xref: /netbsd-src/external/lgpl3/mpfr/dist/src/ubf.c (revision ec6772edaf0cdcb5f52a48f4aca5e33a8fb8ecfd)
1 /* Functions to work with unbounded floats (limited low-level interface).
2 
3 Copyright 2016-2023 Free Software Foundation, Inc.
4 Contributed by the AriC and Caramba projects, INRIA.
5 
6 This file is part of the GNU MPFR Library.
7 
8 The GNU MPFR Library is free software; you can redistribute it and/or modify
9 it under the terms of the GNU Lesser General Public License as published by
10 the Free Software Foundation; either version 3 of the License, or (at your
11 option) any later version.
12 
13 The GNU MPFR Library is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
15 or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
16 License for more details.
17 
18 You should have received a copy of the GNU Lesser General Public License
19 along with the GNU MPFR Library; see the file COPYING.LESSER.  If not, see
20 https://www.gnu.org/licenses/ or write to the Free Software Foundation, Inc.,
21 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA. */
22 
23 #define MPFR_NEED_LONGLONG_H
24 #include "mpfr-impl.h"
25 
26 /* Note: In MPFR math functions, even if UBF code is not called first,
27    we may still need to handle special values NaN and infinities here.
28    Indeed, for MAXR * MAXR + (-inf), even though this is a special case,
29    the computation will also generate an overflow due to MAXR * MAXR,
30    so that UBF code will be called anyway (except if special cases are
31    detected and handled separately, but for polynomial, this should not
32    be needed). */
33 
34 /* Get the exponent of a regular MPFR number or UBF as a mpz_t, which is
35    initialized by this function. The flags are not changed. */
36 static void
mpfr_init_get_zexp(mpz_ptr ez,mpfr_srcptr x)37 mpfr_init_get_zexp (mpz_ptr ez, mpfr_srcptr x)
38 {
39   mpz_init (ez);
40 
41   if (MPFR_IS_UBF (x))
42     mpz_set (ez, MPFR_ZEXP (x));
43   else
44     {
45       mpfr_exp_t ex = MPFR_GET_EXP (x);
46 
47 #if _MPFR_EXP_FORMAT <= 3
48       /* mpfr_exp_t fits in a long */
49       mpz_set_si (ez, ex);
50 #else
51       mp_limb_t e_limb[MPFR_EXP_LIMB_SIZE];
52       mpfr_t e;
53       int inex;
54       MPFR_SAVE_EXPO_DECL (expo);
55 
56       MPFR_TMP_INIT1 (e_limb, e, sizeof (mpfr_exp_t) * CHAR_BIT);
57       MPFR_SAVE_EXPO_MARK (expo);
58       MPFR_DBGRES (inex = mpfr_set_exp_t (e, ex, MPFR_RNDN));
59       MPFR_ASSERTD (inex == 0);
60       MPFR_DBGRES (inex = mpfr_get_z (ez, e, MPFR_RNDN));
61       MPFR_ASSERTD (inex == 0);
62       MPFR_SAVE_EXPO_FREE (expo);
63 #endif
64     }
65 }
66 
67 /* Exact product. The number a is assumed to have enough allocated memory,
68    where the trailing bits are regarded as being part of the input numbers
69    (no reallocation is attempted and no check is performed as MPFR_TMP_INIT
70    could have been used). The arguments b and c may actually be UBF numbers
71    (mpfr_srcptr can be seen a bit like void *, but is stronger).
72    This function does not change the flags, except in case of NaN. */
73 void
mpfr_ubf_mul_exact(mpfr_ubf_ptr a,mpfr_srcptr b,mpfr_srcptr c)74 mpfr_ubf_mul_exact (mpfr_ubf_ptr a, mpfr_srcptr b, mpfr_srcptr c)
75 {
76   MPFR_LOG_FUNC
77     (("b[%Pd]=%.*Rg c[%Pd]=%.*Rg",
78       mpfr_get_prec (b), mpfr_log_prec, b,
79       mpfr_get_prec (c), mpfr_log_prec, c),
80      ("a[%Pd]=%.*Rg",
81       mpfr_get_prec ((mpfr_ptr) a), mpfr_log_prec, a));
82 
83   MPFR_ASSERTD ((mpfr_ptr) a != b);
84   MPFR_ASSERTD ((mpfr_ptr) a != c);
85   MPFR_SIGN (a) = MPFR_MULT_SIGN (MPFR_SIGN (b), MPFR_SIGN (c));
86 
87   if (MPFR_ARE_SINGULAR (b, c))
88     {
89       if (MPFR_IS_NAN (b) || MPFR_IS_NAN (c))
90         MPFR_SET_NAN (a);
91       else if (MPFR_IS_INF (b))
92         {
93           if (MPFR_NOTZERO (c))
94             MPFR_SET_INF (a);
95           else
96             MPFR_SET_NAN (a);
97         }
98       else if (MPFR_IS_INF (c))
99         {
100           if (!MPFR_IS_ZERO (b))
101             MPFR_SET_INF (a);
102           else
103             MPFR_SET_NAN (a);
104         }
105       else
106         {
107           MPFR_ASSERTD (MPFR_IS_ZERO(b) || MPFR_IS_ZERO(c));
108           MPFR_SET_ZERO (a);
109         }
110     }
111   else
112     {
113       mpfr_exp_t e;
114       mp_size_t bn, cn;
115       mpfr_limb_ptr ap;
116       mp_limb_t u, v;
117       int m;
118 
119       /* Note about the code below: For the choice of the precision of
120        * the result a, one could choose PREC(b) + PREC(c), instead of
121        * taking whole limbs into account, but in most cases where one
122        * would gain one limb, one would need to copy the significand
123        * instead of a no-op (see the mul.c code).
