summaryrefslogtreecommitdiffstats
path: root/src/core/arm/interpreter/vfp/vfpsingle.cpp
blob: 602713cff7666608b1fd9c3372192a14563b918d (plain) (blame)
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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
/*
    vfp/vfpsingle.c - ARM VFPv3 emulation unit - SoftFloat single instruction
    Copyright (C) 2003 Skyeye Develop Group
    for help please send mail to <skyeye-developer@lists.gro.clinux.org>

    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation; either version 2 of the License, or
    (at your option) any later version.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with this program; if not, write to the Free Software
    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
*/

/*
 * This code is derived in part from :
 * - Android kernel
 * - John R. Housers softfloat library, which
 * carries the following notice:
 *
 * ===========================================================================
 * This C source file is part of the SoftFloat IEC/IEEE Floating-point
 * Arithmetic Package, Release 2.
 *
 * 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://HTTP.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 ANY
 * AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE.
 *
 * Derivative works are acceptable, even for commercial purposes, so long as
 * (1) they include prominent notice that the work is derivative, and (2) they
 * include prominent notice akin to these three paragraphs for those parts of
 * this code that are retained.
 * ===========================================================================
 */

#include "core/arm/interpreter/vfp/vfp_helper.h"
#include "core/arm/interpreter/vfp/asm_vfp.h"
#include "core/arm/interpreter/vfp/vfp.h"

static struct vfp_single vfp_single_default_qnan = {
	//.exponent	= 255,
	//.sign		= 0,
	//.significand	= VFP_SINGLE_SIGNIFICAND_QNAN,
};

static void vfp_single_dump(const char *str, struct vfp_single *s)
{
	pr_debug("VFP: %s: sign=%d exponent=%d significand=%08x\n",
		 str, s->sign != 0, s->exponent, s->significand);
}

static void vfp_single_normalise_denormal(struct vfp_single *vs)
{
	int bits = 31 - vfp_fls(vs->significand);

	vfp_single_dump("normalise_denormal: in", vs);

	if (bits) {
		vs->exponent -= bits - 1;
		vs->significand <<= bits;
	}

	vfp_single_dump("normalise_denormal: out", vs);
}


u32 vfp_single_normaliseround(ARMul_State* state, int sd, struct vfp_single *vs, u32 fpscr, u32 exceptions, const char *func)
{
	u32 significand, incr, rmode;
	int exponent, shift, underflow;

	vfp_single_dump("pack: in", vs);

	/*
	 * Infinities and NaNs are a special case.
	 */
	if (vs->exponent == 255 && (vs->significand == 0 || exceptions))
		goto pack;

	/*
	 * Special-case zero.
	 */
	if (vs->significand == 0) {
		vs->exponent = 0;
		goto pack;
	}

	exponent = vs->exponent;
	significand = vs->significand;

	/*
	 * Normalise first.  Note that we shift the significand up to
	 * bit 31, so we have VFP_SINGLE_LOW_BITS + 1 below the least
	 * significant bit.
	 */
	shift = 32 - vfp_fls(significand);
	if (shift < 32 && shift) {
		exponent -= shift;
		significand <<= shift;
	}

#if 1
	vs->exponent = exponent;
	vs->significand = significand;
	vfp_single_dump("pack: normalised", vs);
#endif

	/*
	 * Tiny number?
	 */
	underflow = exponent < 0;
	if (underflow) {
		significand = vfp_shiftright32jamming(significand, -exponent);
		exponent = 0;
#if 1
		vs->exponent = exponent;
		vs->significand = significand;
		vfp_single_dump("pack: tiny number", vs);
#endif
		if (!(significand & ((1 << (VFP_SINGLE_LOW_BITS + 1)) - 1)))
			underflow = 0;
	}

	/*
	 * Select rounding increment.
	 */
	incr = 0;
	rmode = fpscr & FPSCR_RMODE_MASK;

	if (rmode == FPSCR_ROUND_NEAREST) {
		incr = 1 << VFP_SINGLE_LOW_BITS;
		if ((significand & (1 << (VFP_SINGLE_LOW_BITS + 1))) == 0)
			incr -= 1;
	} else if (rmode == FPSCR_ROUND_TOZERO) {
		incr = 0;
	} else if ((rmode == FPSCR_ROUND_PLUSINF) ^ (vs->sign != 0))
		incr = (1 << (VFP_SINGLE_LOW_BITS + 1)) - 1;

	pr_debug("VFP: rounding increment = 0x%08x\n", incr);

	/*
	 * Is our rounding going to overflow?
	 */
	if ((significand + incr) < significand) {
		exponent += 1;
		significand = (significand >> 1) | (significand & 1);
		incr >>= 1;
#if 1
		vs->exponent = exponent;
		vs->significand = significand;
		vfp_single_dump("pack: overflow", vs);
#endif
	}

	/*
	 * If any of the low bits (which will be shifted out of the
	 * number) are non-zero, the result is inexact.
	 */
	if (significand & ((1 << (VFP_SINGLE_LOW_BITS + 1)) - 1))
		exceptions |= FPSCR_IXC;

	/*
	 * Do our rounding.
	 */
	significand += incr;

