
/*============================================================================
This C source file is part of the SoftFloat IEC/IEEE Floating-point Arithmetic
Package, Release 2b.
Written by John R. Hauser. This work was made possible in part by the
International Computer Science Institute, located at Suite 600, 1947 Center
Street, Berkeley, California 94704. Funding was partially provided by the
National Science Foundation under grant MIP-9311980. The original version
of this code was written as part of a project to build a fixed-point vector
processor in collaboration with the University of California at Berkeley,
overseen by Profs. Nelson Morgan and John Wawrzynek. More information
is available through the Web page `http://www.cs.berkeley.edu/~jhauser/
arithmetic/SoftFloat.html'.
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.
=============================================================================*/
/* softfloat (and in particular the code in softfloat-specialize.h) is
* target-dependent and needs the TARGET_* macros.
*/
#include "config.h"
#include "fpu/softfloat.h"
/* We only need stdlib for abort() */
#include <stdlib.h>
/*----------------------------------------------------------------------------
| Primitive arithmetic functions, including multi-word arithmetic, and
| division and square root approximations. (Can be specialized to target if
| desired.)
*----------------------------------------------------------------------------*/
#include "softfloat-macros.h"
/*----------------------------------------------------------------------------
| Functions and definitions to determine: (1) whether tininess for underflow
| is detected before or after rounding by default, (2) what (if anything)
| happens when exceptions are raised, (3) how signaling NaNs are distinguished
| from quiet NaNs, (4) the default generated quiet NaNs, and (5) how NaNs
| are propagated from function inputs to output. These details are target-
| specific.
*----------------------------------------------------------------------------*/
#include "softfloat-specialize.h"
/*----------------------------------------------------------------------------
| Returns the fraction bits of the half-precision floating-point value `a'.
*----------------------------------------------------------------------------*/
INLINE uint32_t extractFloat16Frac(float16 a)
{
return float16_val(a) & 0x3ff;
}
/*----------------------------------------------------------------------------
| Returns the exponent bits of the half-precision floating-point value `a'.
*----------------------------------------------------------------------------*/
INLINE int_fast16_t extractFloat16Exp(float16 a)
{
return (float16_val(a) >> 10) & 0x1f;
}
/*----------------------------------------------------------------------------
| Returns the sign bit of the single-precision floating-point value `a'.
*----------------------------------------------------------------------------*/
INLINE flag extractFloat16Sign(float16 a)
{
return float16_val(a)>>15;
}
/*----------------------------------------------------------------------------
| Takes a 64-bit fixed-point value `absZ' with binary point between bits 6
| and 7, and returns the properly rounded 32-bit integer corresponding to the
| input. If `zSign' is 1, the input is negated before being converted to an
| integer. Bit 63 of `absZ' must be zero. Ordinarily, the fixed-point input
| is simply rounded to an integer, with the inexact exception raised if the
| input cannot be represented exactly as an integer. However, if the fixed-
| point input is too large, the invalid exception is raised and the largest
| positive or negative integer is returned.
*----------------------------------------------------------------------------*/
static int32 roundAndPackInt32( flag zSign, uint64_t absZ STATUS_PARAM)
{
int8 roundingMode;
flag roundNearestEven;
int8 roundIncrement, roundBits;
int32_t z;
roundingMode = STATUS(float_rounding_mode);
roundNearestEven = ( roundingMode == float_round_nearest_even );
switch (roundingMode) {
case float_round_nearest_even:
case float_round_ties_away:
roundIncrement = 0x40;
break;
case float_round_to_zero:
roundIncrement = 0;
break;
case float_round_up:
roundIncrement = zSign ? 0 : 0x7f;
break;
case float_round_down:
roundIncrement = zSign ? 0x7f : 0;
break;
default:
abort();
}
roundBits = absZ & 0x7F;
absZ = ( absZ + roundIncrement )>>7;
absZ &= ~ ( ( ( roundBits ^ 0x40 ) == 0 ) & roundNearestEven );
z = absZ;
if ( zSign ) z = - z;
if ( ( absZ>>32 ) || ( z && ( ( z < 0 ) ^ zSign ) ) ) {
float_raise( float_flag_invalid STATUS_VAR);
return zSign ? (int32_t) 0x80000000 : 0x7FFFFFFF;
}
if ( roundBits ) STATUS(float_exception_flags) |= float_flag_inexact;
return z;
}
/*----------------------------------------------------------------------------
| Takes the 128-bit fixed-point value formed by concatenating `absZ0' and
| `absZ1', with binary point between bits 63 and 64 (between the input words),
| and returns the properly rounded 64-bit integer corresponding to the input.
| If `zSign' is 1, the input is negated before being converted to an integer.
| Ordinarily, the fixed-point input is simply rounded to an integer, with
| the inexact exception raised if the input cannot be represented exactly as
| an integer. However, if the fixed-point input is too large, the invalid
| exception is raised and the largest positive or negative integer is
| returned.
