1 /* mpn_dcpi1_bdiv_q -- divide-and-conquer Hensel division with precomputed 2 inverse, returning quotient. 3 4 Contributed to the GNU project by Niels Möller and Torbjorn Granlund. 5 6 THE FUNCTIONS IN THIS FILE ARE INTERNAL WITH MUTABLE INTERFACES. IT IS ONLY 7 SAFE TO REACH THEM THROUGH DOCUMENTED INTERFACES. IN FACT, IT IS ALMOST 8 GUARANTEED THAT THEY WILL CHANGE OR DISAPPEAR IN A FUTURE GMP RELEASE. 9 10 Copyright 2006, 2007, 2009-2011, 2017 Free Software Foundation, Inc. 11 12 This file is part of the GNU MP Library. 13 14 The GNU MP Library is free software; you can redistribute it and/or modify 15 it under the terms of either: 16 17 * the GNU Lesser General Public License as published by the Free 18 Software Foundation; either version 3 of the License, or (at your 19 option) any later version. 20 21 or 22 23 * the GNU General Public License as published by the Free Software 24 Foundation; either version 2 of the License, or (at your option) any 25 later version. 26 27 or both in parallel, as here. 28 29 The GNU MP Library is distributed in the hope that it will be useful, but 30 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY 31 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 32 for more details. 33 34 You should have received copies of the GNU General Public License and the 35 GNU Lesser General Public License along with the GNU MP Library. If not, 36 see https://www.gnu.org/licenses/. */ 37 38 #include "gmp-impl.h" 39 40 41 static mp_size_t 42 mpn_dcpi1_bdiv_q_n_itch (mp_size_t n) 43 { 44 /* NOTE: Depends on mullo_n and mpn_dcpi1_bdiv_qr_n interface */ 45 return n; 46 } 47 48 /* Computes Q = - N / D mod B^n, destroys N. 49 50 N = {np,n} 51 D = {dp,n} 52 */ 53 54 static void 55 mpn_dcpi1_bdiv_q_n (mp_ptr qp, 56 mp_ptr np, mp_srcptr dp, mp_size_t n, 57 mp_limb_t dinv, mp_ptr tp) 58 { 59 while (ABOVE_THRESHOLD (n, DC_BDIV_Q_THRESHOLD)) 60 { 61 mp_size_t lo, hi; 62 mp_limb_t cy; 63 64 lo = n >> 1; /* floor(n/2) */ 65 hi = n - lo; /* ceil(n/2) */ 66 67 cy = mpn_dcpi1_bdiv_qr_n (qp, np, dp, lo, dinv, tp); 68 69 mpn_mullo_n (tp, qp, dp + hi, lo); 70 mpn_add_n (np + hi, np + hi, tp, lo); 71 72 if (lo < hi) 73 { 74 cy += mpn_addmul_1 (np + lo, qp, lo, dp[lo]); 75 np[n - 1] += cy; 76 } 77 qp += lo; 78 np += lo; 79 n -= lo; 80 } 81 mpn_sbpi1_bdiv_q (qp, np, n, dp, n, dinv); 82 } 83 84 /* Computes Q = - N / D mod B^nn, destroys N. 85 86 N = {np,nn} 87 D = {dp,dn} 88 */ 89 90 void 91 mpn_dcpi1_bdiv_q (mp_ptr qp, 92 mp_ptr np, mp_size_t nn, 93 mp_srcptr dp, mp_size_t dn, 94 mp_limb_t dinv) 95 { 96 mp_size_t qn; 97 mp_limb_t cy; 98 mp_ptr tp; 99 TMP_DECL; 100 101 TMP_MARK; 102 103 ASSERT (dn >= 2); 104 ASSERT (nn - dn >= 0); 105 ASSERT (dp[0] & 1); 106 107 tp = TMP_SALLOC_LIMBS (dn); 108 109 qn = nn; 110 111 if (qn > dn) 112 { 113 /* Reduce qn mod dn in a super-efficient manner. */ 114 do 115 qn -= dn; 116 while (qn > dn); 117 118 /* Perform the typically smaller block first. */ 119 if (BELOW_THRESHOLD (qn, DC_BDIV_QR_THRESHOLD)) 120 cy = mpn_sbpi1_bdiv_qr (qp, np, 2 * qn, dp, qn, dinv); 121 else 122 cy = mpn_dcpi1_bdiv_qr_n (qp, np, dp, qn, dinv, tp); 123 124 if (qn != dn) 125 { 126 if (qn > dn - qn) 127 mpn_mul (tp, qp, qn, dp + qn, dn - qn); 128 else 129 mpn_mul (tp, dp + qn, dn - qn, qp, qn); 130 mpn_incr_u (tp + qn, cy); 131 132 mpn_add (np + qn, np + qn, nn - qn, tp, dn); 133 cy = 0; 134 } 135 136 np += qn; 137 qp += qn; 138 139 qn = nn - qn; 140 while (qn > dn) 141 { 142 mpn_add_1 (np + dn, np + dn, qn - dn, cy); 143 cy = mpn_dcpi1_bdiv_qr_n (qp, np, dp, dn, dinv, tp); 144 qp += dn; 145 np += dn; 146 qn -= dn; 147 } 148 mpn_dcpi1_bdiv_q_n (qp, np, dp, dn, dinv, tp); 149 } 150 else 151 { 152 if (BELOW_THRESHOLD (qn, DC_BDIV_Q_THRESHOLD)) 153 mpn_sbpi1_bdiv_q (qp, np, qn, dp, qn, dinv); 154 else 155 mpn_dcpi1_bdiv_q_n (qp, np, dp, qn, dinv, tp); 156 } 157 158 TMP_FREE; 159 } 160