1 /* mpn_dc_divappr_q -- divide-and-conquer division, returning only approximate
2 quotient. The quotient retuened is either correct, or unity too large.
4 Contributed to the GNU project by Torbjörn Granlund.
6 THE FUNCTIONS IN THIS FILE ARE INTERNAL WITH A MUTABLE INTERFACE. IT IS
7 ONLY SAFE TO REACH THEM THROUGH DOCUMENTED INTERFACES. IN FACT, IT IS
8 ALMOST GUARANTEED THAT THEY WILL CHANGE OR DISAPPEAR IN A FUTURE GMP
11 Copyright 2006, 2007 Free Software Foundation, Inc.
13 This file is part of the GNU MP Library.
15 The GNU MP Library is free software; you can redistribute it and/or modify
16 it under the terms of the GNU Lesser General Public License as published by
17 the Free Software Foundation; either version 3 of the License, or (at your
18 option) any later version.
20 The GNU MP Library is distributed in the hope that it will be useful, but
21 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
22 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public
23 License for more details.
25 You should have received a copy of the GNU Lesser General Public License
26 along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. */
33 mpn_dc_divappr_q_n (mp_ptr qp, mp_ptr np, mp_srcptr dp, mp_size_t n,
34 mp_srcptr dip, mp_ptr tp)
39 lo = n >> 1; /* floor(n/2) */
40 hi = n - lo; /* ceil(n/2) */
42 if (BELOW_THRESHOLD (hi, DC_DIV_QR_THRESHOLD))
43 qh = mpn_sb_div_qr (qp + lo, np + 2 * lo, 2 * hi, dp + lo, hi, dip);
45 qh = mpn_dc_div_qr_n (qp + lo, np + 2 * lo, dp + lo, hi, dip, tp);
47 mpn_mul (tp, qp + lo, hi, dp, lo);
49 cy = mpn_sub_n (np + lo, np + lo, tp, n);
51 cy += mpn_sub_n (np + n, np + n, dp, lo);
55 qh -= mpn_sub_1 (qp + lo, qp + lo, hi, 1);
56 cy -= mpn_add_n (np + lo, np + lo, dp, n);
59 if (BELOW_THRESHOLD (lo, DC_DIVAPPR_Q_THRESHOLD))
60 ql = mpn_sb_divappr_q (qp, np + hi, 2 * lo, dp + hi, lo, dip);
62 ql = mpn_dc_divappr_q_n (qp, np + hi, dp + hi, lo, dip, tp);
64 if (UNLIKELY (ql != 0))
67 for (i = 0; i < lo; i++)
68 qp[i] = GMP_NUMB_MASK;
75 mpn_preinv_dc_divappr_q (mp_ptr qp,
76 mp_ptr np, mp_size_t nn,
77 mp_srcptr dp, mp_size_t dn,
81 mp_limb_t qh, cy, qsave;
87 tp = TMP_SALLOC_LIMBS (dn+1);
96 qn++; /* pretend we'll need an extra limb */
97 /* Reduce qn mod dn without division, optimizing small operations. */
102 qp -= qn; /* point at low limb of next quotient block */
103 np -= qn; /* point in the middle of partial remainder */
105 /* Perform the typically smaller block first. */
106 if (BELOW_THRESHOLD (qn, DC_DIV_QR_THRESHOLD))
107 qh = mpn_sb_div_qr (qp, np - qn, 2 * qn, dp - qn, qn, dip);
109 qh = mpn_dc_div_qr_n (qp, np - qn, dp - qn, qn, dip, tp);
114 mpn_mul (tp, qp, qn, dp - dn, dn - qn);
116 mpn_mul (tp, dp - dn, dn - qn, qp, qn);
118 cy = mpn_sub_n (np - dn, np - dn, tp, dn);
120 cy += mpn_sub_n (np - dn + qn, np - dn + qn, dp - dn, dn - qn);
124 qh -= mpn_sub_1 (qp, qp, qn, 1);
125 cy -= mpn_add_n (np - dn, np - dn, dp - dn, dn);
129 qn = nn - dn - qn + 1;
134 mpn_dc_div_qr_n (qp, np - dn, dp - dn, dn, dip, tp);
138 /* Since we pretended we'd need an extra quotient limb before, we now
139 have made sure the code above left just dn-1=qn quotient limbs to
140 develop. Develop that plus a guard limb. */
145 mpn_dc_divappr_q_n (qp, np - dn, dp - dn, dn, dip, tp);
146 MPN_COPY_INCR (qp, qp + 1, qn);
153 qh = mpn_cmp (np - dn, dp - dn, dn) >= 0;
155 mpn_sub_n (np - dn, np - dn, dp - dn, dn);
160 qp -= qn; /* point at low limb of next quotient block */
161 np -= qn; /* point in the middle of partial remainder */
163 if (BELOW_THRESHOLD (qn, DC_DIVAPPR_Q_THRESHOLD))
164 /* Full precision. Optimal? */
165 qh = mpn_sb_divappr_q (qp, np - dn, nn, dp - dn, dn, dip);
168 /* Put quotient in tp, use qp as temporary, since qp lacks a limb. */
169 qh = mpn_dc_divappr_q_n (tp, np - qn - 2, dp - (qn + 1), qn + 1, dip, qp);
170 MPN_COPY (qp, tp + 1, qn);
179 mpn_dc_divappr_q (mp_ptr qp, mp_ptr np, mp_size_t nn, mp_srcptr dp, mp_size_t dn)
182 mp_limb_t xp[2], dip[2];
186 cy = mpn_add_1 (xp, dp + dn - 2, 2, 1);
191 mp_limb_t scratch[10]; /* FIXME */
192 mpn_invert (dip, xp, 2, scratch);
195 return mpn_preinv_dc_divappr_q (qp, np, nn, dp, dn, dip);