124        * But in the case MPFR_LIMB_MSB (u) == 0, if the result fits in
125        * an-1 limbs, one could actually do
126        *   mpn_rshift (ap, ap, k, GMP_NUMB_BITS - 1)
127        * instead of
128        *   mpn_lshift (ap, ap, k, 1)
129        * to gain one limb (and reduce the precision), replacing a shift
130        * by another one. Would this be interesting?
131        */
132 
133       bn = MPFR_LIMB_SIZE (b);
134       cn = MPFR_LIMB_SIZE (c);
135 
136       ap = MPFR_MANT (a);
137 
138       if (bn == 1 && cn == 1)
139         {
140           umul_ppmm (ap[1], ap[0], MPFR_MANT(b)[0], MPFR_MANT(c)[0]);
141           if (ap[1] & MPFR_LIMB_HIGHBIT)
142             m = 0;
143           else
144             {
145               ap[1] = (ap[1] << 1) | (ap[0] >> (GMP_NUMB_BITS - 1));
146               ap[0] = ap[0] << 1;
147               m = 1;
148             }
149         }
150       else
151         {
152           if (b == c)
153             {
154               mpn_sqr (ap, MPFR_MANT (b), bn);
155               u = ap[2 * bn - 1];
156             }
157           else
158             u = (bn >= cn) ?
159               mpn_mul (ap, MPFR_MANT (b), bn, MPFR_MANT (c), cn) :
160               mpn_mul (ap, MPFR_MANT (c), cn, MPFR_MANT (b), bn);
161           if (MPFR_LIMB_MSB (u) == 0)
162             {
163               m = 1;
164               MPFR_DBGRES (v = mpn_lshift (ap, ap, bn + cn, 1));
165               MPFR_ASSERTD (v == 0);
166             }
167           else
168             m = 0;
169         }
170 
171       if (! MPFR_IS_UBF (b) && ! MPFR_IS_UBF (c) &&
172           (e = MPFR_GET_EXP (b) + MPFR_GET_EXP (c) - m,
173            MPFR_EXP_IN_RANGE (e)))
174         {
175           MPFR_SET_EXP (a, e);
176         }
177       else
178         {
179           mpz_t be, ce;
180 
181           mpz_init (MPFR_ZEXP (a));
182 
183           /* This may involve copies of mpz_t, but exponents should not be
184              very large integers anyway. */
185           mpfr_init_get_zexp (be, b);
186           mpfr_init_get_zexp (ce, c);
187           mpz_add (MPFR_ZEXP (a), be, ce);
188           mpz_clear (be);
189           mpz_clear (ce);
190           mpz_sub_ui (MPFR_ZEXP (a), MPFR_ZEXP (a), m);
191           MPFR_SET_UBF (a);
192         }
193     }
194 }
195 
196 /* Compare the exponents of two numbers, which can be either MPFR numbers
197    or UBF numbers. If both numbers can be MPFR numbers, it is better to
198    use the MPFR_UBF_EXP_LESS_P wrapper macro, which is optimized for this
199    common case. */
200 int
mpfr_ubf_exp_less_p(mpfr_srcptr x,mpfr_srcptr y)201 mpfr_ubf_exp_less_p (mpfr_srcptr x, mpfr_srcptr y)
202 {
203   mpz_t xe, ye;
204   int c;
205 
206   mpfr_init_get_zexp (xe, x);
207   mpfr_init_get_zexp (ye, y);
208   c = mpz_cmp (xe, ye) < 0;
209   mpz_clear (xe);
210   mpz_clear (ye);
211   return c;
212 }
213 
214 /* Convert an mpz_t to an mpfr_exp_t, saturated to
215    the interval [MPFR_EXP_MIN,MPFR_EXP_MAX]. */
216 mpfr_exp_t
mpfr_ubf_zexp2exp(mpz_ptr ez)217 mpfr_ubf_zexp2exp (mpz_ptr ez)
218 {
219   mp_size_t n;
220   mpfr_eexp_t e;
221   mpfr_t d;
222   int inex;
223   MPFR_SAVE_EXPO_DECL (expo);
224 
225   n = ABSIZ (ez); /* limb size of ez */
226   if (n == 0)
227     return 0;
228 
229   MPFR_SAVE_EXPO_MARK (expo);
230   mpfr_init2 (d, n * GMP_NUMB_BITS);
231   MPFR_DBGRES (inex = mpfr_set_z (d, ez, MPFR_RNDN));
232   MPFR_ASSERTD (inex == 0);
233   e = mpfr_get_exp_t (d, MPFR_RNDZ);
234   mpfr_clear (d);
235   MPFR_SAVE_EXPO_FREE (expo);
236   if (MPFR_UNLIKELY (e < MPFR_EXP_MIN))
237     return MPFR_EXP_MIN;
238   if (MPFR_UNLIKELY (e > MPFR_EXP_MAX))
239     return MPFR_EXP_MAX;
240   return e;
241 }
242 
243 /* Return the difference of the exponents of x and y, saturated to
244    the interval [MPFR_EXP_MIN,MPFR_EXP_MAX]. */
245 mpfr_exp_t
mpfr_ubf_diff_exp(mpfr_srcptr x,mpfr_srcptr y)246 mpfr_ubf_diff_exp (mpfr_srcptr x, mpfr_srcptr y)
247 {
248   mpz_t xe, ye;
249   mpfr_exp_t e;
250 
251   mpfr_init_get_zexp (xe, x);
252   mpfr_init_get_zexp (ye, y);
253   mpz_sub (xe, xe, ye);
254   mpz_clear (ye);
255   e = mpfr_ubf_zexp2exp (xe);
256   mpz_clear (xe);
257   return e;
258 }
259