*----------------------------------------------------------------------------*/
static int64 roundAndPackInt64( flag zSign, uint64_t absZ0, uint64_t absZ1 STATUS_PARAM)
{
int8 roundingMode;
flag roundNearestEven, increment;
int64_t z;
roundingMode = STATUS(float_rounding_mode);
roundNearestEven = ( roundingMode == float_round_nearest_even );
switch (roundingMode) {
case float_round_nearest_even:
case float_round_ties_away:
increment = ((int64_t) absZ1 < 0);
break;
case float_round_to_zero:
increment = 0;
break;
case float_round_up:
increment = !zSign && absZ1;
break;
case float_round_down:
increment = zSign && absZ1;
break;
default:
abort();
}
if ( increment ) {
++absZ0;
if ( absZ0 == 0 ) goto overflow;
absZ0 &= ~ ( ( (uint64_t) ( absZ1<<1 ) == 0 ) & roundNearestEven );
}
z = absZ0;
if ( zSign ) z = - z;
if ( z && ( ( z < 0 ) ^ zSign ) ) {
overflow:
float_raise( float_flag_invalid STATUS_VAR);
return
zSign ? (int64_t) LIT64( 0x8000000000000000 )
: LIT64( 0x7FFFFFFFFFFFFFFF );
}
if ( absZ1 ) STATUS(float_exception_flags) |= float_flag_inexact;
return z;
}
/*----------------------------------------------------------------------------
| Takes the 128-bit fixed-point value formed by concatenating `absZ0' and
| `absZ1', with binary point between bits 63 and 64 (between the input words),
| and returns the properly rounded 64-bit unsigned integer corresponding to the
| input. Ordinarily, the fixed-point input is simply rounded to an integer,
| with the inexact exception raised if the input cannot be represented exactly
| as an integer. However, if the fixed-point input is too large, the invalid
| exception is raised and the largest unsigned integer is returned.
*----------------------------------------------------------------------------*/
static int64 roundAndPackUint64(flag zSign, uint64_t absZ0,
uint64_t absZ1 STATUS_PARAM)
{
int8 roundingMode;
flag roundNearestEven, increment;
roundingMode = STATUS(float_rounding_mode);
roundNearestEven = (roundingMode == float_round_nearest_even);
switch (roundingMode) {
case float_round_nearest_even:
case float_round_ties_away:
increment = ((int64_t)absZ1 < 0);
break;
case float_round_to_zero:
increment = 0;
break;
case float_round_up:
increment = !zSign && absZ1;
break;
case float_round_down:
increment = zSign && absZ1;
break;
default:
abort();
}
if (increment) {
++absZ0;
if (absZ0 == 0) {
float_raise(float_flag_invalid STATUS_VAR);
return LIT64(0xFFFFFFFFFFFFFFFF);
}
absZ0 &= ~(((uint64_t)(absZ1<<1) == 0) & roundNearestEven);
}
if (zSign && absZ0) {
float_raise(float_flag_invalid STATUS_VAR);
return 0;
}
if (absZ1) {
STATUS(float_exception_flags) |= float_flag_inexact;
}
return absZ0;
}
/*----------------------------------------------------------------------------
| Returns the fraction bits of the single-precision floating-point value `a'.
*----------------------------------------------------------------------------*/
INLINE uint32_t extractFloat32Frac( float32 a )
{
return float32_val(a) & 0x007FFFFF;
}
/*----------------------------------------------------------------------------
| Returns the exponent bits of the single-precision floating-point value `a'.
*----------------------------------------------------------------------------*/
INLINE int_fast16_t extractFloat32Exp(float32 a)
{
return ( float32_val(a)>>23 ) & 0xFF;
}
/*----------------------------------------------------------------------------
| Returns the sign bit of the single-precision floating-point value `a'.