	/*
	 * Infinity?
	 */
	if (exponent >= 254) {
		exceptions |= FPSCR_OFC | FPSCR_IXC;
		if (incr == 0) {
			vs->exponent = 253;
			vs->significand = 0x7fffffff;
		} else {
			vs->exponent = 255;		/* infinity */
			vs->significand = 0;
		}
	} else {
		if (significand >> (VFP_SINGLE_LOW_BITS + 1) == 0)
			exponent = 0;
		if (exponent || significand > 0x80000000)
			underflow = 0;
		if (underflow)
			exceptions |= FPSCR_UFC;
		vs->exponent = exponent;
		vs->significand = significand >> 1;
	}

 pack:
	vfp_single_dump("pack: final", vs);
	{
		s32 d = vfp_single_pack(vs);
#if 1
		pr_debug("VFP: %s: d(s%d)=%08x exceptions=%08x\n", func,
			 sd, d, exceptions);
#endif
		vfp_put_float(state, d, sd);
	}

	return exceptions;
}

/*
 * Propagate the NaN, setting exceptions if it is signalling.
 * 'n' is always a NaN.  'm' may be a number, NaN or infinity.
 */
static u32
vfp_propagate_nan(struct vfp_single *vsd, struct vfp_single *vsn,
		  struct vfp_single *vsm, u32 fpscr)
{
	struct vfp_single *nan;
	int tn, tm = 0;

	tn = vfp_single_type(vsn);

	if (vsm)
		tm = vfp_single_type(vsm);

	if (fpscr & FPSCR_DEFAULT_NAN)
		/*
		 * Default NaN mode - always returns a quiet NaN
		 */
		nan = &vfp_single_default_qnan;
	else {
		/*
		 * Contemporary mode - select the first signalling
		 * NAN, or if neither are signalling, the first
		 * quiet NAN.
		 */
		if (tn == VFP_SNAN || (tm != VFP_SNAN && tn == VFP_QNAN))
			nan = vsn;
		else
			nan = vsm;
		/*
		 * Make the NaN quiet.
		 */
		nan->significand |= VFP_SINGLE_SIGNIFICAND_QNAN;
	}

	*vsd = *nan;

	/*
	 * If one was a signalling NAN, raise invalid operation.
	 */
	return tn == VFP_SNAN || tm == VFP_SNAN ? FPSCR_IOC : VFP_NAN_FLAG;
}


/*
 * Extended operations
 */
static u32 vfp_single_fabs(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	vfp_put_float(state, vfp_single_packed_abs(m), sd);
	return 0;
}

static u32 vfp_single_fcpy(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	vfp_put_float(state, m, sd);
	return 0;
}

static u32 vfp_single_fneg(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	vfp_put_float(state, vfp_single_packed_negate(m), sd);
	return 0;
}

static const u16 sqrt_oddadjust[] = {
	0x0004, 0x0022, 0x005d, 0x00b1, 0x011d, 0x019f, 0x0236, 0x02e0,
	0x039c, 0x0468, 0x0545, 0x0631, 0x072b, 0x0832, 0x0946, 0x0a67
};

static const u16 sqrt_evenadjust[] = {
	0x0a2d, 0x08af, 0x075a, 0x0629, 0x051a, 0x0429, 0x0356, 0x029e,
	0x0200, 0x0179, 0x0109, 0x00af, 0x0068, 0x0034, 0x0012, 0x0002
};

u32 vfp_estimate_sqrt_significand(u32 exponent, u32 significand)
{
	int index;
	u32 z, a;

	if ((significand & 0xc0000000) != 0x40000000) {
		pr_debug("VFP: estimate_sqrt: invalid significand\n");
	}

	a = significand << 1;
	index = (a >> 27) & 15;
	if (exponent & 1) {
		z = 0x4000 + (a >> 17) - sqrt_oddadjust[index];
		z = ((a / z) << 14) + (z << 15);
		a >>= 1;
	} else {
		z = 0x8000 + (a >> 17) - sqrt_evenadjust[index];
		z = a / z + z;
		z = (z >= 0x20000) ? 0xffff8000 : (z << 15);
		if (z <= a)
			return (s32)a >> 1;
	}
	{
		u64 v = (u64)a << 31;
		do_div(v, z);
		return v + (z >> 1);
	}
}

static u32 vfp_single_fsqrt(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	struct vfp_single vsm, vsd, *vsp;
	int ret, tm;

	vfp_single_unpack(&vsm, m);
	tm = vfp_single_type(&vsm);
	if (tm & (VFP_NAN|VFP_INFINITY)) {
		vsp = &vsd;

		if (tm & VFP_NAN)
			ret = vfp_propagate_nan(vsp, &vsm, NULL, fpscr);
		else if (vsm.sign == 0) {
 sqrt_copy:
			vsp = &vsm;
			ret = 0;
		} else {
 sqrt_invalid:
			vsp = &vfp_single_default_qnan;
			ret = FPSCR_IOC;
		}
		vfp_put_float(state, vfp_single_pack(vsp), sd);
		return ret;
	}

	/*
	 * sqrt(+/- 0) == +/- 0
	 */
	if (tm & VFP_ZERO)
		goto sqrt_copy;

	/*
	 * Normalise a denormalised number
	 */
	if (tm & VFP_DENORMAL)
		vfp_single_normalise_denormal(&vsm);

