aboutsummaryrefslogtreecommitdiffstats
path: root/arch/powerpc/math-emu/op-common.h
blob: afb82b6498ce8b84ff440e99cec9c8e2a4c0b984 (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
#define _FP_DECL(wc, X)			\
  _FP_I_TYPE X##_c, X##_s, X##_e;	\
  _FP_FRAC_DECL_##wc(X)

/*
 * Finish truely unpacking a native fp value by classifying the kind
 * of fp value and normalizing both the exponent and the fraction.
 */

#define _FP_UNPACK_CANONICAL(fs, wc, X)					\
do {									\
  switch (X##_e)							\
  {									\
  default:								\
    _FP_FRAC_HIGH_##wc(X) |= _FP_IMPLBIT_##fs;				\
    _FP_FRAC_SLL_##wc(X, _FP_WORKBITS);					\
    X##_e -= _FP_EXPBIAS_##fs;						\
    X##_c = FP_CLS_NORMAL;						\
    break;								\
									\
  case 0:								\
    if (_FP_FRAC_ZEROP_##wc(X))						\
      X##_c = FP_CLS_ZERO;						\
    else								\
      {									\
	/* a denormalized number */					\
	_FP_I_TYPE _shift;						\
	_FP_FRAC_CLZ_##wc(_shift, X);					\
	_shift -= _FP_FRACXBITS_##fs;					\
	_FP_FRAC_SLL_##wc(X, (_shift+_FP_WORKBITS));			\
	X##_e -= _FP_EXPBIAS_##fs - 1 + _shift;				\
	X##_c = FP_CLS_NORMAL;						\
      }									\
    break;								\
									\
  case _FP_EXPMAX_##fs:							\
    if (_FP_FRAC_ZEROP_##wc(X))						\
      X##_c = FP_CLS_INF;						\
    else								\
      /* we don't differentiate between signaling and quiet nans */	\
      X##_c = FP_CLS_NAN;						\
    break;								\
  }									\
} while (0)


/*
 * Before packing the bits back into the native fp result, take care
 * of such mundane things as rounding and overflow.  Also, for some
 * kinds of fp values, the original parts may not have been fully
 * extracted -- but that is ok, we can regenerate them now.
 */

#define _FP_PACK_CANONICAL(fs, wc, X)				\
({int __ret = 0;						\
  switch (X##_c)						\
  {								\
  case FP_CLS_NORMAL:						\
    X##_e += _FP_EXPBIAS_##fs;					\
    if (X##_e > 0)						\
      {								\
	__ret |= _FP_ROUND(wc, X);				\
	if (_FP_FRAC_OVERP_##wc(fs, X))				\
	  {							\
	    _FP_FRAC_SRL_##wc(X, (_FP_WORKBITS+1));		\
	    X##_e++;						\
	  }							\
	else							\
	  _FP_FRAC_SRL_##wc(X, _FP_WORKBITS);			\
	if (X##_e >= _FP_EXPMAX_##fs)				\
	  {							\
	    /* overflow to infinity */				\
	    X##_e = _FP_EXPMAX_##fs;				\
	    _FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc);		\
            __ret |= EFLAG_OVERFLOW;				\
	  }							\
      }								\
    else							\
      {								\
	/* we've got a denormalized number */			\
	X##_e = -X##_e + 1;					\
	if (X##_e <= _FP_WFRACBITS_##fs)			\
	  {							\
	    _FP_FRAC_SRS_##wc(X, X##_e, _FP_WFRACBITS_##fs);	\
	    _FP_FRAC_SLL_##wc(X, 1);				\
	    if (_FP_FRAC_OVERP_##wc(fs, X))			\
	      {							\
	        X##_e = 1;					\
	        _FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc);	\
	      }							\
	    else						\
	      {							\
		X##_e = 0;					\
		_FP_FRAC_SRL_##wc(X, _FP_WORKBITS+1);		\
                __ret |= EFLAG_UNDERFLOW;			\
	      }							\
	  }							\
	else							\
	  {							\
	    /* underflow to zero */				\
	    X##_e = 0;						\
	    _FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc);		\
            __ret |= EFLAG_UNDERFLOW;				\
	  }							\
      }								\
    break;							\
								\
  case FP_CLS_ZERO:						\
    X##_e = 0;							\
    _FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc);			\
    break;							\
								\
  case FP_CLS_INF:						\
    X##_e = _FP_EXPMAX_##fs;					\
    _FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc);			\
    break;							\
								\
  case FP_CLS_NAN:						\
    X##_e = _FP_EXPMAX_##fs;					\
    if (!_FP_KEEPNANFRACP)					\
      {								\
	_FP_FRAC_SET_##wc(X, _FP_NANFRAC_##fs);			\
	X##_s = 0;						\
      }								\
    else							\
      _FP_FRAC_HIGH_##wc(X) |= _FP_QNANBIT_##fs;		\
    break;							\
  }								\
  __ret;							\
})


