summaryrefslogtreecommitdiff
path: root/add.c
blob: d0b4d2a3bdfcd9ab5b9a6f365f17f8a549458771 (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
/* (c) PolKA project, Inria Lorraine */

/* written by Paul Zimmermann, February 1999 */

/* to do: - destination may be identical to operands
*/


#include <stdio.h>
#include "gmp.h"
#include "gmp-impl.h"
#include "mpfr.h"

extern mpfr_sub1(mpfr_ptr, mpfr_srcptr, mpfr_srcptr, unsigned char, int);

/* #define DEBUG2 */

#ifdef DEBUG2
mp_limb_t *ap0, *ap1;
#define ck(x, dx, i) /* mp_limb_t *x; int dx; */ \
{ \
  if (x<ap0 || x+dx-1>ap1) { \
    printf("error %d\n",i); \
    printf("%1.20e,%d,%1.20e,%d,%d,%d\n",mpfr_get_d(b),PREC(b),mpfr_get_d(c), \
	   PREC(c),PREC(a),rnd_mode); \
    exit(1); } \
}
#else
#define ck(x, dx, i) {}
#endif

/* signs of b and c are supposed equal,
   diff_exp is the difference between the exponents of b and c,
   which is supposed >= 0 */
void mpfr_add1(a, b, c, rnd_mode, diff_exp) 
mpfr_ptr a; mpfr_srcptr b, c; unsigned char rnd_mode; int diff_exp;
{
  mp_limb_t *ap, *bp, *cp, cc, c2, c3=0; unsigned int an,bn,cn; int sh,dif,k;

