summaryrefslogtreecommitdiff
path: root/src/arch/arm/insts/vfp.hh
blob: caa22376e9a62061ca5dfdda5a4d647dead07d99 (plain)
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
/*
 * Copyright (c) 2010-2013 ARM Limited
 * All rights reserved
 *
 * The license below extends only to copyright in the software and shall
 * not be construed as granting a license to any other intellectual
 * property including but not limited to intellectual property relating
 * to a hardware implementation of the functionality of the software
 * licensed hereunder.  You may use the software subject to the license
 * terms below provided that you ensure that this notice is replicated
 * unmodified and in its entirety in all distributions of the software,
 * modified or unmodified, in source code or in binary form.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are
 * met: redistributions of source code must retain the above copyright
 * notice, this list of conditions and the following disclaimer;
 * redistributions in binary form must reproduce the above copyright
 * notice, this list of conditions and the following disclaimer in the
 * documentation and/or other materials provided with the distribution;
 * neither the name of the copyright holders nor the names of its
 * contributors may be used to endorse or promote products derived from
 * this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * Authors: Gabe Black
 */

#ifndef __ARCH_ARM_INSTS_VFP_HH__
#define __ARCH_ARM_INSTS_VFP_HH__

#include <fenv.h>

#include <cmath>

#include "arch/arm/insts/misc.hh"
#include "arch/arm/miscregs.hh"

namespace ArmISA
{

enum VfpMicroMode {
    VfpNotAMicroop,
    VfpMicroop,
    VfpFirstMicroop,
    VfpLastMicroop
};

template<class T>
static inline void
setVfpMicroFlags(VfpMicroMode mode, T &flags)
{
    switch (mode) {
      case VfpMicroop:
        flags[StaticInst::IsMicroop] = true;
        break;
      case VfpFirstMicroop:
        flags[StaticInst::IsMicroop] =
            flags[StaticInst::IsFirstMicroop] = true;
        break;
      case VfpLastMicroop:
        flags[StaticInst::IsMicroop] =
            flags[StaticInst::IsLastMicroop] = true;
        break;
      case VfpNotAMicroop:
        break;
    }
    if (mode == VfpMicroop || mode == VfpFirstMicroop) {
        flags[StaticInst::IsDelayedCommit] = true;
    }
}

enum FeExceptionBit
{
    FeDivByZero = FE_DIVBYZERO,
    FeInexact = FE_INEXACT,
    FeInvalid = FE_INVALID,
    FeOverflow = FE_OVERFLOW,
    FeUnderflow = FE_UNDERFLOW,
    FeAllExceptions = FE_ALL_EXCEPT
};

enum FeRoundingMode
{
    FeRoundDown = FE_DOWNWARD,
    FeRoundNearest = FE_TONEAREST,
    FeRoundZero = FE_TOWARDZERO,
    FeRoundUpward = FE_UPWARD
};

enum VfpRoundingMode
{
    VfpRoundNearest = 0,
    VfpRoundUpward = 1,
    VfpRoundDown = 2,
    VfpRoundZero = 3,
    VfpRoundAway = 4
};

static inline float bitsToFp(uint64_t, float);
static inline double bitsToFp(uint64_t, double);
static inline uint32_t fpToBits(float);
static inline uint64_t fpToBits(double);

template <class fpType>
static inline bool
flushToZero(fpType &op)
{
    fpType junk = 0.0;
    if (std::fpclassify(op) == FP_SUBNORMAL) {
        uint64_t bitMask = ULL(0x1) << (sizeof(fpType) * 8 - 1);
        op = bitsToFp(fpToBits(op) & bitMask, junk);
        return true;
    }
    return false;
}

template <class fpType>
static inline bool
flushToZero(fpType &op1, fpType &op2)
{
    bool flush1 = flushToZero(op1);
    bool flush2 = flushToZero(op2);
    return flush1 || flush2;
}

template <class fpType>
static inline void
vfpFlushToZero(FPSCR &fpscr, fpType &op)
{
    if (fpscr.fz == 1 && flushToZero(op)) {
        fpscr.idc = 1;
    }
}