	/*
	 * sqrt(<0) = invalid
	 */
	if (vsm.sign)
		goto sqrt_invalid;

	vfp_single_dump("sqrt", &vsm);

	/*
	 * Estimate the square root.
	 */
	vsd.sign = 0;
	vsd.exponent = ((vsm.exponent - 127) >> 1) + 127;
	vsd.significand = vfp_estimate_sqrt_significand(vsm.exponent, vsm.significand) + 2;

	vfp_single_dump("sqrt estimate", &vsd);

	/*
	 * And now adjust.
	 */
	if ((vsd.significand & VFP_SINGLE_LOW_BITS_MASK) <= 5) {
		if (vsd.significand < 2) {
			vsd.significand = 0xffffffff;
		} else {
			u64 term;
			s64 rem;
			vsm.significand <<= !(vsm.exponent & 1);
			term = (u64)vsd.significand * vsd.significand;
			rem = ((u64)vsm.significand << 32) - term;

			pr_debug("VFP: term=%016llx rem=%016llx\n", term, rem);

			while (rem < 0) {
				vsd.significand -= 1;
				rem += ((u64)vsd.significand << 1) | 1;
			}
			vsd.significand |= rem != 0;
		}
	}
	vsd.significand = vfp_shiftright32jamming(vsd.significand, 1);

	return vfp_single_normaliseround(state, sd, &vsd, fpscr, 0, "fsqrt");
}

/*
 * Equal	:= ZC
 * Less than	:= N
 * Greater than	:= C
 * Unordered	:= CV
 */
static u32 vfp_compare(ARMul_State* state, int sd, int signal_on_qnan, s32 m, u32 fpscr)
{
	s32 d;
	u32 ret = 0;

	d = vfp_get_float(state, sd);
	if (vfp_single_packed_exponent(m) == 255 && vfp_single_packed_mantissa(m)) {
		ret |= FPSCR_C | FPSCR_V;
		if (signal_on_qnan || !(vfp_single_packed_mantissa(m) & (1 << (VFP_SINGLE_MANTISSA_BITS - 1))))
			/*
			 * Signalling NaN, or signalling on quiet NaN
			 */
			ret |= FPSCR_IOC;
	}

	if (vfp_single_packed_exponent(d) == 255 && vfp_single_packed_mantissa(d)) {
		ret |= FPSCR_C | FPSCR_V;
		if (signal_on_qnan || !(vfp_single_packed_mantissa(d) & (1 << (VFP_SINGLE_MANTISSA_BITS - 1))))
			/*
			 * Signalling NaN, or signalling on quiet NaN
			 */
			ret |= FPSCR_IOC;
	}

	if (ret == 0) {
		if (d == m || vfp_single_packed_abs(d | m) == 0) {
			/*
			 * equal
			 */
			ret |= FPSCR_Z | FPSCR_C;
		} else if (vfp_single_packed_sign(d ^ m)) {
			/*
			 * different signs
			 */
			if (vfp_single_packed_sign(d))
				/*
				 * d is negative, so d < m
				 */
				ret |= FPSCR_N;
			else
				/*
				 * d is positive, so d > m
				 */
				ret |= FPSCR_C;
		} else if ((vfp_single_packed_sign(d) != 0) ^ (d < m)) {
			/*
			 * d < m
			 */
			ret |= FPSCR_N;
		} else if ((vfp_single_packed_sign(d) != 0) ^ (d > m)) {
			/*
			 * d > m
			 */
			ret |= FPSCR_C;
		}
	}
	return ret;
}

static u32 vfp_single_fcmp(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	return vfp_compare(state, sd, 0, m, fpscr);
}

static u32 vfp_single_fcmpe(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	return vfp_compare(state, sd, 1, m, fpscr);
}

static u32 vfp_single_fcmpz(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	return vfp_compare(state, sd, 0, 0, fpscr);
}

static u32 vfp_single_fcmpez(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	return vfp_compare(state, sd, 1, 0, fpscr);
}

static u32 vfp_single_fcvtd(ARMul_State* state, int dd, int unused, s32 m, u32 fpscr)
{
	struct vfp_single vsm;
	struct vfp_double vdd;
	int tm;
	u32 exceptions = 0;

	vfp_single_unpack(&vsm, m);

	tm = vfp_single_type(&vsm);

	/*
	 * If we have a signalling NaN, signal invalid operation.
	 */
	if (tm == VFP_SNAN)
		exceptions = FPSCR_IOC;

	if (tm & VFP_DENORMAL)
		vfp_single_normalise_denormal(&vsm);

	vdd.sign = vsm.sign;
	vdd.significand = (u64)vsm.significand << 32;