*----------------------------------------------------------------------------*/
INLINE flag extractFloat32Sign( float32 a )
{
return float32_val(a)>>31;
}
/*----------------------------------------------------------------------------
| If `a' is denormal and we are in flush-to-zero mode then set the
| input-denormal exception and return zero. Otherwise just return the value.
*----------------------------------------------------------------------------*/
float32 float32_squash_input_denormal(float32 a STATUS_PARAM)
{
if (STATUS(flush_inputs_to_zero)) {
if (extractFloat32Exp(a) == 0 && extractFloat32Frac(a) != 0) {
float_raise(float_flag_input_denormal STATUS_VAR);
return make_float32(float32_val(a) & 0x80000000);
}
}
return a;
}
/*----------------------------------------------------------------------------
| Normalizes the subnormal single-precision floating-point value represented
| by the denormalized significand `aSig'. The normalized exponent and
| significand are stored at the locations pointed to by `zExpPtr' and
| `zSigPtr', respectively.
*----------------------------------------------------------------------------*/
static void
normalizeFloat32Subnormal(uint32_t aSig, int_fast16_t *zExpPtr, uint32_t *zSigPtr)
{
int8 shiftCount;
shiftCount = countLeadingZeros32( aSig ) - 8;
*zSigPtr = aSig<<shiftCount;
*zExpPtr = 1 - shiftCount;
}
/*----------------------------------------------------------------------------
| Packs the sign `zSign', exponent `zExp', and significand `zSig' into a
| single-precision floating-point value, returning the result. After being
| shifted into the proper positions, the three fields are simply added
| together to form the result. This means that any integer portion of `zSig'
| will be added into the exponent. Since a properly normalized significand
| will have an integer portion equal to 1, the `zExp' input should be 1 less
| than the desired result exponent whenever `zSig' is a complete, normalized
| significand.
*----------------------------------------------------------------------------*/
INLINE float32 packFloat32(flag zSign, int_fast16_t zExp, uint32_t zSig)
{
return make_float32(
( ( (uint32_t) zSign )<<31 ) + ( ( (uint32_t) zExp )<<23 ) + zSig);
}
/*----------------------------------------------------------------------------
| Takes an abstract floating-point value having sign `zSign', exponent `zExp',
| and significand `zSig', and returns the proper single-precision floating-
| point value corresponding to the abstract input. Ordinarily, the abstract
| value is simply rounded and packed into the single-precision format, with
| the inexact exception raised if the abstract input cannot be represented
| exactly. However, if the abstract value is too large, the overflow and
| inexact exceptions are raised and an infinity or maximal finite value is
| returned. If the abstract value is too small, the input value is rounded to
| a subnormal number, and the underflow and inexact exceptions are raised if
| the abstract input cannot be represented exactly as a subnormal single-
| precision floating-point number.
| The input significand `zSig' has its binary point between bits 30
| and 29, which is 7 bits to the left of the usual location. This shifted
| significand must be normalized or smaller. If `zSig' is not normalized,
| `zExp' must be 0; in that case, the result returned is a subnormal number,
| and it must not require rounding. In the usual case that `zSig' is
| normalized, `zExp' must be 1 less than the ``true'' floating-point exponent.
| The handling of underflow and overflow follows the IEC/IEEE Standard for
| Binary Floating-Point Arithmetic.