#ifdef DEBUG2
  printf("b=  "); mpfr_print_raw(b); putchar('\n');
  printf(" =  %1.20e\n",mpfr_get_d(b));
  printf("c=  "); for (k=0;k<diff_exp;k++) putchar(' '); mpfr_print_raw(c);
  putchar('\n');
#endif
  ap = MANT(a);
  bp = MANT(b);
  an = (PREC(a)-1)/mp_bits_per_limb+1; /* number of significant limbs of a */
#ifdef DEBUG2
ap0=ap; ap1=ap+an-1;
#endif
  sh = an*mp_bits_per_limb-PREC(a); /* non-significant bits in low limb */
  bn = (PREC(b)-1)/mp_bits_per_limb+1; /* number of significant limbs of b */
  EXP(a) = EXP(b);
  if (SIGN(a)!=SIGN(b)) CHANGE_SIGN(a);
  /* case 1: diff_exp>=prec(a), i.e. c only affects the last bit
     through rounding */
  dif = PREC(a)-diff_exp;
#ifdef DEBUG2
  if (bn & ((mp_limb_t)1<<(mp_bits_per_limb-1))==0) {
    printf("Error in mpfr_add: b not msb-normalized\n"); exit(1);
  }
printf("an=%u prec(a)=%d bn=%u prec(b)=%d prec(c)=%d diff_exp=%u dif=%d\n",
       an,PREC(a),bn,PREC(b),PREC(c),diff_exp,dif);
#endif
  if (dif<=0) { /* diff_exp>=PREC(a): c does not overlap with a */
    /* either PREC(b)<=PREC(a), and we can copy the mantissa of b directly 
       into that of a, or PREC(b)>PREC(a) and we have to round b+c */
    if (PREC(b)<=PREC(a)) {
ck(ap+an-bn,bn,1);
      MPN_COPY(ap+(an-bn), bp, bn);
      /* fill low significant limbs with zero */
      for (bp=ap;bn<an;bn++) { ck(bp,1,2); *bp++=0; }
      /* now take c into account */
      if (rnd_mode==GMP_RNDN) { /* to nearest */
	/* if diff_exp > PREC(a), no change */
	if (diff_exp==PREC(a)) {
	  /* if c is not zero, then as it is normalized, we have to add
	     one to the lsb of a if c>1/2, or c=1/2 and lsb(a)=1 (round to
	     even) */
	  cp=MANT(c);
	  if (NOTZERO(c)) { /* c is not zero */
	    /* check whether mant(c)=1/2 or not */
	    cc = *cp - (1<<(mp_bits_per_limb-1));
	    if (cc==0) {
	      bp = cp+(PREC(c)-1)/mp_bits_per_limb;
	      while (cp<bp && cc==0) cc = *++cp;
	    }
ck(ap+an-1, 1, 3);
	    if (cc || (ap[an-1] & 1<<sh)) goto add_one_ulp;
	      /* mant(c) != 1/2 or mant(c) = 1/2: add 1 iff lsb(a)=1 */
	  }
	}
      }
      else if ((ISNONNEG(b) && rnd_mode==GMP_RNDU) || 
	       (ISNEG(b) && rnd_mode==GMP_RNDD)) {
	/* round up */
	if (NOTZERO(c)) goto add_one_ulp;
      }
      /* in the other cases (round to zero, or up/down with sign -/+),
         nothing to do */
    }
    else { /* PREC(b)>PREC(a) : we have to round b+c */
      k=bn-an;
      /* first copy the 'an' most significant limbs of b to a */
ck(ap, an, 4);
      MPN_COPY(ap, bp+k, an);
      if (rnd_mode==GMP_RNDN) { /* to nearest */
	/* first check whether the truncated bits from b are 1/2*lsb(a) */
	if (sh) {
ck(ap, 1, 5);
	  cc = *ap & ((1<<sh)-1);
	  *ap &= ~cc; /* truncate last bits */
	  cc -= 1<<(sh-1);
	}
	else /* no bit to truncate */
	  cc = bp[--k] - (1<<(mp_bits_per_limb-1));
	if ((long)cc>0) goto add_one_ulp; /* trunc(b)>1/2*lsb(a) -> round up */
	else if (cc==0) {
	  while (k>1 && cc==0) cc=bp[--k];
	  /* now if the truncated part of b = 1/2*lsb(a), check whether c=0 */
ck(ap, 1, 6);
	  if (NOTZERO(c) || (*ap & (1<<sh))) goto add_one_ulp;
	  /* if trunc(b)+c is exactly 1/2*lsb(a) : round to even lsb */
	}
	/* if cc<0 : trunc(b) < 1/2*lsb(a) -> round down, i.e. do nothing */
      }
      else if ((ISNONNEG(b) && rnd_mode==GMP_RNDU) || 
	       (ISNEG(b) && rnd_mode==GMP_RNDD)) {
	/* first check whether trunc(b)+c is zero or not */
	if (sh) {
ck(ap, 1, 7);
	  cc = *ap & ((1<<sh)-1); *ap &= ~cc; /* truncate last bits */
	}
	else cc = bp[--k] - (1<<(mp_bits_per_limb-1));
	while (cc==0 && k>1) cc=bp[--k];
	if (cc || NOTZERO(c)) goto add_one_ulp;
      }
      /* in the other cases (round to zero, or up/down with sign -/+),
         nothing to do, since b and c don't overlap, there can't be any
	 carry */
    }
  }
  else { /* diff_exp < PREC(a) : c overlaps with a by dif bits */
    /* first copy upper part of c into a (after shift) */
    unsigned char overlap;
    k = (dif-1)/mp_bits_per_limb + 1; /* only the highest k limbs from c
					 have to be considered */
    cp = MANT(c);
    cn = (PREC(c)-1)/mp_bits_per_limb + 1;
ck(ap+k, an-k, 8);
    MPN_ZERO(ap+k, an-k); /* do it now otherwise ap[k] may be destroyed
			     in case dif<0 */
#ifdef DEBUG2