template <class fpType>
static inline void
vfpFlushToZero(FPSCR &fpscr, fpType &op1, fpType &op2)
{
    vfpFlushToZero(fpscr, op1);
    vfpFlushToZero(fpscr, op2);
}

static inline uint32_t
fpToBits(float fp)
{
    union
    {
        float fp;
        uint32_t bits;
    } val;
    val.fp = fp;
    return val.bits;
}

static inline uint64_t
fpToBits(double fp)
{
    union
    {
        double fp;
        uint64_t bits;
    } val;
    val.fp = fp;
    return val.bits;
}

static inline float
bitsToFp(uint64_t bits, float junk)
{
    union
    {
        float fp;
        uint32_t bits;
    } val;
    val.bits = bits;
    return val.fp;
}

static inline double
bitsToFp(uint64_t bits, double junk)
{
    union
    {
        double fp;
        uint64_t bits;
    } val;
    val.bits = bits;
    return val.fp;
}

template <class fpType>
static inline bool
isSnan(fpType val)
{
    const bool single = (sizeof(fpType) == sizeof(float));
    const uint64_t qnan =
        single ? 0x7fc00000 : ULL(0x7ff8000000000000);
    return std::isnan(val) && ((fpToBits(val) & qnan) != qnan);
}

typedef int VfpSavedState;

VfpSavedState prepFpState(uint32_t rMode);
void finishVfp(FPSCR &fpscr, VfpSavedState state, bool flush, FPSCR mask = FpscrExcMask);

template <class fpType>
fpType fixDest(FPSCR fpscr, fpType val, fpType op1);

template <class fpType>
fpType fixDest(FPSCR fpscr, fpType val, fpType op1, fpType op2);

template <class fpType>
fpType fixDivDest(FPSCR fpscr, fpType val, fpType op1, fpType op2);

float fixFpDFpSDest(FPSCR fpscr, double val);
double fixFpSFpDDest(FPSCR fpscr, float val);

uint16_t vcvtFpSFpH(FPSCR &fpscr, bool flush, bool defaultNan,
                    uint32_t rMode, bool ahp, float op);
uint16_t vcvtFpDFpH(FPSCR &fpscr, bool flush, bool defaultNan,
                    uint32_t rMode, bool ahp, double op);

float  vcvtFpHFpS(FPSCR &fpscr, bool defaultNan, bool ahp, uint16_t op);
double vcvtFpHFpD(FPSCR &fpscr, bool defaultNan, bool ahp, uint16_t op);

static inline double
makeDouble(uint32_t low, uint32_t high)
{
    double junk = 0.0;
    return bitsToFp((uint64_t)low | ((uint64_t)high << 32), junk);
}

static inline uint32_t
lowFromDouble(double val)
{
    return fpToBits(val);
}

static inline uint32_t
highFromDouble(double val)
{
    return fpToBits(val) >> 32;
}

static inline void
setFPExceptions(int exceptions) {
    feclearexcept(FeAllExceptions);
    feraiseexcept(exceptions);
}

template <typename T>
uint64_t
vfpFpToFixed(T val, bool isSigned, uint8_t width, uint8_t imm, bool
             useRmode = true, VfpRoundingMode roundMode = VfpRoundZero,
             bool aarch64 = false)
{
    int  rmode;
    bool roundAwayFix = false;

    if (!useRmode) {
        rmode = fegetround();
    } else {
        switch (roundMode)
        {
          case VfpRoundNearest:
            rmode = FeRoundNearest;
            break;
          case VfpRoundUpward:
            rmode = FeRoundUpward;
            break;
          case VfpRoundDown:
            rmode = FeRoundDown;
            break;
          case VfpRoundZero:
            rmode = FeRoundZero;
            break;
          case VfpRoundAway:
            // There is no equivalent rounding mode, use round down and we'll
            // fix it later
            rmode        = FeRoundDown;
            roundAwayFix = true;
            break;
          default:
            panic("Unsupported roundMode %d\n", roundMode);
        }
    }
    __asm__ __volatile__("" : "=m" (rmode) : "m" (rmode));
    fesetround(FeRoundNearest);
    val = val * pow(2.0, imm);
    __asm__ __volatile__("" : "=m" (val) : "m" (val));
    fesetround(rmode);
    feclearexcept(FeAllExceptions);
    __asm__ __volatile__("" : "=m" (val) : "m" (val));
    T origVal = val;
    val = rint(val);
    __asm__ __volatile__("" : "=m" (val) : "m" (val));