/*
 * Main addition routine.  The input values should be cooked.
 */

#define _FP_ADD(fs, wc, R, X, Y)					     \
do {									     \
  switch (_FP_CLS_COMBINE(X##_c, Y##_c))				     \
  {									     \
  case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NORMAL):			     \
    {									     \
      /* shift the smaller number so that its exponent matches the larger */ \
      _FP_I_TYPE diff = X##_e - Y##_e;					     \
									     \
      if (diff < 0)							     \
	{								     \
	  diff = -diff;							     \
	  if (diff <= _FP_WFRACBITS_##fs)				     \
	    _FP_FRAC_SRS_##wc(X, diff, _FP_WFRACBITS_##fs);		     \
	  else if (!_FP_FRAC_ZEROP_##wc(X))				     \
	    _FP_FRAC_SET_##wc(X, _FP_MINFRAC_##wc);			     \
	  else								     \
	    _FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc);			     \
	  R##_e = Y##_e;						     \
	}								     \
      else								     \
	{								     \
	  if (diff > 0)							     \
	    {								     \
	      if (diff <= _FP_WFRACBITS_##fs)				     \
	        _FP_FRAC_SRS_##wc(Y, diff, _FP_WFRACBITS_##fs);		     \
	      else if (!_FP_FRAC_ZEROP_##wc(Y))				     \
	        _FP_FRAC_SET_##wc(Y, _FP_MINFRAC_##wc);			     \
	      else							     \
	        _FP_FRAC_SET_##wc(Y, _FP_ZEROFRAC_##wc);		     \
	    }								     \
	  R##_e = X##_e;						     \
	}								     \
									     \
      R##_c = FP_CLS_NORMAL;						     \
									     \
      if (X##_s == Y##_s)						     \
	{								     \
	  R##_s = X##_s;						     \
	  _FP_FRAC_ADD_##wc(R, X, Y);					     \
	  if (_FP_FRAC_OVERP_##wc(fs, R))				     \
	    {								     \
	      _FP_FRAC_SRS_##wc(R, 1, _FP_WFRACBITS_##fs);		     \
	      R##_e++;							     \
	    }								     \
	}								     \
      else								     \
	{								     \
	  R##_s = X##_s;						     \
	  _FP_FRAC_SUB_##wc(R, X, Y);					     \
	  if (_FP_FRAC_ZEROP_##wc(R))					     \
	    {								     \
	      /* return an exact zero */				     \
	      if (FP_ROUNDMODE == FP_RND_MINF)				     \
		R##_s |= Y##_s;						     \
	      else							     \
		R##_s &= Y##_s;						     \
	      R##_c = FP_CLS_ZERO;					     \
	    }								     \
	  else								     \
	    {								     \
	      if (_FP_FRAC_NEGP_##wc(R))				     \
		{							     \
		  _FP_FRAC_SUB_##wc(R, Y, X);				     \
		  R##_s = Y##_s;					     \
		}							     \
									     \
	      /* renormalize after subtraction */			     \
	      _FP_FRAC_CLZ_##wc(diff, R);				     \
	      diff -= _FP_WFRACXBITS_##fs;				     \
	      if (diff)							     \
		{							     \
		  R##_e -= diff;					     \
		  _FP_FRAC_SLL_##wc(R, diff);				     \
		}							     \
	    }								     \
	}								     \
      break;								     \
    }									     \
									     \
  case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NAN):				     \
    _FP_CHOOSENAN(fs, wc, R, X, Y);					     \
    break;								     \
									     \
  case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_ZERO):			     \
    R##_e = X##_e;							     \
  case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NORMAL):			     \
  case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_INF):				     \
  case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_ZERO):				     \
    _FP_FRAC_COPY_##wc(R, X);						     \
    R##_s = X##_s;							     \
    R##_c = X##_c;							     \
    break;								     \
									     \
  case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NORMAL):			     \
    R##_e = Y##_e;							     \
  case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NAN):			     \
  case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NAN):				     \
  case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NAN):				     \
    _FP_FRAC_COPY_##wc(R, Y);						     \
    R##_s = Y##_s;							     \
    R##_c = Y##_c;							     \
    break;								     \
									     \
  case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_INF):				     \
    if (X##_s != Y##_s)							     \
      {									     \
	/* +INF + -INF => NAN */					     \
	_FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs);				     \
	R##_s = X##_s ^ Y##_s;						     \
	R##_c = FP_CLS_NAN;						     \
	break;								     \
      }									     \
    /* FALLTHRU */							     \
									     \
  case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NORMAL):			     \
  case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_ZERO):				     \
    R##_s = X##_s;							     \
    R##_c = FP_CLS_INF;							     \
    break;								     \
									     \
  case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_INF):			     \
  case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_INF):				     \
    R##_s = Y##_s;							     \
    R##_c = FP_CLS_INF;							     \
    break;								     \
									     \
  case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_ZERO):			     \
    /* make sure the sign is correct */					     \
    if (FP_ROUNDMODE == FP_RND_MINF)					     \
      R##_s = X##_s | Y##_s;						     \
    else								     \
      R##_s = X##_s & Y##_s;						     \
    R##_c = FP_CLS_ZERO;						     \
    break;								     \
									     \
  default:								     \
    abort();								     \
  }									     \
} while (0)