	/*
	 * If we have an infinity or NaN, the exponent must be 2047.
	 */
	if (tm & (VFP_INFINITY|VFP_NAN)) {
		vdd.exponent = 2047;
		if (tm == VFP_QNAN)
			vdd.significand |= VFP_DOUBLE_SIGNIFICAND_QNAN;
		goto pack_nan;
	} else if (tm & VFP_ZERO)
		vdd.exponent = 0;
	else
		vdd.exponent = vsm.exponent + (1023 - 127);

	return vfp_double_normaliseround(state, dd, &vdd, fpscr, exceptions, "fcvtd");

 pack_nan:
	vfp_put_double(state, vfp_double_pack(&vdd), dd);
	return exceptions;
}

static u32 vfp_single_fuito(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	struct vfp_single vs;

	vs.sign = 0;
	vs.exponent = 127 + 31 - 1;
	vs.significand = (u32)m;

	return vfp_single_normaliseround(state, sd, &vs, fpscr, 0, "fuito");
}

static u32 vfp_single_fsito(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	struct vfp_single vs;

	vs.sign = (m & 0x80000000) >> 16;
	vs.exponent = 127 + 31 - 1;
	vs.significand = vs.sign ? -m : m;

	return vfp_single_normaliseround(state, sd, &vs, fpscr, 0, "fsito");
}

static u32 vfp_single_ftoui(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	struct vfp_single vsm;
	u32 d, exceptions = 0;
	int rmode = fpscr & FPSCR_RMODE_MASK;
	int tm;

	vfp_single_unpack(&vsm, m);
	vfp_single_dump("VSM", &vsm);

	/*
	 * Do we have a denormalised number?
	 */
	tm = vfp_single_type(&vsm);
	if (tm & VFP_DENORMAL)
		exceptions |= FPSCR_IDC;

	if (tm & VFP_NAN)
		vsm.sign = 0;

	if (vsm.exponent >= 127 + 32) {
		d = vsm.sign ? 0 : 0xffffffff;
		exceptions = FPSCR_IOC;
	} else if (vsm.exponent >= 127 - 1) {
		int shift = 127 + 31 - vsm.exponent;
		u32 rem, incr = 0;

		/*
		 * 2^0 <= m < 2^32-2^8
		 */
		d = (vsm.significand << 1) >> shift;
		rem = vsm.significand << (33 - shift);

		if (rmode == FPSCR_ROUND_NEAREST) {
			incr = 0x80000000;
			if ((d & 1) == 0)
				incr -= 1;
		} else if (rmode == FPSCR_ROUND_TOZERO) {
			incr = 0;
		} else if ((rmode == FPSCR_ROUND_PLUSINF) ^ (vsm.sign != 0)) {
			incr = ~0;
		}

		if ((rem + incr) < rem) {
			if (d < 0xffffffff)
				d += 1;
			else
				exceptions |= FPSCR_IOC;
		}

		if (d && vsm.sign) {
			d = 0;
			exceptions |= FPSCR_IOC;
		} else if (rem)
			exceptions |= FPSCR_IXC;
	} else {
		d = 0;
		if (vsm.exponent | vsm.significand) {
			exceptions |= FPSCR_IXC;
			if (rmode == FPSCR_ROUND_PLUSINF && vsm.sign == 0)
				d = 1;
			else if (rmode == FPSCR_ROUND_MINUSINF && vsm.sign) {
				d = 0;
				exceptions |= FPSCR_IOC;
			}
		}
	}

	pr_debug("VFP: ftoui: d(s%d)=%08x exceptions=%08x\n", sd, d, exceptions);

	vfp_put_float(state, d, sd);

	return exceptions;
}

static u32 vfp_single_ftouiz(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	return vfp_single_ftoui(state, sd, unused, m, FPSCR_ROUND_TOZERO);
}

static u32 vfp_single_ftosi(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	struct vfp_single vsm;
	u32 d, exceptions = 0;
	int rmode = fpscr & FPSCR_RMODE_MASK;
	int tm;

	vfp_single_unpack(&vsm, m);
	vfp_single_dump("VSM", &vsm);

	/*
	 * Do we have a denormalised number?
	 */
	tm = vfp_single_type(&vsm);
	if (vfp_single_type(&vsm) & VFP_DENORMAL)
		exceptions |= FPSCR_IDC;

	if (tm & VFP_NAN) {
		d = 0;
		exceptions |= FPSCR_IOC;
	} else if (vsm.exponent >= 127 + 32) {
		/*
		 * m >= 2^31-2^7: invalid
		 */
		d = 0x7fffffff;
		if (vsm.sign)
			d = ~d;
		exceptions |= FPSCR_IOC;
	} else if (vsm.exponent >= 127 - 1) {
		int shift = 127 + 31 - vsm.exponent;
		u32 rem, incr = 0;

		/* 2^0 <= m <= 2^31-2^7 */
		d = (vsm.significand << 1) >> shift;
		rem = vsm.significand << (33 - shift);

		if (rmode == FPSCR_ROUND_NEAREST) {
			incr = 0x80000000;
			if ((d & 1) == 0)
				incr -= 1;
		} else if (rmode == FPSCR_ROUND_TOZERO) {
			incr = 0;
		} else if ((rmode == FPSCR_ROUND_PLUSINF) ^ (vsm.sign != 0)) {
			incr = ~0;
		}

		if ((rem + incr) < rem && d < 0xffffffff)
			d += 1;
		if (d > 0x7fffffff + (vsm.sign != 0)) {
			d = 0x7fffffff + (vsm.sign != 0);
			exceptions |= FPSCR_IOC;
		} else if (rem)
			exceptions |= FPSCR_IXC;

		if (vsm.sign)
			d = -d;
	} else {
		d = 0;
		if (vsm.exponent | vsm.significand) {
			exceptions |= FPSCR_IXC;
			if (rmode == FPSCR_ROUND_PLUSINF && vsm.sign == 0)
				d = 1;
			else if (rmode == FPSCR_ROUND_MINUSINF && vsm.sign)
				d = -1;
		}
	}