    int exceptions = fetestexcept(FeAllExceptions);

    int fpType = std::fpclassify(val);
    if (fpType == FP_SUBNORMAL || fpType == FP_NAN) {
        if (fpType == FP_NAN) {
            exceptions |= FeInvalid;
        }
        val = 0.0;
    } else if (origVal != val) {
        switch (rmode) {
          case FeRoundNearest:
            if (origVal - val > 0.5)
                val += 1.0;
            else if (val - origVal > 0.5)
                val -= 1.0;
            break;
          case FeRoundDown:
            if (roundAwayFix) {
                // The ordering on the subtraction looks a bit odd in that we
                // don't do the obvious origVal - val, instead we do
                // -(val - origVal). This is required to get the corruct bit
                // exact behaviour when very close to the 0.5 threshold.
                volatile T error = val;
                error -= origVal;
                error = -error;
                if ( (error >  0.5) ||
                    ((error == 0.5) && (val >= 0)) )
                    val += 1.0;
            } else {
                if (origVal < val)
                    val -= 1.0;
            }
            break;
          case FeRoundUpward:
            if (origVal > val)
                val += 1.0;
            break;
        }
        exceptions |= FeInexact;
    }

    __asm__ __volatile__("" : "=m" (val) : "m" (val));

    if (isSigned) {
        bool     outOfRange = false;
        int64_t  result     = (int64_t) val;
        uint64_t finalVal;

        if (!aarch64) {
            if (width == 16) {
                finalVal = (int16_t)val;
            } else if (width == 32) {
                finalVal =(int32_t)val;
            } else if (width == 64) {
                finalVal = result;
            } else {
                panic("Unsupported width %d\n", width);
            }

            // check if value is in range
            int64_t minVal = ~mask(width-1);
            if ((double)val < minVal) {
                outOfRange = true;
                finalVal = minVal;
            }
            int64_t maxVal = mask(width-1);
            if ((double)val > maxVal) {
                outOfRange = true;
                finalVal = maxVal;
            }
        } else {
            bool isNeg = val < 0;
            finalVal = result & mask(width);
            // If the result is supposed to be less than 64 bits check that the
            // upper bits that got thrown away are just sign extension bits
            if (width != 64) {
                outOfRange = ((uint64_t) result >> (width - 1)) !=
                             (isNeg ? mask(64-width+1) : 0);
            }
            // Check if the original floating point value doesn't matches the
            // integer version we are also out of range. So create a saturated
            // result.
            if (isNeg) {
                outOfRange |= val < result;
                if (outOfRange) {
                    finalVal = 1LL << (width-1);
                }
            } else {
                outOfRange |= val > result;
                if (outOfRange) {
                    finalVal = mask(width-1);
                }
            }
        }

        // Raise an exception if the value was out of range
        if (outOfRange) {
            exceptions |= FeInvalid;
            exceptions &= ~FeInexact;
        }
        setFPExceptions(exceptions);
        return finalVal;
    } else {
        if ((double)val < 0) {
            exceptions |= FeInvalid;
            exceptions &= ~FeInexact;
            setFPExceptions(exceptions);
            return 0;
        }

        uint64_t result = ((uint64_t) val) & mask(width);
        if (val > result) {
            exceptions |= FeInvalid;
            exceptions &= ~FeInexact;
            setFPExceptions(exceptions);
            return mask(width);
        }

        setFPExceptions(exceptions);
        return result;
    }
};


float vfpUFixedToFpS(bool flush, bool defaultNan,
        uint64_t val, uint8_t width, uint8_t imm);
float vfpSFixedToFpS(bool flush, bool defaultNan,
        int64_t val, uint8_t width, uint8_t imm);

double vfpUFixedToFpD(bool flush, bool defaultNan,
        uint64_t val, uint8_t width, uint8_t imm);
double vfpSFixedToFpD(bool flush, bool defaultNan,
        int64_t val, uint8_t width, uint8_t imm);

float fprSqrtEstimate(FPSCR &fpscr, float op);
uint32_t unsignedRSqrtEstimate(uint32_t op);

float fpRecipEstimate(FPSCR &fpscr, float op);
uint32_t unsignedRecipEstimate(uint32_t op);

class VfpMacroOp : public PredMacroOp
{
  public:
    static bool
    inScalarBank(IntRegIndex idx)
    {
        return (idx % 32) < 8;
    }

  protected:
    bool wide;