/*
 * Main negation routine.  FIXME -- when we care about setting exception
 * bits reliably, this will not do.  We should examine all of the fp classes.
 */

#define _FP_NEG(fs, wc, R, X)		\
  do {					\
    _FP_FRAC_COPY_##wc(R, X);		\
    R##_c = X##_c;			\
    R##_e = X##_e;			\
    R##_s = 1 ^ X##_s;			\
  } while (0)


/*
 * Main multiplication routine.  The input values should be cooked.
 */

#define _FP_MUL(fs, wc, R, X, Y)			\
do {							\
  R##_s = X##_s ^ Y##_s;				\
  switch (_FP_CLS_COMBINE(X##_c, Y##_c))		\
  {							\
  case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NORMAL):	\
    R##_c = FP_CLS_NORMAL;				\
    R##_e = X##_e + Y##_e + 1;				\
							\
    _FP_MUL_MEAT_##fs(R,X,Y);				\
							\
    if (_FP_FRAC_OVERP_##wc(fs, R))			\
      _FP_FRAC_SRS_##wc(R, 1, _FP_WFRACBITS_##fs);	\
    else						\
      R##_e--;						\
    break;						\
							\
  case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NAN):		\
    _FP_CHOOSENAN(fs, wc, R, X, Y);			\
    break;						\
							\
  case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NORMAL):	\
  case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_INF):		\
  case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_ZERO):		\
    R##_s = X##_s;					\
							\
  case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_INF):		\
  case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NORMAL):	\
  case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NORMAL):	\
  case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_ZERO):	\
    _FP_FRAC_COPY_##wc(R, X);				\
    R##_c = X##_c;					\
    break;						\
							\
  case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NAN):	\
  case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NAN):		\
  case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NAN):		\
    R##_s = Y##_s;					\
							\
  case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_INF):	\
  case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_ZERO):	\
    _FP_FRAC_COPY_##wc(R, Y);				\
    R##_c = Y##_c;					\
    break;						\
							\
  case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_ZERO):		\
  case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_INF):		\
    R##_c = FP_CLS_NAN;					\
    _FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs);		\
    break;						\
							\
  default:						\
    abort();						\
  }							\
} while (0)