	pr_debug("VFP: ftosi: d(s%d)=%08x exceptions=%08x\n", sd, d, exceptions);

	vfp_put_float(state, (s32)d, sd);

	return exceptions;
}

static u32 vfp_single_ftosiz(ARMul_State* state, int sd, int unused, s32 m, u32 fpscr)
{
	return vfp_single_ftosi(state, sd, unused, m, FPSCR_ROUND_TOZERO);
}

static struct op fops_ext[] = {
    { vfp_single_fcpy,   0 },                 //0x00000000 - FEXT_FCPY
    { vfp_single_fabs,   0 },                 //0x00000001 - FEXT_FABS
    { vfp_single_fneg,   0 },                 //0x00000002 - FEXT_FNEG
    { vfp_single_fsqrt,  0 },                 //0x00000003 - FEXT_FSQRT
    { NULL, 0 },
    { NULL, 0 },
    { NULL, 0 },
    { NULL, 0 },
    { vfp_single_fcmp,   OP_SCALAR },         //0x00000008 - FEXT_FCMP
    { vfp_single_fcmpe,  OP_SCALAR },         //0x00000009 - FEXT_FCMPE
    { vfp_single_fcmpz,  OP_SCALAR },         //0x0000000A - FEXT_FCMPZ
    { vfp_single_fcmpez, OP_SCALAR },         //0x0000000B - FEXT_FCMPEZ
    { NULL, 0 },
    { NULL, 0 },
    { NULL, 0 },
    { vfp_single_fcvtd,  OP_SCALAR|OP_DD },   //0x0000000F - FEXT_FCVT
    { vfp_single_fuito,  OP_SCALAR },         //0x00000010 - FEXT_FUITO
    { vfp_single_fsito,  OP_SCALAR },         //0x00000011 - FEXT_FSITO
    { NULL, 0 },
    { NULL, 0 },
    { NULL, 0 },
    { NULL, 0 },
    { NULL, 0 },
    { NULL, 0 },
    { vfp_single_ftoui,  OP_SCALAR },         //0x00000018 - FEXT_FTOUI
    { vfp_single_ftouiz, OP_SCALAR },         //0x00000019 - FEXT_FTOUIZ
    { vfp_single_ftosi,  OP_SCALAR },         //0x0000001A - FEXT_FTOSI
    { vfp_single_ftosiz, OP_SCALAR },         //0x0000001B - FEXT_FTOSIZ
};





static u32
vfp_single_fadd_nonnumber(struct vfp_single *vsd, struct vfp_single *vsn,
			  struct vfp_single *vsm, u32 fpscr)
{
	struct vfp_single *vsp;
	u32 exceptions = 0;
	int tn, tm;

	tn = vfp_single_type(vsn);
	tm = vfp_single_type(vsm);

	if (tn & tm & VFP_INFINITY) {
		/*
		 * Two infinities.  Are they different signs?
		 */
		if (vsn->sign ^ vsm->sign) {
			/*
			 * different signs -> invalid
			 */
			exceptions = FPSCR_IOC;
			vsp = &vfp_single_default_qnan;
		} else {
			/*
			 * same signs -> valid
			 */
			vsp = vsn;
		}
	} else if (tn & VFP_INFINITY && tm & VFP_NUMBER) {
		/*
		 * One infinity and one number -> infinity
		 */
		vsp = vsn;
	} else {
		/*
		 * 'n' is a NaN of some type
		 */
		return vfp_propagate_nan(vsd, vsn, vsm, fpscr);
	}
	*vsd = *vsp;
	return exceptions;
}

static u32
vfp_single_add(struct vfp_single *vsd, struct vfp_single *vsn,
	       struct vfp_single *vsm, u32 fpscr)
{
	u32 exp_diff, m_sig;

	if (vsn->significand & 0x80000000 ||
	    vsm->significand & 0x80000000) {
		pr_info("VFP: bad FP values\n");
		vfp_single_dump("VSN", vsn);
		vfp_single_dump("VSM", vsm);
	}

	/*
	 * Ensure that 'n' is the largest magnitude number.  Note that
	 * if 'n' and 'm' have equal exponents, we do not swap them.
	 * This ensures that NaN propagation works correctly.
	 */
	if (vsn->exponent < vsm->exponent) {
		struct vfp_single *t = vsn;
		vsn = vsm;
		vsm = t;
	}

	/*
	 * Is 'n' an infinity or a NaN?  Note that 'm' may be a number,
	 * infinity or a NaN here.
	 */
	if (vsn->exponent == 255)
		return vfp_single_fadd_nonnumber(vsd, vsn, vsm, fpscr);

	/*
	 * We have two proper numbers, where 'vsn' is the larger magnitude.
	 *
	 * Copy 'n' to 'd' before doing the arithmetic.
	 */
	*vsd = *vsn;

	/*
	 * Align both numbers.
	 */
	exp_diff = vsn->exponent - vsm->exponent;
	m_sig = vfp_shiftright32jamming(vsm->significand, exp_diff);