    printf("k=%d cn=%d\n",k,cn);
#endif
    if (dif<=PREC(c)) { /* c has to be truncated */
      dif = dif % mp_bits_per_limb;
      dif = (dif) ? mp_bits_per_limb-dif-sh : -sh;
      /* we have to shift by dif bits to the right */
      if (dif>0) { ck(ap, k, 9); mpn_rshift(ap, cp+(cn-k), k, dif); }
      else if (dif<0) {
ck(ap, k+1, 10);
	ap[k] = mpn_lshift(ap, cp+(cn-k), k, -dif);
	/* put the non-significant bits in low limb for further rounding */
	ap[0] += cp[cn-k-1]>>(mp_bits_per_limb+dif);
      }
      else { ck(ap, k, 11); MPN_COPY(ap, cp+(cn-k), k); }
      overlap=1;
    }
    else { /* c is not truncated, but we have to fill low limbs with 0 */
      k = diff_exp/mp_bits_per_limb;
      overlap = diff_exp%mp_bits_per_limb;
      /* warning: a shift of zero bit is not allowed */
      if (overlap) { 
	cc=mpn_rshift(ap+(an-k-cn), cp, cn, overlap); 
	if (an-k-cn>0) ap[an-k-cn-1]=cc;
      }
      else { ck(ap+(an-k-cn), cn, 13); MPN_COPY(ap+(an-k-cn), cp, cn); }
      overlap=0;
    }
    /* here overlap=1 iff ulp(c)<ulp(a) */
    /* then put high limbs to zero */
    /* now add 'an' upper limbs of b in place */
    if (PREC(b)<=PREC(a)) {
      overlap += 2;
ck(ap+(an-bn), bn, 14);
      cc = mpn_add_n(ap+(an-bn), ap+(an-bn), bp, bn);
    }
    else /* PREC(b) > PREC(a): we have to truncate b */
{ ck(ap, an, 15);      cc = mpn_add_n(ap, ap, bp+(bn-an), an); }
    if (cc) { /* shift one bit to the right */
ck(ap, an, 16);
      c3 = (ap[0]&1) && (PREC(a)%mp_bits_per_limb==0);
      mpn_rshift(ap, ap, an, 1);
      ap[an-1] += (mp_limb_t)1<<(mp_bits_per_limb-1);
      EXP(a)++;
    }
    /* remains to do the rounding */
    if (rnd_mode==GMP_RNDN) { /* to nearest */
      int kc;
      /* four cases: overlap =
         (0) PREC(b) > PREC(a) and diff_exp+PREC(c) <= PREC(a)
         (1) PREC(b) > PREC(a) and diff_exp+PREC(c) > PREC(a)
         (2) PREC(b) <= PREC(a) and diff_exp+PREC(c) <= PREC(a)
         (3)  PREC(b) <= PREC(a) and diff_exp+PREC(c) > PREC(a) */
#ifdef DEBUG2
printf("overlap=%d\n",overlap);
#endif
      switch (overlap)
	{
        case 1: /* both b and c to round */
	  kc = cn-k; /* remains kc limbs from c */
	  k = bn-an; /* remains k limbs from b */
	  /* truncate last bits and store the difference with 1/2*ulp in cc */
ck(ap, 1, 17);
	  cc = *ap & ((1<<sh)-1);
	  *ap &= ~cc; /* truncate last bits */
	  cc -= 1<<(sh-1);
	  while ((cc==0 || cc==-1) && k!=0 && kc!=0) {
	    kc--;
	    cc += mpn_add_1(&c2, bp+(--k), 1,(cp[kc+1]<<(mp_bits_per_limb-dif))
			    +(cp[kc]>>dif));
	    if (cc==0 || cc==-1) cc=c2;
	  }
	  if ((long)cc>0) goto add_one_ulp;
	  else if ((long)cc<-1) return; /* the carry can be at most 1 */
	  else if (kc==0) goto round_b;
	  /* else round c: go through */
	case 3: /* only c to round */
	  bp=cp; k=cn-k; goto to_nearest;
	case 0: /* only b to round */
        round_b:
	  k=bn-an; dif=0; goto to_nearest;
        /* otherwise the result is exact: nothing to do */
	}
    }
    else if ((ISNONNEG(b) && rnd_mode==GMP_RNDU) || 
	     (ISNEG(b) && rnd_mode==GMP_RNDD)) {
ck(ap, 1, 18);
      cc = *ap & ((1<<sh)-1);
      *ap &= ~cc; /* truncate last bits */
      if (cc || c3) goto add_one_ulp; /* will happen most of the time */
      else { /* same four cases too */
	int kc = cn-k; /* remains kc limbs from c */
	switch (overlap)
	{
        case 1: /* both b and c to round */
	  k = bn-an; /* remains k limbs from b */
	  while (cc==0 && k!=0 && kc!=0) {
	    kc--;
	    cc = mpn_add_1(&c2, bp+(--k), 1,(cp[kc+1]<<(mp_bits_per_limb-dif))
			   + (cp[kc]>>dif));
	  }
	  if (cc) goto add_one_ulp;
	  else if (kc==0) goto round_b2;
	  /* else round c: go through */
	case 3: /* only c to round */
	  while (kc) if (cp[--kc]) goto add_one_ulp;
	  /* if dif>0 : remains to check last dif bits from c */
	  if (dif>0 && (cp[0]<<(mp_bits_per_limb-dif))) goto add_one_ulp;
	  break;
	case 0: /* only b to round */
        round_b2:
	  k=bn-an;
	  while (k) if (bp[--k]) goto add_one_ulp;
        /* otherwise the result is exact: nothing to do */
	}
      }
    }
    /* else nothing to do: round towards zero, i.e. truncate last sh bits */
    else {
ck(ap, 1, 19);
      *ap &= ~((1<<sh)-1);
    }
  }
  goto end_of_add;
    