/*
 * Main division routine.  The input values should be cooked.
 */

#define _FP_DIV(fs, wc, R, X, Y)			\
do {							\
  R##_s = X##_s ^ Y##_s;				\
  switch (_FP_CLS_COMBINE(X##_c, Y##_c))		\
  {							\
  case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NORMAL):	\
    R##_c = FP_CLS_NORMAL;				\
    R##_e = X##_e - Y##_e;				\
							\
    _FP_DIV_MEAT_##fs(R,X,Y);				\
    break;						\
							\
  case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NAN):		\
    _FP_CHOOSENAN(fs, wc, R, X, Y);			\
    break;						\
							\
  case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NORMAL):	\
  case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_INF):		\
  case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_ZERO):		\
    R##_s = X##_s;					\
    _FP_FRAC_COPY_##wc(R, X);				\
    R##_c = X##_c;					\
    break;						\
							\
  case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NAN):	\
  case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NAN):		\
  case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NAN):		\
    R##_s = Y##_s;					\
    _FP_FRAC_COPY_##wc(R, Y);				\
    R##_c = Y##_c;					\
    break;						\
							\
  case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_INF):	\
  case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_INF):		\
  case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NORMAL):	\
    R##_c = FP_CLS_ZERO;				\
    break;						\
							\
  case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_ZERO):	\
  case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_ZERO):		\
  case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NORMAL):	\
    R##_c = FP_CLS_INF;					\
    break;						\
							\
  case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_INF):		\
  case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_ZERO):	\
    R##_c = FP_CLS_NAN;					\
    _FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs);		\
    break;						\
							\
  default:						\
    abort();						\
  }							\
} while (0)


/*
 * Main differential comparison routine.  The inputs should be raw not
 * cooked.  The return is -1,0,1 for normal values, 2 otherwise.
 */

#define _FP_CMP(fs, wc, ret, X, Y, un)					\
  do {									\
    /* NANs are unordered */						\
    if ((X##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(X))		\
	|| (Y##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(Y)))	\
      {									\
	ret = un;							\
      }									\
    else								\
      {									\
        int __x_zero = (!X##_e && _FP_FRAC_ZEROP_##wc(X)) ? 1 : 0;	\
        int __y_zero = (!Y##_e && _FP_FRAC_ZEROP_##wc(Y)) ? 1 : 0;	\
									\
	if (__x_zero && __y_zero)					\
	  ret = 0;							\
	else if (__x_zero)						\
	  ret = Y##_s ? 1 : -1;						\
	else if (__y_zero)						\
	  ret = X##_s ? -1 : 1;						\
	else if (X##_s != Y##_s)					\
	  ret = X##_s ? -1 : 1;						\
	else if (X##_e > Y##_e)						\
	  ret = X##_s ? -1 : 1;						\
	else if (X##_e < Y##_e)						\
	  ret = X##_s ? 1 : -1;						\
	else if (_FP_FRAC_GT_##wc(X, Y))				\
	  ret = X##_s ? -1 : 1;						\
	else if (_FP_FRAC_GT_##wc(Y, X))				\
	  ret = X##_s ? 1 : -1;						\
	else								\
	  ret = 0;							\
      }									\
  } while (0)


/* Simplification for strict equality.  */

#define _FP_CMP_EQ(fs, wc, ret, X, Y)					  \
  do {									  \
    /* NANs are unordered */						  \
    if ((X##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(X))		  \
	|| (Y##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(Y)))	  \
      {									  \
	ret = 1;							  \
      }									  \
    else								  \
      {									  \
	ret = !(X##_e == Y##_e						  \
		&& _FP_FRAC_EQ_##wc(X, Y)				  \
		&& (X##_s == Y##_s || !X##_e && _FP_FRAC_ZEROP_##wc(X))); \
      }									  \
  } while (0)