    VfpMacroOp(const char *mnem, ExtMachInst _machInst,
            OpClass __opClass, bool _wide) :
        PredMacroOp(mnem, _machInst, __opClass), wide(_wide)
    {}

    IntRegIndex addStride(IntRegIndex idx, unsigned stride);
    void nextIdxs(IntRegIndex &dest, IntRegIndex &op1, IntRegIndex &op2);
    void nextIdxs(IntRegIndex &dest, IntRegIndex &op1);
    void nextIdxs(IntRegIndex &dest);
};

template <typename T>
static inline T
fpAdd(T a, T b)
{
    return a + b;
};

template <typename T>
static inline T
fpSub(T a, T b)
{
    return a - b;
};

static inline float
fpAddS(float a, float b)
{
    return a + b;
}

static inline double
fpAddD(double a, double b)
{
    return a + b;
}

static inline float
fpSubS(float a, float b)
{
    return a - b;
}

static inline double
fpSubD(double a, double b)
{
    return a - b;
}

static inline float
fpDivS(float a, float b)
{
    return a / b;
}

static inline double
fpDivD(double a, double b)
{
    return a / b;
}

template <typename T>
static inline T
fpDiv(T a, T b)
{
    return a / b;
};

template <typename T>
static inline T
fpMulX(T a, T b)
{
    uint64_t opData;
    uint32_t sign1;
    uint32_t sign2;
    const bool single = (sizeof(T) == sizeof(float));
    if (single) {
        opData = (fpToBits(a));
        sign1 = opData>>31;
        opData = (fpToBits(b));
        sign2 = opData>>31;
    } else {
        opData = (fpToBits(a));
        sign1 = opData>>63;
        opData = (fpToBits(b));
        sign2 = opData>>63;
    }
    bool inf1 = (std::fpclassify(a) == FP_INFINITE);
    bool inf2 = (std::fpclassify(b) == FP_INFINITE);
    bool zero1 = (std::fpclassify(a) == FP_ZERO);
    bool zero2 = (std::fpclassify(b) == FP_ZERO);
    if ((inf1 && zero2) || (zero1 && inf2)) {
        if (sign1 ^ sign2)
            return (T)(-2.0);
        else
            return (T)(2.0);
    } else {
        return (a * b);
    }
};


template <typename T>
static inline T
fpMul(T a, T b)
{
    return a * b;
};

static inline float
fpMulS(float a, float b)
{
    return a * b;
}

static inline double
fpMulD(double a, double b)
{
    return a * b;
}

template <typename T>
static inline T
// @todo remove this when all calls to it have been replaced with the new fplib implementation
fpMulAdd(T op1, T op2, T addend)
{
    T result;

    if (sizeof(T) == sizeof(float))
        result = fmaf(op1, op2, addend);
    else
        result = fma(op1, op2, addend);

    // ARM doesn't generate signed nan's from this opperation, so fix up the result
    if (std::isnan(result) && !std::isnan(op1) &&
        !std::isnan(op2) && !std::isnan(addend))
    {
        uint64_t bitMask = ULL(0x1) << ((sizeof(T) * 8) - 1);
        result = bitsToFp(fpToBits(result) & ~bitMask, op1);
    }
    return result;
}

template <typename T>
static inline T
fpRIntX(T a, FPSCR &fpscr)
{
    T rVal;

    rVal = rint(a);
    if (rVal != a && !std::isnan(a))
        fpscr.ixc = 1;
    return (rVal);
};

template <typename T>
static inline T
fpMaxNum(T a, T b)
{
    const bool     single = (sizeof(T) == sizeof(float));
    const uint64_t qnan   = single ? 0x7fc00000 : ULL(0x7ff8000000000000);