  to_nearest: /* 0 <= sh < mp_bits_per_limb : number of bits of a to truncate
                 bp[k] : last significant limb from b */
        if (sh) {
ck(ap, 1, 20);
	  cc = *ap & ((1<<sh)-1);
	  *ap &= ~cc; /* truncate last bits */
	  c2 = 1<<(sh-1);
	}
	else /* no bit to truncate */
	  { cc = bp[--k]; c2 = 1<<(mp_bits_per_limb-1); }
#ifdef DEBUG2
printf("cc=%lu c2=%lu k=%u\n",cc,c2,k);
#endif
	if (cc>c2) goto add_one_ulp; /* trunc(b)>1/2*lsb(a) -> round up */
	else if (cc==c2) {
	  cc=0; while (k && cc==0) cc=bp[--k];
#ifdef DEBUG2
printf("cc=%lu\n",cc);
#endif
	  /* special case of rouding c shifted to the right */
	  if (cc==0 && dif>0) cc=cp[0]<<(mp_bits_per_limb-dif);
	  /* now if the truncated part of b = 1/2*lsb(a), check whether c=0 */
ck(ap, 1, 21);
	  if (cc || (*ap & (1<<sh))) goto add_one_ulp;
	}
        goto end_of_add;

  add_one_ulp: /* add one unit in last place to a */
ck(ap, an, 22);
    cc = mpn_add_1(ap, ap, an, 1<<sh);
    if (cc) { printf("carry(3) in mpfr_add\n"); exit(1); }

 end_of_add:
#ifdef DEBUG2
printf("b+c="); mpfr_print_raw(a); putchar('\n');
#endif
  return;
}

void mpfr_add(a, b, c, rnd_mode) 
mpfr_ptr a; mpfr_srcptr b, c; unsigned char rnd_mode;
{
  int diff_exp;

  diff_exp = EXP(b)-EXP(c);
  if (SIGN(b) != SIGN(c)) { /* signs differ, it's a subtraction */
    if (diff_exp<0) {
      mpfr_sub1(a, c, b, rnd_mode, -diff_exp);
    }
    else if (diff_exp>0) mpfr_sub1(a, b, c, rnd_mode, diff_exp);
    else { /* diff_exp=0 */
      CHANGE_SIGN(c);
      diff_exp = mpfr_cmp(b,c);
      CHANGE_SIGN(c);
      /* if b>0 and diff_exp>0 or b<0 and diff_exp<0: abs(b) > abs(c) */
      if (diff_exp*SIGN(b)>=0) mpfr_sub1(a, b, c, rnd_mode, 0);
      else {
	mpfr_sub1(a, c, b, rnd_mode, 0);
      }
    }
  }
  else /* signs are equal, it's an addition */
    if (diff_exp<0) mpfr_add1(a, c, b, rnd_mode, -diff_exp);
    else mpfr_add1(a, b, c, rnd_mode, diff_exp);
}