/*
 * Main square root routine.  The input value should be cooked.
 */

#define _FP_SQRT(fs, wc, R, X)						\
do {									\
    _FP_FRAC_DECL_##wc(T); _FP_FRAC_DECL_##wc(S);			\
    _FP_W_TYPE q;							\
    switch (X##_c)							\
    {									\
    case FP_CLS_NAN:							\
    	R##_s = 0;							\
    	R##_c = FP_CLS_NAN;						\
    	_FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc);			\
    	break;								\
    case FP_CLS_INF:							\
    	if (X##_s)							\
    	  {								\
    	    R##_s = 0;							\
	    R##_c = FP_CLS_NAN; /* sNAN */				\
    	  }								\
    	else								\
    	  {								\
    	    R##_s = 0;							\
    	    R##_c = FP_CLS_INF; /* sqrt(+inf) = +inf */			\
    	  }								\
    	break;								\
    case FP_CLS_ZERO:							\
	R##_s = X##_s;							\
    	R##_c = FP_CLS_ZERO; /* sqrt(+-0) = +-0 */			\
	break;								\
    case FP_CLS_NORMAL:							\
    	R##_s = 0;							\
        if (X##_s)							\
          {								\
	    R##_c = FP_CLS_NAN; /* sNAN */				\
	    break;							\
          }								\
    	R##_c = FP_CLS_NORMAL;						\
        if (X##_e & 1)							\
          _FP_FRAC_SLL_##wc(X, 1);					\
        R##_e = X##_e >> 1;						\
        _FP_FRAC_SET_##wc(S, _FP_ZEROFRAC_##wc);			\
        _FP_FRAC_SET_##wc(R, _FP_ZEROFRAC_##wc);			\
        q = _FP_OVERFLOW_##fs;						\
        _FP_FRAC_SLL_##wc(X, 1);					\
        _FP_SQRT_MEAT_##wc(R, S, T, X, q);				\
        _FP_FRAC_SRL_##wc(R, 1);					\
    }									\
  } while (0)

/*
 * Convert from FP to integer
 */

/* "When a NaN, infinity, large positive argument >= 2147483648.0, or
 * large negative argument <= -2147483649.0 is converted to an integer,
 * the invalid_current bit...should be set and fp_exception_IEEE_754 should
 * be raised. If the floating point invalid trap is disabled, no trap occurs
 * and a numerical result is generated: if the sign bit of the operand
 * is 0, the result is 2147483647; if the sign bit of the operand is 1,
 * the result is -2147483648."
 * Similarly for conversion to extended ints, except that the boundaries
 * are >= 2^63, <= -(2^63 + 1), and the results are 2^63 + 1 for s=0 and
 * -2^63 for s=1.
 * -- SPARC Architecture Manual V9, Appendix B, which specifies how
 * SPARCs resolve implementation dependencies in the IEEE-754 spec.
 * I don't believe that the code below follows this. I'm not even sure
 * it's right!
 * It doesn't cope with needing to convert to an n bit integer when there
 * is no n bit integer type. Fortunately gcc provides long long so this
 * isn't a problem for sparc32.
 * I have, however, fixed its NaN handling to conform as above.
 *         -- PMM 02/1998
 * NB: rsigned is not 'is r declared signed?' but 'should the value stored
 * in r be signed or unsigned?'. r is always(?) declared unsigned.
 * Comments below are mine, BTW -- PMM
 */
#define _FP_TO_INT(fs, wc, r, X, rsize, rsigned)			\
  do {									\
    switch (X##_c)							\
      {									\
      case FP_CLS_NORMAL:						\
	if (X##_e < 0)							\
	  {								\
	  /* case FP_CLS_NAN: see above! */				\
	  case FP_CLS_ZERO:						\
	    r = 0;							\
	  }								\
	else if (X##_e >= rsize - (rsigned != 0))			\
	  {	/* overflow */						\
	  case FP_CLS_NAN:                                              \
          case FP_CLS_INF:						\
	    if (rsigned)						\
	      {								\
		r = 1;							\
		r <<= rsize - 1;					\
		r -= 1 - X##_s;						\
	      }								\
	    else							\
	      {								\
		r = 0;							\
		if (!X##_s)						\
		  r = ~r;						\
	      }								\
	  }								\
	else								\
	  {								\
	    if (_FP_W_TYPE_SIZE*wc < rsize)				\
	      {								\
		_FP_FRAC_ASSEMBLE_##wc(r, X, rsize);			\
		r <<= X##_e - _FP_WFRACBITS_##fs;			\
	      }								\
	    else							\
	      {								\
		if (X##_e >= _FP_WFRACBITS_##fs)			\
		  _FP_FRAC_SLL_##wc(X, (X##_e - _FP_WFRACBITS_##fs + 1));\
		else							\
		  _FP_FRAC_SRL_##wc(X, (_FP_WFRACBITS_##fs - X##_e - 1));\
		_FP_FRAC_ASSEMBLE_##wc(r, X, rsize);			\
	      }								\
	    if (rsigned && X##_s)					\
	      r = -r;							\
	  }								\
	break;								\
      }									\
  } while (0)

