1 /* mpn_toom_eval_pm2exp -- Evaluate a polynomial in +2^k and -2^k
2
3 Contributed to the GNU project by Niels Möller
4
5 THE FUNCTION IN THIS FILE IS INTERNAL WITH A MUTABLE INTERFACE. IT IS ONLY
6 SAFE TO REACH IT THROUGH DOCUMENTED INTERFACES. IN FACT, IT IS ALMOST
7 GUARANTEED THAT IT WILL CHANGE OR DISAPPEAR IN A FUTURE GNU MP RELEASE.
8
9 Copyright 2009 Free Software Foundation, Inc.
10
11 This file is part of the GNU MP Library.
12
13 The GNU MP Library is free software; you can redistribute it and/or modify
14 it under the terms of either:
15
16 * the GNU Lesser General Public License as published by the Free
17 Software Foundation; either version 3 of the License, or (at your
18 option) any later version.
19
20 or
21
22 * the GNU General Public License as published by the Free Software
23 Foundation; either version 2 of the License, or (at your option) any
24 later version.
25
26 or both in parallel, as here.
27
28 The GNU MP Library is distributed in the hope that it will be useful, but
29 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
30 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
31 for more details.
32
33 You should have received copies of the GNU General Public License and the
34 GNU Lesser General Public License along with the GNU MP Library. If not,
35 see https://www.gnu.org/licenses/. */
36
37
38 #include "gmp-impl.h"
39
40 /* Evaluates a polynomial of degree k > 2, in the points +2^shift and -2^shift. */
41 int
mpn_toom_eval_pm2exp(mp_ptr xp2,mp_ptr xm2,unsigned k,mp_srcptr xp,mp_size_t n,mp_size_t hn,unsigned shift,mp_ptr tp)42 mpn_toom_eval_pm2exp (mp_ptr xp2, mp_ptr xm2, unsigned k,
43 mp_srcptr xp, mp_size_t n, mp_size_t hn, unsigned shift,
44 mp_ptr tp)
45 {
46 unsigned i;
47 int neg;
48 #if HAVE_NATIVE_mpn_addlsh_n
49 mp_limb_t cy;
50 #endif
51
52 ASSERT (k >= 3);
53 ASSERT (shift*k < GMP_NUMB_BITS);
54
55 ASSERT (hn > 0);
56 ASSERT (hn <= n);
57
58 /* The degree k is also the number of full-size coefficients, so
59 * that last coefficient, of size hn, starts at xp + k*n. */
60
61 #if HAVE_NATIVE_mpn_addlsh_n
62 xp2[n] = mpn_addlsh_n (xp2, xp, xp + 2*n, n, 2*shift);
63 for (i = 4; i < k; i += 2)
64 xp2[n] += mpn_addlsh_n (xp2, xp2, xp + i*n, n, i*shift);
65
66 tp[n] = mpn_lshift (tp, xp+n, n, shift);
67 for (i = 3; i < k; i+= 2)
68 tp[n] += mpn_addlsh_n (tp, tp, xp+i*n, n, i*shift);
69
70 if (k & 1)
71 {
72 cy = mpn_addlsh_n (tp, tp, xp+k*n, hn, k*shift);
73 MPN_INCR_U (tp + hn, n+1 - hn, cy);
74 }
75 else
76 {
77 cy = mpn_addlsh_n (xp2, xp2, xp+k*n, hn, k*shift);
78 MPN_INCR_U (xp2 + hn, n+1 - hn, cy);
79 }
80
81 #else /* !HAVE_NATIVE_mpn_addlsh_n */
82 xp2[n] = mpn_lshift (tp, xp+2*n, n, 2*shift);
83 xp2[n] += mpn_add_n (xp2, xp, tp, n);
84 for (i = 4; i < k; i += 2)
85 {
86 xp2[n] += mpn_lshift (tp, xp + i*n, n, i*shift);
87 xp2[n] += mpn_add_n (xp2, xp2, tp, n);
88 }
89
90 tp[n] = mpn_lshift (tp, xp+n, n, shift);
91 for (i = 3; i < k; i+= 2)
92 {
93 tp[n] += mpn_lshift (xm2, xp + i*n, n, i*shift);
94 tp[n] += mpn_add_n (tp, tp, xm2, n);
95 }
96
97 xm2[hn] = mpn_lshift (xm2, xp + k*n, hn, k*shift);
98 if (k & 1)
99 mpn_add (tp, tp, n+1, xm2, hn+1);
100 else
101 mpn_add (xp2, xp2, n+1, xm2, hn+1);
102 #endif /* !HAVE_NATIVE_mpn_addlsh_n */
103
104 neg = (mpn_cmp (xp2, tp, n + 1) < 0) ? ~0 : 0;
105
106 #if HAVE_NATIVE_mpn_add_n_sub_n
107 if (neg)
108 mpn_add_n_sub_n (xp2, xm2, tp, xp2, n + 1);
109 else
110 mpn_add_n_sub_n (xp2, xm2, xp2, tp, n + 1);
111 #else /* !HAVE_NATIVE_mpn_add_n_sub_n */
112 if (neg)
113 mpn_sub_n (xm2, tp, xp2, n + 1);
114 else
115 mpn_sub_n (xm2, xp2, tp, n + 1);
116
117 mpn_add_n (xp2, xp2, tp, n + 1);
118 #endif /* !HAVE_NATIVE_mpn_add_n_sub_n */
119
120 /* FIXME: the following asserts are useless if (k+1)*shift >= GMP_LIMB_BITS */
121 ASSERT ((k+1)*shift >= GMP_LIMB_BITS ||
122 xp2[n] < ((CNST_LIMB(1)<<((k+1)*shift))-1)/((CNST_LIMB(1)<<shift)-1));
123 ASSERT ((k+2)*shift >= GMP_LIMB_BITS ||
124 xm2[n] < ((CNST_LIMB(1)<<((k+2)*shift))-((k&1)?(CNST_LIMB(1)<<shift):1))/((CNST_LIMB(1)<<(2*shift))-1));
125
126 return neg;
127 }
128