    if (std::isnan(a))
        return ((fpToBits(a) & qnan) == qnan) ? b : a;
    if (std::isnan(b))
        return ((fpToBits(b) & qnan) == qnan) ? a : b;
    // Handle comparisons of +0 and -0.
    if (!std::signbit(a) && std::signbit(b))
        return a;
    return fmax(a, b);
};

template <typename T>
static inline T
fpMax(T a, T b)
{
    if (std::isnan(a))
        return a;
    if (std::isnan(b))
        return b;
    return fpMaxNum<T>(a, b);
};

template <typename T>
static inline T
fpMinNum(T a, T b)
{
    const bool     single = (sizeof(T) == sizeof(float));
    const uint64_t qnan   = single ? 0x7fc00000 : ULL(0x7ff8000000000000);

    if (std::isnan(a))
        return ((fpToBits(a) & qnan) == qnan) ? b : a;
    if (std::isnan(b))
        return ((fpToBits(b) & qnan) == qnan) ? a : b;
    // Handle comparisons of +0 and -0.
    if (std::signbit(a) && !std::signbit(b))
        return a;
    return fmin(a, b);
};

template <typename T>
static inline T
fpMin(T a, T b)
{
    if (std::isnan(a))
        return a;
    if (std::isnan(b))
        return b;
    return fpMinNum<T>(a, b);
};

template <typename T>
static inline T
fpRSqrts(T a, T b)
{
    int fpClassA = std::fpclassify(a);
    int fpClassB = std::fpclassify(b);
    T aXb;
    int fpClassAxB;

    if ((fpClassA == FP_ZERO && fpClassB == FP_INFINITE) ||
        (fpClassA == FP_INFINITE && fpClassB == FP_ZERO)) {
        return 1.5;
    }
    aXb = a*b;
    fpClassAxB = std::fpclassify(aXb);
    if (fpClassAxB == FP_SUBNORMAL) {
       feraiseexcept(FeUnderflow);
       return 1.5;
    }
    return (3.0 - (a * b)) / 2.0;
};

template <typename T>
static inline T
fpRecps(T a, T b)
{
    int fpClassA = std::fpclassify(a);
    int fpClassB = std::fpclassify(b);
    T aXb;
    int fpClassAxB;

    if ((fpClassA == FP_ZERO && fpClassB == FP_INFINITE) ||
        (fpClassA == FP_INFINITE && fpClassB == FP_ZERO)) {
        return 2.0;
    }
    aXb = a*b;
    fpClassAxB = std::fpclassify(aXb);
    if (fpClassAxB == FP_SUBNORMAL) {
       feraiseexcept(FeUnderflow);
       return 2.0;
    }
    return 2.0 - (a * b);
};


static inline float
fpRSqrtsS(float a, float b)
{
    int fpClassA = std::fpclassify(a);
    int fpClassB = std::fpclassify(b);
    float aXb;
    int fpClassAxB;

    if ((fpClassA == FP_ZERO && fpClassB == FP_INFINITE) ||
        (fpClassA == FP_INFINITE && fpClassB == FP_ZERO)) {
        return 1.5;
    }
    aXb = a*b;
    fpClassAxB = std::fpclassify(aXb);
    if (fpClassAxB == FP_SUBNORMAL) {
       feraiseexcept(FeUnderflow);
       return 1.5;
    }
    return (3.0 - (a * b)) / 2.0;
}

static inline float
fpRecpsS(float a, float b)
{
    int fpClassA = std::fpclassify(a);
    int fpClassB = std::fpclassify(b);
    float aXb;
    int fpClassAxB;

    if ((fpClassA == FP_ZERO && fpClassB == FP_INFINITE) ||
        (fpClassA == FP_INFINITE && fpClassB == FP_ZERO)) {
        return 2.0;
    }
    aXb = a*b;
    fpClassAxB = std::fpclassify(aXb);
    if (fpClassAxB == FP_SUBNORMAL) {
       feraiseexcept(FeUnderflow);
       return 2.0;
    }
    return 2.0 - (a * b);
}

