1 /* mpn_mod_1s_4p (ap, n, b, cps) 2 Divide (ap,,n) by b. Return the single-limb remainder. 3 Requires that d < B / 4. 4 5 Contributed to the GNU project by Torbjorn Granlund. 6 Based on a suggestion by Peter L. Montgomery. 7 8 THE FUNCTIONS IN THIS FILE ARE INTERNAL WITH MUTABLE INTERFACES. IT IS ONLY 9 SAFE TO REACH THEM THROUGH DOCUMENTED INTERFACES. IN FACT, IT IS ALMOST 10 GUARANTEED THAT THEY WILL CHANGE OR DISAPPEAR IN A FUTURE GNU MP RELEASE. 11 12 Copyright 2008, 2009, 2010 Free Software Foundation, Inc. 13 14 This file is part of the GNU MP Library. 15 16 The GNU MP Library is free software; you can redistribute it and/or modify 17 it under the terms of the GNU Lesser General Public License as published by 18 the Free Software Foundation; either version 3 of the License, or (at your 19 option) any later version. 20 21 The GNU MP Library is distributed in the hope that it will be useful, but 22 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY 23 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public 24 License for more details. 25 26 You should have received a copy of the GNU Lesser General Public License 27 along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. */ 28 29 #include "gmp.h" 30 #include "gmp-impl.h" 31 #include "longlong.h" 32 33 #include "mpn/sparc64/sparc64.h" 34 35 void 36 mpn_mod_1s_4p_cps (mp_limb_t cps[7], mp_limb_t b) 37 { 38 mp_limb_t bi; 39 mp_limb_t B1modb, B2modb, B3modb, B4modb, B5modb; 40 int cnt; 41 42 ASSERT (b <= (~(mp_limb_t) 0) / 4); 43 44 count_leading_zeros (cnt, b); 45 46 b <<= cnt; 47 invert_limb (bi, b); 48 49 B1modb = -b * ((bi >> (GMP_LIMB_BITS-cnt)) | (CNST_LIMB(1) << cnt)); 50 ASSERT (B1modb <= b); /* NB: not fully reduced mod b */ 51 udiv_rnnd_preinv (B2modb, B1modb, 0, b, bi); 52 udiv_rnnd_preinv (B3modb, B2modb, 0, b, bi); 53 udiv_rnnd_preinv (B4modb, B3modb, 0, b, bi); 54 udiv_rnnd_preinv (B5modb, B4modb, 0, b, bi); 55 56 cps[0] = bi; 57 cps[1] = cnt; 58 cps[2] = B1modb >> cnt; 59 cps[3] = B2modb >> cnt; 60 cps[4] = B3modb >> cnt; 61 cps[5] = B4modb >> cnt; 62 cps[6] = B5modb >> cnt; 63 64 #if WANT_ASSERT 65 { 66 int i; 67 b = cps[2]; 68 for (i = 3; i <= 6; i++) 69 { 70 b += cps[i]; 71 ASSERT (b >= cps[i]); 72 } 73 } 74 #endif 75 } 76 77 mp_limb_t 78 mpn_mod_1s_4p (mp_srcptr ap, mp_size_t n, mp_limb_t b, mp_limb_t cps[7]) 79 { 80 mp_limb_t rh, rl, bi, ph, pl, ch, cl, r; 81 mp_limb_t B1modb, B2modb, B3modb, B4modb, B5modb; 82 mp_size_t i; 83 int cnt; 84 85 ASSERT (n >= 1); 86 87 B1modb = cps[2]; 88 B2modb = cps[3]; 89 B3modb = cps[4]; 90 B4modb = cps[5]; 91 B5modb = cps[6]; 92 93 if ((b >> 32) == 0) 94 { 95 switch (n & 3) 96 { 97 case 0: 98 umul_ppmm_s (ph, pl, ap[n - 3], B1modb); 99 add_ssaaaa (ph, pl, ph, pl, 0, ap[n - 4]); 100 umul_ppmm_s (ch, cl, ap[n - 2], B2modb); 101 add_ssaaaa (ph, pl, ph, pl, ch, cl); 102 umul_ppmm_s (rh, rl, ap[n - 1], B3modb); 103 add_ssaaaa (rh, rl, rh, rl, ph, pl); 104 n -= 4; 105 break; 106 case 1: 107 rh = 0; 108 rl = ap[n - 1]; 109 n -= 1; 110 break; 111 case 2: 112 rh = ap[n - 1]; 113 rl = ap[n - 2]; 114 n -= 2; 115 break; 116 case 3: 117 umul_ppmm_s (ph, pl, ap[n - 2], B1modb); 118 add_ssaaaa (ph, pl, ph, pl, 0, ap[n - 3]); 119 umul_ppmm_s (rh, rl, ap[n - 1], B2modb); 120 add_ssaaaa (rh, rl, rh, rl, ph, pl); 121 n -= 3; 122 break; 123 } 124 125 for (i = n - 4; i >= 0; i -= 4) 126 { 127 /* rr = ap[i] < B 128 + ap[i+1] * (B mod b) <= (B-1)(b-1) 129 + ap[i+2] * (B^2 mod b) <= (B-1)(b-1) 130 + ap[i+3] * (B^3 mod b) <= (B-1)(b-1) 131 + LO(rr) * (B^4 mod b) <= (B-1)(b-1) 132 + HI(rr) * (B^5 mod b) <= (B-1)(b-1) 133 */ 134 umul_ppmm_s (ph, pl, ap[i + 1], B1modb); 135 add_ssaaaa (ph, pl, ph, pl, 0, ap[i + 0]); 136 137 umul_ppmm_s (ch, cl, ap[i + 2], B2modb); 138 add_ssaaaa (ph, pl, ph, pl, ch, cl); 139 140 umul_ppmm_s (ch, cl, ap[i + 3], B3modb); 141 add_ssaaaa (ph, pl, ph, pl, ch, cl); 142 143 umul_ppmm_s (ch, cl, rl, B4modb); 144 add_ssaaaa (ph, pl, ph, pl, ch, cl); 145 146 umul_ppmm_s (rh, rl, rh, B5modb); 147 add_ssaaaa (rh, rl, rh, rl, ph, pl); 148 } 149 150 umul_ppmm_s (rh, cl, rh, B1modb); 151 add_ssaaaa (rh, rl, rh, rl, 0, cl); 152 } 153 else 154 { 155 switch (n & 3) 156 { 157 case 0: 158 umul_ppmm (ph, pl, ap[n - 3], B1modb); 159 add_ssaaaa (ph, pl, ph, pl, 0, ap[n - 4]); 160 umul_ppmm (ch, cl, ap[n - 2], B2modb); 161 add_ssaaaa (ph, pl, ph, pl, ch, cl); 162 umul_ppmm (rh, rl, ap[n - 1], B3modb); 163 add_ssaaaa (rh, rl, rh, rl, ph, pl); 164 n -= 4; 165 break; 166 case 1: 167 rh = 0; 168 rl = ap[n - 1]; 169 n -= 1; 170 break; 171 case 2: 172 rh = ap[n - 1]; 173 rl = ap[n - 2]; 174 n -= 2; 175 break; 176 case 3: 177 umul_ppmm (ph, pl, ap[n - 2], B1modb); 178 add_ssaaaa (ph, pl, ph, pl, 0, ap[n - 3]); 179 umul_ppmm (rh, rl, ap[n - 1], B2modb); 180 add_ssaaaa (rh, rl, rh, rl, ph, pl); 181 n -= 3; 182 break; 183 } 184 185 for (i = n - 4; i >= 0; i -= 4) 186 { 187 /* rr = ap[i] < B 188 + ap[i+1] * (B mod b) <= (B-1)(b-1) 189 + ap[i+2] * (B^2 mod b) <= (B-1)(b-1) 190 + ap[i+3] * (B^3 mod b) <= (B-1)(b-1) 191 + LO(rr) * (B^4 mod b) <= (B-1)(b-1) 192 + HI(rr) * (B^5 mod b) <= (B-1)(b-1) 193 */ 194 umul_ppmm (ph, pl, ap[i + 1], B1modb); 195 add_ssaaaa (ph, pl, ph, pl, 0, ap[i + 0]); 196 197 umul_ppmm (ch, cl, ap[i + 2], B2modb); 198 add_ssaaaa (ph, pl, ph, pl, ch, cl); 199 200 umul_ppmm (ch, cl, ap[i + 3], B3modb); 201 add_ssaaaa (ph, pl, ph, pl, ch, cl); 202 203 umul_ppmm (ch, cl, rl, B4modb); 204 add_ssaaaa (ph, pl, ph, pl, ch, cl); 205 206 umul_ppmm (rh, rl, rh, B5modb); 207 add_ssaaaa (rh, rl, rh, rl, ph, pl); 208 } 209 210 umul_ppmm (rh, cl, rh, B1modb); 211 add_ssaaaa (rh, rl, rh, rl, 0, cl); 212 } 213 214 bi = cps[0]; 215 cnt = cps[1]; 216 217 r = (rh << cnt) | (rl >> (GMP_LIMB_BITS - cnt)); 218 udiv_rnnd_preinv (r, r, rl << cnt, b, bi); 219 220 return r >> cnt; 221 } 222