	/*
	 * If the signs are different, we are really subtracting.
	 */
	if (vsn->sign ^ vsm->sign) {
		m_sig = vsn->significand - m_sig;
		if ((s32)m_sig < 0) {
			vsd->sign = vfp_sign_negate(vsd->sign);
			m_sig = -m_sig;
		} else if (m_sig == 0) {
			vsd->sign = (fpscr & FPSCR_RMODE_MASK) ==
				      FPSCR_ROUND_MINUSINF ? 0x8000 : 0;
		}
	} else {
		m_sig = vsn->significand + m_sig;
	}
	vsd->significand = m_sig;

	return 0;
}

static u32
vfp_single_multiply(struct vfp_single *vsd, struct vfp_single *vsn, struct vfp_single *vsm, u32 fpscr)
{
	vfp_single_dump("VSN", vsn);
	vfp_single_dump("VSM", vsm);

	/*
	 * Ensure that 'n' is the largest magnitude number.  Note that
	 * if 'n' and 'm' have equal exponents, we do not swap them.
	 * This ensures that NaN propagation works correctly.
	 */
	if (vsn->exponent < vsm->exponent) {
		struct vfp_single *t = vsn;
		vsn = vsm;
		vsm = t;
		pr_debug("VFP: swapping M <-> N\n");
	}

	vsd->sign = vsn->sign ^ vsm->sign;

	/*
	 * If 'n' is an infinity or NaN, handle it.  'm' may be anything.
	 */
	if (vsn->exponent == 255) {
		if (vsn->significand || (vsm->exponent == 255 && vsm->significand))
			return vfp_propagate_nan(vsd, vsn, vsm, fpscr);
		if ((vsm->exponent | vsm->significand) == 0) {
			*vsd = vfp_single_default_qnan;
			return FPSCR_IOC;
		}
		vsd->exponent = vsn->exponent;
		vsd->significand = 0;
		return 0;
	}

	/*
	 * If 'm' is zero, the result is always zero.  In this case,
	 * 'n' may be zero or a number, but it doesn't matter which.
	 */
	if ((vsm->exponent | vsm->significand) == 0) {
		vsd->exponent = 0;
		vsd->significand = 0;
		return 0;
	}

	/*
	 * We add 2 to the destination exponent for the same reason as
	 * the addition case - though this time we have +1 from each
	 * input operand.
	 */
	vsd->exponent = vsn->exponent + vsm->exponent - 127 + 2;
	vsd->significand = vfp_hi64to32jamming((u64)vsn->significand * vsm->significand);

	vfp_single_dump("VSD", vsd);
	return 0;
}

#define NEG_MULTIPLY	(1 << 0)
#define NEG_SUBTRACT	(1 << 1)

static u32
vfp_single_multiply_accumulate(ARMul_State* state, int sd, int sn, s32 m, u32 fpscr, u32 negate, char *func)
{
	struct vfp_single vsd, vsp, vsn, vsm;
	u32 exceptions;
	s32 v;

	v = vfp_get_float(state, sn);
	pr_debug("VFP: s%u = %08x\n", sn, v);
	vfp_single_unpack(&vsn, v);
	if (vsn.exponent == 0 && vsn.significand)
		vfp_single_normalise_denormal(&vsn);

	vfp_single_unpack(&vsm, m);
	if (vsm.exponent == 0 && vsm.significand)
		vfp_single_normalise_denormal(&vsm);

	exceptions = vfp_single_multiply(&vsp, &vsn, &vsm, fpscr);
	if (negate & NEG_MULTIPLY)
		vsp.sign = vfp_sign_negate(vsp.sign);

	v = vfp_get_float(state, sd);
	pr_debug("VFP: s%u = %08x\n", sd, v);
	vfp_single_unpack(&vsn, v);
	if (negate & NEG_SUBTRACT)
		vsn.sign = vfp_sign_negate(vsn.sign);

	exceptions |= vfp_single_add(&vsd, &vsn, &vsp, fpscr);

	return vfp_single_normaliseround(state, sd, &vsd, fpscr, exceptions, func);
}

/*
 * Standard operations
 */

/*
 * sd = sd + (sn * sm)
 */
static u32 vfp_single_fmac(ARMul_State* state, int sd, int sn, s32 m, u32 fpscr)
{
	pr_debug("In %sVFP: s%u = %08x\n", __FUNCTION__, sn, sd);
	return vfp_single_multiply_accumulate(state, sd, sn, m, fpscr, 0, "fmac");
}

/*
 * sd = sd - (sn * sm)
 */
static u32 vfp_single_fnmac(ARMul_State* state, int sd, int sn, s32 m, u32 fpscr)
{
	pr_debug("In %sVFP: s%u = %08x\n", __FUNCTION__, sd, sn);
	return vfp_single_multiply_accumulate(state, sd, sn, m, fpscr, NEG_MULTIPLY, "fnmac");
}

/*
 * sd = -sd + (sn * sm)
 */
static u32 vfp_single_fmsc(ARMul_State* state, int sd, int sn, s32 m, u32 fpscr)
{
	pr_debug("In %sVFP: s%u = %08x\n", __FUNCTION__, sn, sd);
	return vfp_single_multiply_accumulate(state, sd, sn, m, fpscr, NEG_SUBTRACT, "fmsc");
}