template <typename T>
static inline T
roundNEven(T a) {
    T val;

    val = round(a);
    if (a - val == 0.5) {
        if ( (((int) a) & 1) == 0 ) val += 1.0;
    }
    else if (a - val == -0.5) {
        if ( (((int) a) & 1) == 0 ) val -= 1.0;
    }
    return val;
}



class FpOp : public PredOp
{
  protected:
    FpOp(const char *mnem, ExtMachInst _machInst, OpClass __opClass) :
        PredOp(mnem, _machInst, __opClass)
    {}

    virtual float
    doOp(float op1, float op2) const
    {
        panic("Unimplemented version of doOp called.\n");
    }

    virtual float
    doOp(float op1) const
    {
        panic("Unimplemented version of doOp called.\n");
    }

    virtual double
    doOp(double op1, double op2) const
    {
        panic("Unimplemented version of doOp called.\n");
    }

    virtual double
    doOp(double op1) const
    {
        panic("Unimplemented version of doOp called.\n");
    }

    double
    dbl(uint32_t low, uint32_t high) const
    {
        double junk = 0.0;
        return bitsToFp((uint64_t)low | ((uint64_t)high << 32), junk);
    }

    uint32_t
    dblLow(double val) const
    {
        return fpToBits(val);
    }

    uint32_t
    dblHi(double val) const
    {
        return fpToBits(val) >> 32;
    }

    template <class fpType>
    fpType
    processNans(FPSCR &fpscr, bool &done, bool defaultNan,
                fpType op1, fpType op2) const;

    template <class fpType>
    fpType
    ternaryOp(FPSCR &fpscr, fpType op1, fpType op2, fpType op3,
              fpType (*func)(fpType, fpType, fpType),
              bool flush, bool defaultNan, uint32_t rMode) const;

    template <class fpType>
    fpType
    binaryOp(FPSCR &fpscr, fpType op1, fpType op2,
            fpType (*func)(fpType, fpType),
            bool flush, bool defaultNan, uint32_t rMode) const;

    template <class fpType>
    fpType
    unaryOp(FPSCR &fpscr, fpType op1,
            fpType (*func)(fpType),
            bool flush, uint32_t rMode) const;

    void
    advancePC(PCState &pcState) const
    {
        if (flags[IsLastMicroop]) {
            pcState.uEnd();
        } else if (flags[IsMicroop]) {
            pcState.uAdvance();
        } else {
            pcState.advance();
        }
    }

    float
    fpSqrt (FPSCR fpscr,float x) const
    {

        return unaryOp(fpscr,x,sqrtf,fpscr.fz,fpscr.rMode);

    }

    double
    fpSqrt (FPSCR fpscr,double x) const
    {

        return unaryOp(fpscr,x,sqrt,fpscr.fz,fpscr.rMode);

    }
};

class FpCondCompRegOp : public FpOp
{
  protected:
    IntRegIndex op1, op2;
    ConditionCode condCode;
    uint8_t defCc;

    FpCondCompRegOp(const char *mnem, ExtMachInst _machInst,
                       OpClass __opClass, IntRegIndex _op1, IntRegIndex _op2,
                       ConditionCode _condCode, uint8_t _defCc) :
        FpOp(mnem, _machInst, __opClass),
        op1(_op1), op2(_op2), condCode(_condCode), defCc(_defCc)
    {}

    std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const;
};

class FpCondSelOp : public FpOp
{
  protected:
    IntRegIndex dest, op1, op2;
    ConditionCode condCode;

    FpCondSelOp(const char *mnem, ExtMachInst _machInst, OpClass __opClass,
                IntRegIndex _dest, IntRegIndex _op1, IntRegIndex _op2,
                ConditionCode _condCode) :
        FpOp(mnem, _machInst, __opClass),
        dest(_dest), op1(_op1), op2(_op2), condCode(_condCode)
    {}

    std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const;
};

class FpRegRegOp : public FpOp
{
  protected:
    IntRegIndex dest;
    IntRegIndex op1;

