xref: /netbsd-src/external/lgpl3/gmp/dist/mpn/generic/toom52_mul.c (revision 72c7faa4dbb41dbb0238d6b4a109da0d4b236dd4)
1 /* mpn_toom52_mul -- Multiply {ap,an} and {bp,bn} where an is nominally 4/3
2    times as large as bn.  Or more accurately, bn < an < 2 bn.
3 
4    Contributed to the GNU project by Marco Bodrato.
5 
6    The idea of applying toom to unbalanced multiplication is due to Marco
7    Bodrato and Alberto Zanoni.
8 
9    THE FUNCTION IN THIS FILE IS INTERNAL WITH A MUTABLE INTERFACE.  IT IS ONLY
10    SAFE TO REACH IT THROUGH DOCUMENTED INTERFACES.  IN FACT, IT IS ALMOST
11    GUARANTEED THAT IT WILL CHANGE OR DISAPPEAR IN A FUTURE GNU MP RELEASE.
12 
13 Copyright 2009 Free Software Foundation, Inc.
14 
15 This file is part of the GNU MP Library.
16 
17 The GNU MP Library is free software; you can redistribute it and/or modify
18 it under the terms of either:
19 
20   * the GNU Lesser General Public License as published by the Free
21     Software Foundation; either version 3 of the License, or (at your
22     option) any later version.
23 
24 or
25 
26   * the GNU General Public License as published by the Free Software
27     Foundation; either version 2 of the License, or (at your option) any
28     later version.
29 
30 or both in parallel, as here.
31 
32 The GNU MP Library is distributed in the hope that it will be useful, but
33 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
34 or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
35 for more details.
36 
37 You should have received copies of the GNU General Public License and the
38 GNU Lesser General Public License along with the GNU MP Library.  If not,
39 see https://www.gnu.org/licenses/.  */
40 
41 
42 #include "gmp-impl.h"
43 
44 /* Evaluate in: -2, -1, 0, +1, +2, +inf
45 
46   <-s-><--n--><--n--><--n--><--n-->
47    ___ ______ ______ ______ ______
48   |a4_|___a3_|___a2_|___a1_|___a0_|
49 			|b1|___b0_|
50 			<t-><--n-->
51 
52   v0  =  a0                  * b0      #   A(0)*B(0)
53   v1  = (a0+ a1+ a2+ a3+  a4)*(b0+ b1) #   A(1)*B(1)      ah  <= 4   bh <= 1
54   vm1 = (a0- a1+ a2- a3+  a4)*(b0- b1) #  A(-1)*B(-1)    |ah| <= 2   bh  = 0
55   v2  = (a0+2a1+4a2+8a3+16a4)*(b0+2b1) #   A(2)*B(2)      ah  <= 30  bh <= 2
56   vm2 = (a0-2a1+4a2-8a3+16a4)*(b0-2b1) #  A(-2)*B(-2)    |ah| <= 20 |bh|<= 1
57   vinf=                   a4 *     b1  # A(inf)*B(inf)
58 
59   Some slight optimization in evaluation are taken from the paper:
60   "Towards Optimal Toom-Cook Multiplication for Univariate and
61   Multivariate Polynomials in Characteristic 2 and 0."
62 */
63 
64 void
mpn_toom52_mul(mp_ptr pp,mp_srcptr ap,mp_size_t an,mp_srcptr bp,mp_size_t bn,mp_ptr scratch)65 mpn_toom52_mul (mp_ptr pp,
66 		mp_srcptr ap, mp_size_t an,
67 		mp_srcptr bp, mp_size_t bn, mp_ptr scratch)
68 {
69   mp_size_t n, s, t;
70   enum toom6_flags flags;
71 
72 #define a0  ap
73 #define a1  (ap + n)
74 #define a2  (ap + 2 * n)
75 #define a3  (ap + 3 * n)
76 #define a4  (ap + 4 * n)
77 #define b0  bp
78 #define b1  (bp + n)
79 
80   n = 1 + (2 * an >= 5 * bn ? (an - 1) / (size_t) 5 : (bn - 1) >> 1);
81 
82   s = an - 4 * n;
83   t = bn - n;
84 
85   ASSERT (0 < s && s <= n);
86   ASSERT (0 < t && t <= n);
87 
88   /* Ensures that 5 values of n+1 limbs each fits in the product area.