*----------------------------------------------------------------------------*/
static float32 roundAndPackFloat32(flag zSign, int_fast16_t zExp, uint32_t zSig STATUS_PARAM)
{
int8 roundingMode;
flag roundNearestEven;
int8 roundIncrement, roundBits;
flag isTiny;
roundingMode = STATUS(float_rounding_mode);
roundNearestEven = ( roundingMode == float_round_nearest_even );
switch (roundingMode) {
case float_round_nearest_even:
case float_round_ties_away:
roundIncrement = 0x40;
break;
case float_round_to_zero:
roundIncrement = 0;
break;
case float_round_up:
roundIncrement = zSign ? 0 : 0x7f;
break;
case float_round_down:
roundIncrement = zSign ? 0x7f : 0;
break;
default:
abort();
break;
}
roundBits = zSig & 0x7F;
if ( 0xFD <= (uint16_t) zExp ) {
if ( ( 0xFD < zExp )
|| ( ( zExp == 0xFD )
&& ( (int32_t) ( zSig + roundIncrement ) < 0 ) )
) {
float_raise( float_flag_overflow | float_flag_inexact STATUS_VAR);
return packFloat32( zSign, 0xFF, - ( roundIncrement == 0 ));
}
if ( zExp < 0 ) {
if (STATUS(flush_to_zero)) {
float_raise(float_flag_output_denormal STATUS_VAR);
return packFloat32(zSign, 0, 0);
}
isTiny =
( STATUS(float_detect_tininess) == float_tininess_before_rounding )
|| ( zExp < -1 )
|| ( zSig + roundIncrement < 0x80000000 );
shift32RightJamming( zSig, - zExp, &zSig );
zExp = 0;
roundBits = zSig & 0x7F;
if ( isTiny && roundBits ) float_raise( float_flag_underflow STATUS_VAR);
}
}
if ( roundBits ) STATUS(float_exception_flags) |= float_flag_inexact;
zSig = ( zSig + roundIncrement )>>7;
zSig &= ~ ( ( ( roundBits ^ 0x40 ) == 0 ) & roundNearestEven );
if ( zSig == 0 ) zExp = 0;
return packFloat32( zSign, zExp, zSig );
}
/*----------------------------------------------------------------------------
| Takes an abstract floating-point value having sign `zSign', exponent `zExp',
| and significand `zSig', and returns the proper single-precision floating-
| point value corresponding to the abstract input. This routine is just like
| `roundAndPackFloat32' except that `zSig' does not have to be normalized.
| Bit 31 of `zSig' must be zero, and `zExp' must be 1 less than the ``true''
| floating-point exponent.
*----------------------------------------------------------------------------*/
static float32
normalizeRoundAndPackFloat32(flag zSign, int_fast16_t zExp, uint32_t zSig STATUS_PARAM)
{
int8 shiftCount;
shiftCount = countLeadingZeros32( zSig ) - 1;
return roundAndPackFloat32( zSign, zExp - shiftCount, zSig<<shiftCount STATUS_VAR);
}
/*----------------------------------------------------------------------------
| Returns the fraction bits of the double-precision floating-point value `a'.
*----------------------------------------------------------------------------*/
INLINE uint64_t extractFloat64Frac( float64 a )
{
return float64_val(a) & LIT64( 0x000FFFFFFFFFFFFF );
}
/*----------------------------------------------------------------------------
| Returns the exponent bits of the double-precision floating-point value `a'.
*----------------------------------------------------------------------------*/
INLINE int_fast16_t extractFloat64Exp(float64 a)
{
return ( float64_val(a)>>52 ) & 0x7FF;
}
/*----------------------------------------------------------------------------
| Returns the sign bit of the double-precision floating-point value `a'.
*----------------------------------------------------------------------------*/
INLINE flag extractFloat64Sign( float64 a )
{
return float64_val(a)>>63;
}
/*----------------------------------------------------------------------------
| If `a' is denormal and we are in flush-to-zero mode then set the
| input-denormal exception and return zero. Otherwise just return the value.
*----------------------------------------------------------------------------*/
float64 float64_squash_input_denormal(float64 a STATUS_PARAM)
{
if (STATUS(flush_inputs_to_zero)) {
if (extractFloat64Exp(a) == 0 && extractFloat64Frac(a) != 0) {
float_raise(float_flag_input_denormal STATUS_VAR);
return make_float64(float64_val(a) & (1ULL << 63));
}
}
return a;
}
/*----------------------------------------------------------------------------
| Normalizes the subnormal double-precision floating-point value represented
| by the denormalized significand `aSig'. The normalized exponent and
| significand are stored at the locations pointed to by `zExpPtr' and
| `zSigPtr', respectively.
*----------------------------------------------------------------------------*/
static void
normalizeFloat64Subnormal(uint64_t aSig, int_fast16_t *zExpPtr, uint64_t *zSigPtr)
{
int8 shiftCount;
shiftCount = countLeadingZeros64( aSig ) - 11;
*zSigPtr = aSig<<shiftCount;
*zExpPtr = 1 - shiftCount;
}
/*----------------------------------------------------------------------------
| Packs the sign `zSign', exponent `zExp', and significand `zSig' into a
| double-precision floating-point value, returning the result. After being
| shifted into the proper positions, the three fields are simply added
| together to form the result. This means that any integer portion of `zSig'
| will be added into the exponent. Since a properly normalized significand
| will have an integer portion equal to 1, the `zExp' input should be 1 less
| than the desired result exponent whenever `zSig' is a complete, normalized
| significand.