    FpRegRegOp(const char *mnem, ExtMachInst _machInst, OpClass __opClass,
               IntRegIndex _dest, IntRegIndex _op1,
               VfpMicroMode mode = VfpNotAMicroop) :
        FpOp(mnem, _machInst, __opClass), dest(_dest), op1(_op1)
    {
        setVfpMicroFlags(mode, flags);
    }

    std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const;
};

class FpRegImmOp : public FpOp
{
  protected:
    IntRegIndex dest;
    uint64_t imm;

    FpRegImmOp(const char *mnem, ExtMachInst _machInst, OpClass __opClass,
               IntRegIndex _dest, uint64_t _imm,
               VfpMicroMode mode = VfpNotAMicroop) :
        FpOp(mnem, _machInst, __opClass), dest(_dest), imm(_imm)
    {
        setVfpMicroFlags(mode, flags);
    }

    std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const;
};

class FpRegRegImmOp : public FpOp
{
  protected:
    IntRegIndex dest;
    IntRegIndex op1;
    uint64_t imm;

    FpRegRegImmOp(const char *mnem, ExtMachInst _machInst, OpClass __opClass,
                  IntRegIndex _dest, IntRegIndex _op1,
                  uint64_t _imm, VfpMicroMode mode = VfpNotAMicroop) :
        FpOp(mnem, _machInst, __opClass), dest(_dest), op1(_op1), imm(_imm)
    {
        setVfpMicroFlags(mode, flags);
    }

    std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const;
};

class FpRegRegRegOp : public FpOp
{
  protected:
    IntRegIndex dest;
    IntRegIndex op1;
    IntRegIndex op2;

    FpRegRegRegOp(const char *mnem, ExtMachInst _machInst, OpClass __opClass,
                  IntRegIndex _dest, IntRegIndex _op1, IntRegIndex _op2,
                  VfpMicroMode mode = VfpNotAMicroop) :
        FpOp(mnem, _machInst, __opClass), dest(_dest), op1(_op1), op2(_op2)
    {
        setVfpMicroFlags(mode, flags);
    }

    std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const;
};

class FpRegRegRegCondOp : public FpOp
{
  protected:
    IntRegIndex dest;
    IntRegIndex op1;
    IntRegIndex op2;
    ConditionCode cond;

    FpRegRegRegCondOp(const char *mnem, ExtMachInst _machInst,
                      OpClass __opClass, IntRegIndex _dest, IntRegIndex _op1,
                      IntRegIndex _op2, ConditionCode _cond,
                      VfpMicroMode mode = VfpNotAMicroop) :
        FpOp(mnem, _machInst, __opClass), dest(_dest), op1(_op1), op2(_op2),
        cond(_cond)
    {
        setVfpMicroFlags(mode, flags);
    }

    std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const;
};

class FpRegRegRegRegOp : public FpOp
{
  protected:
    IntRegIndex dest;
    IntRegIndex op1;
    IntRegIndex op2;
    IntRegIndex op3;

    FpRegRegRegRegOp(const char *mnem, ExtMachInst _machInst, OpClass __opClass,
                     IntRegIndex _dest, IntRegIndex _op1, IntRegIndex _op2,
                     IntRegIndex _op3, VfpMicroMode mode = VfpNotAMicroop) :
        FpOp(mnem, _machInst, __opClass), dest(_dest), op1(_op1), op2(_op2),
        op3(_op3)
    {
        setVfpMicroFlags(mode, flags);
    }

    std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const;
};

class FpRegRegRegImmOp : public FpOp
{
  protected:
    IntRegIndex dest;
    IntRegIndex op1;
    IntRegIndex op2;
    uint64_t imm;

    FpRegRegRegImmOp(const char *mnem, ExtMachInst _machInst,
                     OpClass __opClass, IntRegIndex _dest,
                     IntRegIndex _op1, IntRegIndex _op2,
                     uint64_t _imm, VfpMicroMode mode = VfpNotAMicroop) :
        FpOp(mnem, _machInst, __opClass),
        dest(_dest), op1(_op1), op2(_op2), imm(_imm)
    {
        setVfpMicroFlags(mode, flags);
    }

    std::string generateDisassembly(Addr pc, const SymbolTable *symtab) const;
};

}

#endif //__ARCH_ARM_INSTS_VFP_HH__