#define _FP_FROM_INT(fs, wc, X, r, rsize, rtype)			\
  do {									\
    if (r)								\
      {									\
	X##_c = FP_CLS_NORMAL;						\
									\
	if ((X##_s = (r < 0)))						\
	  r = -r;							\
	/* Note that `r' is now considered unsigned, so we don't have	\
	   to worry about the single signed overflow case.  */		\
									\
	if (rsize <= _FP_W_TYPE_SIZE)					\
	  __FP_CLZ(X##_e, r);						\
	else								\
	  __FP_CLZ_2(X##_e, (_FP_W_TYPE)(r >> _FP_W_TYPE_SIZE), 	\
		     (_FP_W_TYPE)r);					\
	if (rsize < _FP_W_TYPE_SIZE)					\
		X##_e -= (_FP_W_TYPE_SIZE - rsize);			\
	X##_e = rsize - X##_e - 1;					\
									\
	if (_FP_FRACBITS_##fs < rsize && _FP_WFRACBITS_##fs < X##_e)	\
	  __FP_FRAC_SRS_1(r, (X##_e - _FP_WFRACBITS_##fs), rsize);	\
	r &= ~((_FP_W_TYPE)1 << X##_e);					\
	_FP_FRAC_DISASSEMBLE_##wc(X, ((unsigned rtype)r), rsize);	\
	_FP_FRAC_SLL_##wc(X, (_FP_WFRACBITS_##fs - X##_e - 1));		\
      }									\
    else								\
      {									\
	X##_c = FP_CLS_ZERO, X##_s = 0;					\
      }									\
  } while (0)


#define FP_CONV(dfs,sfs,dwc,swc,D,S)			\
  do {							\
    _FP_FRAC_CONV_##dwc##_##swc(dfs, sfs, D, S);	\
    D##_e = S##_e;					\
    D##_c = S##_c;					\
    D##_s = S##_s;					\
  } while (0)

/*
 * Helper primitives.
 */

/* Count leading zeros in a word.  */

#ifndef __FP_CLZ
#if _FP_W_TYPE_SIZE < 64
/* this is just to shut the compiler up about shifts > word length -- PMM 02/1998 */
#define __FP_CLZ(r, x)				\
  do {						\
    _FP_W_TYPE _t = (x);			\
    r = _FP_W_TYPE_SIZE - 1;			\
    if (_t > 0xffff) r -= 16;			\
    if (_t > 0xffff) _t >>= 16;			\
    if (_t > 0xff) r -= 8;			\
    if (_t > 0xff) _t >>= 8;			\
    if (_t & 0xf0) r -= 4;			\
    if (_t & 0xf0) _t >>= 4;			\
    if (_t & 0xc) r -= 2;			\
    if (_t & 0xc) _t >>= 2;			\
    if (_t & 0x2) r -= 1;			\
  } while (0)
#else /* not _FP_W_TYPE_SIZE < 64 */
#define __FP_CLZ(r, x)				\
  do {						\
    _FP_W_TYPE _t = (x);			\
    r = _FP_W_TYPE_SIZE - 1;			\
    if (_t > 0xffffffff) r -= 32;		\
    if (_t > 0xffffffff) _t >>= 32;		\
    if (_t > 0xffff) r -= 16;			\
    if (_t > 0xffff) _t >>= 16;			\
    if (_t > 0xff) r -= 8;			\
    if (_t > 0xff) _t >>= 8;			\
    if (_t & 0xf0) r -= 4;			\
    if (_t & 0xf0) _t >>= 4;			\
    if (_t & 0xc) r -= 2;			\
    if (_t & 0xc) _t >>= 2;			\
    if (_t & 0x2) r -= 1;			\
  } while (0)
#endif /* not _FP_W_TYPE_SIZE < 64 */
#endif /* ndef __FP_CLZ */

#define _FP_DIV_HELP_imm(q, r, n, d)		\
  do {						\
    q = n / d, r = n % d;			\
  } while (0)