/*
 * sd = -sd - (sn * sm)
 */
static u32 vfp_single_fnmsc(ARMul_State* state, int sd, int sn, s32 m, u32 fpscr)
{
	pr_debug("In %sVFP: s%u = %08x\n", __FUNCTION__, sn, sd);
	return vfp_single_multiply_accumulate(state, sd, sn, m, fpscr, NEG_SUBTRACT | NEG_MULTIPLY, "fnmsc");
}

/*
 * sd = sn * sm
 */
static u32 vfp_single_fmul(ARMul_State* state, int sd, int sn, s32 m, u32 fpscr)
{
	struct vfp_single vsd, vsn, vsm;
	u32 exceptions;
	s32 n = vfp_get_float(state, sn);

	pr_debug("In %sVFP: s%u = %08x\n", __FUNCTION__, sn, n);

	vfp_single_unpack(&vsn, n);
	if (vsn.exponent == 0 && vsn.significand)
		vfp_single_normalise_denormal(&vsn);

	vfp_single_unpack(&vsm, m);
	if (vsm.exponent == 0 && vsm.significand)
		vfp_single_normalise_denormal(&vsm);

	exceptions = vfp_single_multiply(&vsd, &vsn, &vsm, fpscr);
	return vfp_single_normaliseround(state, sd, &vsd, fpscr, exceptions, "fmul");
}

/*
 * sd = -(sn * sm)
 */
static u32 vfp_single_fnmul(ARMul_State* state, int sd, int sn, s32 m, u32 fpscr)
{
	struct vfp_single vsd, vsn, vsm;
	u32 exceptions;
	s32 n = vfp_get_float(state, sn);

	pr_debug("VFP: s%u = %08x\n", sn, n);

	vfp_single_unpack(&vsn, n);
	if (vsn.exponent == 0 && vsn.significand)
		vfp_single_normalise_denormal(&vsn);

	vfp_single_unpack(&vsm, m);
	if (vsm.exponent == 0 && vsm.significand)
		vfp_single_normalise_denormal(&vsm);

	exceptions = vfp_single_multiply(&vsd, &vsn, &vsm, fpscr);
	vsd.sign = vfp_sign_negate(vsd.sign);
	return vfp_single_normaliseround(state, sd, &vsd, fpscr, exceptions, "fnmul");
}

/*
 * sd = sn + sm
 */
static u32 vfp_single_fadd(ARMul_State* state, int sd, int sn, s32 m, u32 fpscr)
{
	struct vfp_single vsd, vsn, vsm;
	u32 exceptions;
	s32 n = vfp_get_float(state, sn);

	pr_debug("VFP: s%u = %08x\n", sn, n);

	/*
	 * Unpack and normalise denormals.
	 */
	vfp_single_unpack(&vsn, n);
	if (vsn.exponent == 0 && vsn.significand)
		vfp_single_normalise_denormal(&vsn);

	vfp_single_unpack(&vsm, m);
	if (vsm.exponent == 0 && vsm.significand)
		vfp_single_normalise_denormal(&vsm);

	exceptions = vfp_single_add(&vsd, &vsn, &vsm, fpscr);

	return vfp_single_normaliseround(state, sd, &vsd, fpscr, exceptions, "fadd");
}

/*
 * sd = sn - sm
 */
static u32 vfp_single_fsub(ARMul_State* state, int sd, int sn, s32 m, u32 fpscr)
{
	pr_debug("In %sVFP: s%u = %08x\n", __FUNCTION__, sn, sd);
	/*
	 * Subtraction is addition with one sign inverted.
	 */
	return vfp_single_fadd(state, sd, sn, vfp_single_packed_negate(m), fpscr);
}

/*
 * sd = sn / sm
 */
static u32 vfp_single_fdiv(ARMul_State* state, int sd, int sn, s32 m, u32 fpscr)
{
	struct vfp_single vsd, vsn, vsm;
	u32 exceptions = 0;
	s32 n = vfp_get_float(state, sn);
	int tm, tn;

	pr_debug("VFP: s%u = %08x\n", sn, n);

	vfp_single_unpack(&vsn, n);
	vfp_single_unpack(&vsm, m);

	vsd.sign = vsn.sign ^ vsm.sign;

	tn = vfp_single_type(&vsn);
	tm = vfp_single_type(&vsm);

	/*
	 * Is n a NAN?
	 */
	if (tn & VFP_NAN)
		goto vsn_nan;

	/*
	 * Is m a NAN?
	 */
	if (tm & VFP_NAN)
		goto vsm_nan;

	/*
	 * If n and m are infinity, the result is invalid
	 * If n and m are zero, the result is invalid
	 */
	if (tm & tn & (VFP_INFINITY|VFP_ZERO))
		goto invalid;

	/*
	 * If n is infinity, the result is infinity
	 */
	if (tn & VFP_INFINITY)
		goto infinity;

	/*
	 * If m is zero, raise div0 exception
	 */
	if (tm & VFP_ZERO)
		goto divzero;

	/*
	 * If m is infinity, or n is zero, the result is zero
	 */
	if (tm & VFP_INFINITY || tn & VFP_ZERO)
		goto zero;

	if (tn & VFP_DENORMAL)
		vfp_single_normalise_denormal(&vsn);
	if (tm & VFP_DENORMAL)
		vfp_single_normalise_denormal(&vsm);