89      Borderline cases are an = 32, bn = 8, n = 7, and an = 36, bn = 9,
90      n = 8. */
91   ASSERT (s+t >= 5);
92 
93 #define v0    pp				/* 2n */
94 #define vm1   (scratch)				/* 2n+1 */
95 #define v1    (pp + 2 * n)			/* 2n+1 */
96 #define vm2   (scratch + 2 * n + 1)		/* 2n+1 */
97 #define v2    (scratch + 4 * n + 2)		/* 2n+1 */
98 #define vinf  (pp + 5 * n)			/* s+t */
99 #define bs1    pp				/* n+1 */
100 #define bsm1  (scratch + 2 * n + 2)		/* n   */
101 #define asm1  (scratch + 3 * n + 3)		/* n+1 */
102 #define asm2  (scratch + 4 * n + 4)		/* n+1 */
103 #define bsm2  (pp + n + 1)			/* n+1 */
104 #define bs2   (pp + 2 * n + 2)			/* n+1 */
105 #define as2   (pp + 3 * n + 3)			/* n+1 */
106 #define as1   (pp + 4 * n + 4)			/* n+1 */
107 
108   /* Scratch need is 6 * n + 3 + 1. We need one extra limb, because
109      products will overwrite 2n+2 limbs. */
110 
111 #define a0a2  scratch
112 #define a1a3  asm1
113 
114   /* Compute as2 and asm2.  */
115   flags = (enum toom6_flags) (toom6_vm2_neg & mpn_toom_eval_pm2 (as2, asm2, 4, ap, n, s, a1a3));
116 
117   /* Compute bs1 and bsm1.  */
118   if (t == n)
119     {
120 #if HAVE_NATIVE_mpn_add_n_sub_n
121       mp_limb_t cy;
122 
123       if (mpn_cmp (b0, b1, n) < 0)
124 	{
125 	  cy = mpn_add_n_sub_n (bs1, bsm1, b1, b0, n);
126 	  flags = (enum toom6_flags) (flags ^ toom6_vm1_neg);
127 	}
128       else
129 	{
130 	  cy = mpn_add_n_sub_n (bs1, bsm1, b0, b1, n);
131 	}
132       bs1[n] = cy >> 1;
133 #else
134       bs1[n] = mpn_add_n (bs1, b0, b1, n);
135       if (mpn_cmp (b0, b1, n) < 0)
136 	{
137 	  mpn_sub_n (bsm1, b1, b0, n);
138 	  flags = (enum toom6_flags) (flags ^ toom6_vm1_neg);
139 	}
140       else
141 	{
142 	  mpn_sub_n (bsm1, b0, b1, n);
143 	}
144 #endif
145     }
146   else
147     {
148       bs1[n] = mpn_add (bs1, b0, n, b1, t);
149       if (mpn_zero_p (b0 + t, n - t) && mpn_cmp (b0, b1, t) < 0)
150 	{
151 	  mpn_sub_n (bsm1, b1, b0, t);
152 	  MPN_ZERO (bsm1 + t, n - t);
153 	  flags = (enum toom6_flags) (flags ^ toom6_vm1_neg);
154 	}
155       else
156 	{
157 	  mpn_sub (bsm1, b0, n, b1, t);
158 	}
159     }
160 
161   /* Compute bs2 and bsm2, recycling bs1 and bsm1. bs2=bs1+b1; bsm2=bsm1-b1  */
162   mpn_add (bs2, bs1, n+1, b1, t);
163   if (flags & toom6_vm1_neg)
164     {
165       bsm2[n] = mpn_add (bsm2, bsm1, n, b1, t);
166       flags = (enum toom6_flags) (flags ^ toom6_vm2_neg);
167     }
168   else
169     {
170       bsm2[n] = 0;
171       if (t == n)
172 	{
173 	  if (mpn_cmp (bsm1, b1, n) < 0)
174 	    {
175 	      mpn_sub_n (bsm2, b1, bsm1, n);
176 	      flags = (enum toom6_flags) (flags ^ toom6_vm2_neg);
177 	    }
178 	  else
179 	    {
180 	      mpn_sub_n (bsm2, bsm1, b1, n);
181 	    }
182 	}
183       else
184 	{
185 	  if (mpn_zero_p (bsm1 + t, n - t) && mpn_cmp (bsm1, b1, t) < 0)
186 	    {
187 	      mpn_sub_n (bsm2, b1, bsm1, t);
188 	      MPN_ZERO (bsm2 + t, n - t);
189 	      flags = (enum toom6_flags) (flags ^ toom6_vm2_neg);
190 	    }
191 	  else
192 	    {
193 	      mpn_sub (bsm2, bsm1, n, b1, t);
194 	    }
195 	}
196     }
197 
198   /* Compute as1 and asm1.  */
199   flags = (enum toom6_flags) (flags ^ (toom6_vm1_neg & mpn_toom_eval_pm1 (as1, asm1, 4, ap, n, s, a0a2)));
200 
201   ASSERT (as1[n] <= 4);
202   ASSERT (bs1[n] <= 1);
203   ASSERT (asm1[n] <= 2);
204 /*   ASSERT (bsm1[n] <= 1); */
205   ASSERT (as2[n] <=30);
206   ASSERT (bs2[n] <= 2);
207   ASSERT (asm2[n] <= 20);
208   ASSERT (bsm2[n] <= 1);
209 
210   /* vm1, 2n+1 limbs */
211   mpn_mul (vm1, asm1, n+1, bsm1, n);  /* W4 */
212 
213   /* vm2, 2n+1 limbs */
214   mpn_mul_n (vm2, asm2, bsm2, n+1);  /* W2 */
215 
216   /* v2, 2n+1 limbs */
217   mpn_mul_n (v2, as2, bs2, n+1);  /* W1 */
218 
219   /* v1, 2n+1 limbs */
220   mpn_mul_n (v1, as1, bs1, n+1);  /* W3 */
221 
222   /* vinf, s+t limbs */   /* W0 */
223   if (s > t)  mpn_mul (vinf, a4, s, b1, t);
224   else        mpn_mul (vinf, b1, t, a4, s);
225 
226   /* v0, 2n limbs */
227   mpn_mul_n (v0, ap, bp, n);  /* W5 */
228 
229   mpn_toom_interpolate_6pts (pp, n, flags, vm1, vm2, v2, t + s);
230 
231 #undef v0
232 #undef vm1
233 #undef v1
234 #undef vm2
235 #undef v2
236 #undef vinf
237 #undef bs1
238 #undef bs2
239 #undef bsm1
240 #undef bsm2
241 #undef asm1
242 #undef asm2
243 #undef as1
244 #undef as2
245 #undef a0a2
246 #undef b0b2
247 #undef a1a3
248 #undef a0
249 #undef a1
250 #undef a2
251 #undef a3
252 #undef b0
253 #undef b1
254 #undef b2
255 
256 }
257