*----------------------------------------------------------------------------*/
INLINE float64 packFloat64(flag zSign, int_fast16_t zExp, uint64_t zSig)
{
return make_float64(
( ( (uint64_t) zSign )<<63 ) + ( ( (uint64_t) zExp )<<52 ) + zSig);
}
/*----------------------------------------------------------------------------
| Takes an abstract floating-point value having sign `zSign', exponent `zExp',
| and significand `zSig', and returns the proper double-precision floating-
| point value corresponding to the abstract input. Ordinarily, the abstract
| value is simply rounded and packed into the double-precision format, with
| the inexact exception raised if the abstract input cannot be represented
| exactly. However, if the abstract value is too large, the overflow and
| inexact exceptions are raised and an infinity or maximal finite value is
| returned. If the abstract value is too small, the input value is rounded
| to a subnormal number, and the underflow and inexact exceptions are raised
| if the abstract input cannot be represented exactly as a subnormal double-
| precision floating-point number.
| The input significand `zSig' has its binary point between bits 62
| and 61, which is 10 bits to the left of the usual location. This shifted
| significand must be normalized or smaller. If `zSig' is not normalized,
| `zExp' must be 0; in that case, the result returned is a subnormal number,
| and it must not require rounding. In the usual case that `zSig' is
| normalized, `zExp' must be 1 less than the ``true'' floating-point exponent.
| The handling of underflow and overflow follows the IEC/IEEE Standard for
| Binary Floating-Point Arithmetic.
*----------------------------------------------------------------------------*/
static float64 roundAndPackFloat64(flag zSign, int_fast16_t zExp, uint64_t zSig STATUS_PARAM)
{
int8 roundingMode;
flag roundNearestEven;
int_fast16_t roundIncrement, roundBits;
flag isTiny;
roundingMode = STATUS(float_rounding_mode);
roundNearestEven = ( roundingMode == float_round_nearest_even );
switch (roundingMode) {
case float_round_nearest_even:
case float_round_ties_away:
roundIncrement = 0x200;
break;
case float_round_to_zero:
roundIncrement = 0;
break;
case float_round_up:
roundIncrement = zSign ? 0 : 0x3ff;
break;
case float_round_down:
roundIncrement = zSign ? 0x3ff : 0;
break;
default:
abort();
}
roundBits = zSig & 0x3FF;
if ( 0x7FD <= (uint16_t) zExp ) {
if ( ( 0x7FD < zExp )
|| ( ( zExp == 0x7FD )
&& ( (int64_t) ( zSig + roundIncrement ) < 0 ) )
) {
float_raise( float_flag_overflow | float_flag_inexact STATUS_VAR);
return packFloat64( zSign, 0x7FF, - ( roundIncrement == 0 ));
}
if ( zExp < 0 ) {
if (STATUS(flush_to_zero)) {
float_raise(float_flag_output_denormal STATUS_VAR);
return packFloat64(zSign, 0, 0);
}
isTiny =
( STATUS(float_detect_tininess) == float_tininess_before_rounding )
|| ( zExp < -1 )
|| ( zSig + roundIncrement < LIT64( 0x8000000000000000 ) );
shift64RightJamming( zSig, - zExp, &zSig );
zExp = 0;
roundBits = zSig & 0x3FF;
if ( isTiny && roundBits ) float_raise( float_flag_underflow STATUS_VAR);
}
}
if ( roundBits ) STATUS(float_exception_flags) |= float_flag_inexact;
zSig = ( zSig + roundIncrement )>>10;
zSig &= ~ ( ( ( roundBits ^ 0x200 ) == 0 ) & roundNearestEven );
if ( zSig == 0 ) zExp = 0;
return packFloat64( zSign, zExp, zSig );
}
/*----------------------------------------------------------------------------
| Takes an abstract floating-point value having sign `zSign', exponent `zExp',
| and significand `zSig', and returns the proper double-precision floating-
| point value corresponding to the abstract input. This routine is just like
| `roundAndPackFloat64' except that `zSig' does not have to be normalized.
| Bit 63 of `zSig' must be zero, and `zExp' must be 1 less than the ``true''
| floating-point exponent.