	/*
	 * Ok, we have two numbers, we can perform division.
	 */
	vsd.exponent = vsn.exponent - vsm.exponent + 127 - 1;
	vsm.significand <<= 1;
	if (vsm.significand <= (2 * vsn.significand)) {
		vsn.significand >>= 1;
		vsd.exponent++;
	}
	{
		u64 significand = (u64)vsn.significand << 32;
		do_div(significand, vsm.significand);
		vsd.significand = significand;
	}
	if ((vsd.significand & 0x3f) == 0)
		vsd.significand |= ((u64)vsm.significand * vsd.significand != (u64)vsn.significand << 32);

	return vfp_single_normaliseround(state, sd, &vsd, fpscr, 0, "fdiv");

 vsn_nan:
	exceptions = vfp_propagate_nan(&vsd, &vsn, &vsm, fpscr);
 pack:
	vfp_put_float(state, vfp_single_pack(&vsd), sd);
	return exceptions;

 vsm_nan:
	exceptions = vfp_propagate_nan(&vsd, &vsm, &vsn, fpscr);
	goto pack;

 zero:
	vsd.exponent = 0;
	vsd.significand = 0;
	goto pack;

 divzero:
	exceptions = FPSCR_DZC;
 infinity:
	vsd.exponent = 255;
	vsd.significand = 0;
	goto pack;

 invalid:
	vfp_put_float(state, vfp_single_pack(&vfp_single_default_qnan), sd);
	return FPSCR_IOC;
}

static struct op fops[] = {
	{ vfp_single_fmac,  0 },
	{ vfp_single_fmsc,  0 },
	{ vfp_single_fmul,  0 },
	{ vfp_single_fadd,  0 },
	{ vfp_single_fnmac, 0 },
	{ vfp_single_fnmsc, 0 },
	{ vfp_single_fnmul, 0 },
	{ vfp_single_fsub,  0 },
	{ vfp_single_fdiv,  0 },
};

#define FREG_BANK(x)	((x) & 0x18)
#define FREG_IDX(x)	((x) & 7)

u32 vfp_single_cpdo(ARMul_State* state, u32 inst, u32 fpscr)
{
	u32 op = inst & FOP_MASK;
	u32 exceptions = 0;
	unsigned int dest;
	unsigned int sn = vfp_get_sn(inst);
	unsigned int sm = vfp_get_sm(inst);
	unsigned int vecitr, veclen, vecstride;
	struct op *fop;
	pr_debug("In %s\n", __FUNCTION__);

	vecstride = 1 + ((fpscr & FPSCR_STRIDE_MASK) == FPSCR_STRIDE_MASK);

	fop = (op == FOP_EXT) ? &fops_ext[FEXT_TO_IDX(inst)] : &fops[FOP_TO_IDX(op)];

	/*
	 * fcvtsd takes a dN register number as destination, not sN.
	 * Technically, if bit 0 of dd is set, this is an invalid
	 * instruction.  However, we ignore this for efficiency.
	 * It also only operates on scalars.
	 */
	if (fop->flags & OP_DD)
		dest = vfp_get_dd(inst);
	else
		dest = vfp_get_sd(inst);

	/*
	 * If destination bank is zero, vector length is always '1'.
	 * ARM DDI0100F C5.1.3, C5.3.2.
	 */
	if ((fop->flags & OP_SCALAR) || FREG_BANK(dest) == 0)
		veclen = 0;
	else
		veclen = fpscr & FPSCR_LENGTH_MASK;

	pr_debug("VFP: vecstride=%u veclen=%u\n", vecstride,
		 (veclen >> FPSCR_LENGTH_BIT) + 1);

	if (!fop->fn) {
		printf("VFP: could not find single op %d, inst=0x%x@0x%x\n", FEXT_TO_IDX(inst), inst, state->Reg[15]);
		exit(-1);
		goto invalid;
	}

	for (vecitr = 0; vecitr <= veclen; vecitr += 1 << FPSCR_LENGTH_BIT) {
		s32 m = vfp_get_float(state, sm);
		u32 except;
		char type;

		type = fop->flags & OP_DD ? 'd' : 's';
		if (op == FOP_EXT)
			pr_debug("VFP: itr%d (%c%u) = op[%u] (s%u=%08x)\n",
				 vecitr >> FPSCR_LENGTH_BIT, type, dest, sn,
				 sm, m);
		else
			pr_debug("VFP: itr%d (%c%u) = (s%u) op[%u] (s%u=%08x)\n",
				 vecitr >> FPSCR_LENGTH_BIT, type, dest, sn,
				 FOP_TO_IDX(op), sm, m);

		except = fop->fn(state, dest, sn, m, fpscr);
		pr_debug("VFP: itr%d: exceptions=%08x\n",
			 vecitr >> FPSCR_LENGTH_BIT, except);

		exceptions |= except;

		/*
		 * CHECK: It appears to be undefined whether we stop when
		 * we encounter an exception.  We continue.
		 */
		dest = FREG_BANK(dest) + ((FREG_IDX(dest) + vecstride) & 7);
		sn = FREG_BANK(sn) + ((FREG_IDX(sn) + vecstride) & 7);
		if (FREG_BANK(sm) != 0)
			sm = FREG_BANK(sm) + ((FREG_IDX(sm) + vecstride) & 7);
	}
	return exceptions;

 invalid:
	return (u32)-1;
}