*----------------------------------------------------------------------------*/
static float64
normalizeRoundAndPackFloat64(flag zSign, int_fast16_t zExp, uint64_t zSig STATUS_PARAM)
{
int8 shiftCount;
shiftCount = countLeadingZeros64( zSig ) - 1;
return roundAndPackFloat64( zSign, zExp - shiftCount, zSig<<shiftCount STATUS_VAR);
}
/*----------------------------------------------------------------------------
| Returns the fraction bits of the extended double-precision floating-point
| value `a'.
*----------------------------------------------------------------------------*/
INLINE uint64_t extractFloatx80Frac( floatx80 a )
{
return a.low;
}
/*----------------------------------------------------------------------------
| Returns the exponent bits of the extended double-precision floating-point
| value `a'.
*----------------------------------------------------------------------------*/
INLINE int32 extractFloatx80Exp( floatx80 a )
{
return a.high & 0x7FFF;
}
/*----------------------------------------------------------------------------
| Returns the sign bit of the extended double-precision floating-point value
| `a'.
*----------------------------------------------------------------------------*/
INLINE flag extractFloatx80Sign( floatx80 a )
{
return a.high>>15;
}
/*----------------------------------------------------------------------------
| Normalizes the subnormal extended double-precision floating-point value
| represented by the denormalized significand `aSig'. The normalized exponent
| and significand are stored at the locations pointed to by `zExpPtr' and
| `zSigPtr', respectively.
*----------------------------------------------------------------------------*/
static void
normalizeFloatx80Subnormal( uint64_t aSig, int32 *zExpPtr, uint64_t *zSigPtr )
{
int8 shiftCount;
shiftCount = countLeadingZeros64( aSig );
*zSigPtr = aSig<<shiftCount;
*zExpPtr = 1 - shiftCount;
}
/*----------------------------------------------------------------------------
| Packs the sign `zSign', exponent `zExp', and significand `zSig' into an
| extended double-precision floating-point value, returning the result.
*----------------------------------------------------------------------------*/
INLINE floatx80 packFloatx80( flag zSign, int32 zExp, uint64_t zSig )
{
floatx80 z;
z.low = zSig;
z.high = ( ( (uint16_t) zSign )<<15 ) + zExp;
return z;
}
/*----------------------------------------------------------------------------
| Takes an abstract floating-point value having sign `zSign', exponent `zExp',
| and extended significand formed by the concatenation of `zSig0' and `zSig1',
| and returns the proper extended double-precision floating-point value
| corresponding to the abstract input. Ordinarily, the abstract value is
| rounded and packed into the extended double-precision format, with the
| inexact exception raised if the abstract input cannot be represented
| exactly. However, if the abstract value is too large, the overflow and
| inexact exceptions are raised and an infinity or maximal finite value is
| returned. If the abstract value is too small, the input value is rounded to
| a subnormal number, and the underflow and inexact exceptions are raised if
| the abstract input cannot be represented exactly as a subnormal extended
| double-precision floating-point number.
| If `roundingPrecision' is 32 or 64, the result is rounded to the same
| number of bits as single or double precision, respectively. Otherwise, the
| result is rounded to the full precision of the extended double-precision
| format.
| The input significand must be normalized or smaller. If the input
| significand is not normalized, `zExp' must be 0; in that case, the result
| returned is a subnormal number, and it must not require rounding. The
| handling of underflow and overflow follows the IEC/IEEE Standard for Binary
| Floating-Point Arithmetic.
*----------------------------------------------------------------------------*/
static floatx80
roundAndPackFloatx80(
int8 roundingPrecision, flag zSign, int32 zExp, uint64_t zSig0, uint64_t zSig1
STATUS_PARAM)
{
int8 roundingMode;
flag roundNearestEven, increment, isTiny;
int64 roundIncrement, roundMask, roundBits;
roundingMode = STATUS(float_rounding_mode);
roundNearestEven = ( roundingMode == float_round_nearest_even );
if ( roundingPrecision == 80 ) goto precision80;
if ( roundingPrecision == 64 ) {
roundIncrement = LIT64( 0x0000000000000400 );
roundMask = LIT64( 0x00000000000007FF );
}
else if ( roundingPrecision == 32 ) {
roundIncrement = LIT64( 0x0000008000000000 );
roundMask = LIT64( 0x000000FFFFFFFFFF );
}
else {
goto precision80;
}
zSig0 |= ( zSig1 != 0 );
switch (roundingMode) {
case float_round_nearest_even:
case float_round_ties_away:
break;
case float_round_to_zero:
roundIncrement = 0;
break;
case float_round_up:
roundIncrement = zSign ? 0 : roundMask;
break;
case float_round_down:
roundIncrement = zSign ? roundMask : 0;
break;
default:
abort();
}
roundBits = zSig0 & roundMask;
if ( 0x7FFD <= (uint32_t) ( zExp - 1 ) ) {
if ( ( 0x7FFE < zExp )
|| ( ( zExp == 0x7FFE ) && ( zSig0 + roundIncrement < zSig0 ) )
) {
goto overflow;
}
if ( zExp <= 0 ) {
if (STATUS(flush_to_zero)) {
float_raise(float_flag_output_denormal STATUS_VAR);
return packFloatx80(zSign, 0, 0);
}
isTiny =
( STATUS(float_detect_tininess) == float_tininess_before_rounding )
|| ( zExp < 0 )
|| ( zSig0 <= zSig0 + roundIncrement );
shift64RightJamming( zSig0, 1 - zExp, &zSig0 );
zExp = 0;
roundBits = zSig0 & roundMask;
if ( isTiny && roundBits ) float_raise( float_flag_underflow STATUS_VAR);
if ( roundBits ) STATUS(float_exception_flags) |= float_flag_inexact;
zSig0 += roundIncrement;
if ( (int64_t) zSig0 < 0 ) zExp = 1;
roundIncrement = roundMask + 1;
if ( roundNearestEven && ( roundBits<<1 == roundIncrement ) ) {
roundMask |= roundIncrement;
}
zSig0 &= ~ roundMask;
return packFloatx80( zSign, zExp, zSig0 );
}
}
if ( roundBits ) STATUS(float_exception_flags) |= float_flag_inexact;
zSig0 += roundIncrement;
if ( zSig0 < roundIncrement ) {
++zExp;
zSig0 = LIT64( 0x8000000000000000 );
}
roundIncrement = roundMask + 1;
if ( roundNearestEven && ( roundBits<<1 == roundIncrement ) ) {
roundMask |= roundIncrement;
}
zSig0 &= ~ roundMask;
if ( zSig0 == 0 ) zExp = 0;
return packFloatx80( zSign, zExp, zSig0 );
precision80:
switch (roundingMode) {
case float_round_nearest_even:
case float_round_ties_away:
increment = ((int64_t)zSig1 < 0);
break;
case float_round_to_zero:
increment = 0;
break;
case float_round_up:
increment = !zSign && zSig1;
break;
case float_round_down:
increment = zSign && zSig1;
break;
default:
abort();
}
if ( 0x7FFD <= (uint32_t) ( zExp - 1 ) ) {
if ( ( 0x7FFE < zExp )
|| ( ( zExp == 0x7FFE )
&& ( zSig0 == LIT64( 0xFFFFFFFFFFFFFFFF ) )
&& increment
)
) {
roundMask = 0;
overflow:
float_raise( float_flag_overflow | float_flag_inexact STATUS_VAR);
if ( ( roundingMode == float_round_to_zero )
|| ( zSign && ( roundingMode == float_round_up ) )
|| ( ! zSign && ( roundingMode == float_round_down ) )
) {
return packFloatx80( zSign, 0x7FFE, ~ roundMask );
}
return packFloatx80( zSign, 0x7FFF, LIT64( 0x8000000000000000 ) );
}
if ( zExp <= 0 ) {
isTiny =
( STATUS(float_detect_tininess) == float_tininess_before_rounding )
|| ( zExp < 0 )
|| ! increment
|| ( zSig0 < LIT64( 0xFFFFFFFFFFFFFFFF ) );
shift64ExtraRightJamming( zSig0, zSig1, 1 - zExp, &zSig0, &zSig1 );
zExp = 0;
if ( isTiny && zSig1 ) float_raise( float_flag_underflow STATUS_VAR);
if ( zSig1 ) STATUS(float_exception_flags) |= float_flag_inexact;
switch (roundingMode) {
case float_round_nearest_even:
case float_round_ties_away:
increment = ((int64_t)zSig1 < 0);
break;
case float_round_to_zero:
increment = 0;
break;
case float_round_up:
increment = !zSign && zSig1;
break;
case float_round_down:
increment = zSign && zSig1;
break;
default:
abort();
}
if ( increment ) {
++zSig0;
zSig0 &=
~ ( ( (uint64_t) ( zSig1<<1 ) == 0 ) & roundNearestEven );
if ( (int64_t) zSig0 < 0 ) zExp = 1;
}
return packFloatx80( zSign, zExp, zSig0 );
}
}
if ( zSig1 ) STATUS(float_exception_flags) |= float_flag_inexact;
if ( increment ) {
++zSig0;
if ( zSig0 == 0 ) {
++zExp;
zSig0 = LIT64( 0x8000000000000000 );
}
else {
zSig0 &= ~ ( ( (uint64_t) ( zSig1<<1 ) == 0 ) & roundNearestEven );
}
}
else {
if ( zSig0 == 0 ) zExp = 0;
}
return packFloatx80( zSign, zExp, zSig0 );
}
/*----------------------------------------------------------------------------
| Takes an abstract floating-point value having sign `zSign', exponent
| `zExp', and significand formed by the concatenation of `zSig0' and `zSig1',
| and returns the proper extended double-precision floating-point value
| corresponding to the abstract input. This routine is just like
| `roundAndPackFloatx80' except that the input significand does not have to be
| normalized.
*----------------------------------------------------------------------------*/
static floatx80
normalizeRoundAndPackFloatx80(
int8 roundingPrecision, flag zSign, int32 zExp, uint64_t zSig0, uint64_t zSig1
STATUS_PARAM)
{
int8 shiftCount;
if ( zSig0 == 0 ) {
zSig0 = zSig1;
zSig1 = 0;
zExp -= 64;
}
shiftCount = countLeadingZeros64( zSig0 );
shortShift128Left( zSig0, zSig1, shiftCount, &zSig0, &zSig1 );
zExp -= shiftCount;
return
roundAndPackFloatx80( roundingPrecision, zSign, zExp, zSig0, zSig1 STATUS_VAR);
}
/*----------------------------------------------------------------------------
| Returns the least-significant 64 fraction bits of the quadruple-precision
| floating-point value `a'.
*----------------------------------------------------------------------------*/
INLINE uint64_t extractFloat128Frac1( float128 a )
{
return a.low;
}
/*----------------------------------------------------------------------------
| Returns the most-significant 48 fraction bits of the quadruple-precision
| floating-point value `a'.
*----------------------------------------------------------------------------*/
INLINE uint64_t extractFloat128Frac0( float128 a )
{
return a.high & LIT64( 0x0000FFFFFFFFFFFF );
}
/*----------------------------------------------------------------------------
| Returns the exponent bits of the quadruple-precision floating-point value
| `a'.
*----------------------------------------------------------------------------*/
INLINE int32 extractFloat128Exp( float128 a )
{
return ( a.high>>48 ) & 0x7FFF;
}
/*----------------------------------------------------------------------------
| Returns the sign bit of the quadruple-precision floating-point value `a'.
*----------------------------------------------------------------------------*/
INLINE flag extractFloat128Sign( float128 a )
{
return a.high>>63;
}
/*----------------------------------------------------------------------------
| Normalizes the subnormal quadruple-precision floating-point value
| represented by the denormalized significand formed by the concatenation of
| `aSig0' and `aSig1'. The normalized exponent is stored at the location
| pointed to by `zExpPtr'. The most significant 49 bits of the normalized
| significand are stored at the location pointed to by `zSig0Ptr', and the
| least significant 64 bits of the normalized significand are stored at the
| location pointed to by `zSig1Ptr'.
*----------------------------------------------------------------------------*/
static void
normalizeFloat128Subnormal(
uint64_t aSig0,
uint64_t aSig1,
int32 *zExpPtr,
uint64_t *zSig0Ptr,
uint64_t *zSig1Ptr
)
{
int8 shiftCount;
if ( aSig0 == 0 ) {
shiftCount = countLeadingZeros64( aSig1 ) - 15;
if ( shiftCount < 0 ) {
*zSig0Ptr = aSig1>>( - shiftCount );
*zSig1Ptr = aSig1<<( shiftCount & 63 );
}
else {
*zSig0Ptr = aSig1<<shiftCount;
*zSig1Ptr = 0;
}
*zExpPtr = - shiftCount - 63;
}
else {
shiftCount = countLeadingZeros64( aSig0 ) - 15;