1 /*
2 * Copyright (c) 2008-2020 Stefan Krah. All rights reserved.
3 *
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
6 * are met:
7 *
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 *
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS "AS IS" AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25 * SUCH DAMAGE.
26 */
27
28
29 #include "mpdecimal.h"
30
31 #include <assert.h>
32
33 #include "constants.h"
34 #include "fourstep.h"
35 #include "numbertheory.h"
36 #include "sixstep.h"
37 #include "umodarith.h"
38
39
40 /* Bignum: Cache efficient Matrix Fourier Transform for arrays of the
41 form 3 * 2**n (See literature/matrix-transform.txt). */
42
43
44 #ifndef PPRO
45 static inline void
std_size3_ntt(mpd_uint_t * x1,mpd_uint_t * x2,mpd_uint_t * x3,mpd_uint_t w3table[3],mpd_uint_t umod)46 std_size3_ntt(mpd_uint_t *x1, mpd_uint_t *x2, mpd_uint_t *x3,
47 mpd_uint_t w3table[3], mpd_uint_t umod)
48 {
49 mpd_uint_t r1, r2;
50 mpd_uint_t w;
51 mpd_uint_t s, tmp;
52
53
54 /* k = 0 -> w = 1 */
55 s = *x1;
56 s = addmod(s, *x2, umod);
57 s = addmod(s, *x3, umod);
58
59 r1 = s;
60
61 /* k = 1 */
62 s = *x1;
63
64 w = w3table[1];
65 tmp = MULMOD(*x2, w);
66 s = addmod(s, tmp, umod);
67
68 w = w3table[2];
69 tmp = MULMOD(*x3, w);
70 s = addmod(s, tmp, umod);
71
72 r2 = s;
73
74 /* k = 2 */
75 s = *x1;
76
77 w = w3table[2];
78 tmp = MULMOD(*x2, w);
79 s = addmod(s, tmp, umod);
80
81 w = w3table[1];
82 tmp = MULMOD(*x3, w);
83 s = addmod(s, tmp, umod);
84
85 *x3 = s;
86 *x2 = r2;
87 *x1 = r1;
88 }
89 #else /* PPRO */
90 static inline void
ppro_size3_ntt(mpd_uint_t * x1,mpd_uint_t * x2,mpd_uint_t * x3,mpd_uint_t w3table[3],mpd_uint_t umod,double * dmod,uint32_t dinvmod[3])91 ppro_size3_ntt(mpd_uint_t *x1, mpd_uint_t *x2, mpd_uint_t *x3, mpd_uint_t w3table[3],
92 mpd_uint_t umod, double *dmod, uint32_t dinvmod[3])
93 {
94 mpd_uint_t r1, r2;
95 mpd_uint_t w;
96 mpd_uint_t s, tmp;
97
98
99 /* k = 0 -> w = 1 */
100 s = *x1;
101 s = addmod(s, *x2, umod);
102 s = addmod(s, *x3, umod);
103
104 r1 = s;
105
106 /* k = 1 */
107 s = *x1;
108
109 w = w3table[1];
110 tmp = ppro_mulmod(*x2, w, dmod, dinvmod);
111 s = addmod(s, tmp, umod);
112
113 w = w3table[2];
114 tmp = ppro_mulmod(*x3, w, dmod, dinvmod);
115 s = addmod(s, tmp, umod);
116
117 r2 = s;
118
119 /* k = 2 */
120 s = *x1;
121
122 w = w3table[2];
123 tmp = ppro_mulmod(*x2, w, dmod, dinvmod);
124 s = addmod(s, tmp, umod);
125
126 w = w3table[1];
127 tmp = ppro_mulmod(*x3, w, dmod, dinvmod);
128 s = addmod(s, tmp, umod);
129
130 *x3 = s;
131 *x2 = r2;
132 *x1 = r1;
133 }
134 #endif
135
136
137 /* forward transform, sign = -1; transform length = 3 * 2**n */
138 int
four_step_fnt(mpd_uint_t * a,mpd_size_t n,int modnum)139 four_step_fnt(mpd_uint_t *a, mpd_size_t n, int modnum)
140 {
141 mpd_size_t R = 3; /* number of rows */
142 mpd_size_t C = n / 3; /* number of columns */
143 mpd_uint_t w3table[3];
144 mpd_uint_t kernel, w0, w1, wstep;
145 mpd_uint_t *s, *p0, *p1, *p2;
146 mpd_uint_t umod;
147 #ifdef PPRO
148 double dmod;
149 uint32_t dinvmod[3];
150 #endif
151 mpd_size_t i, k;
152
153
154 assert(n >= 48);
155 assert(n <= 3*MPD_MAXTRANSFORM_2N);
156
157
158 /* Length R transform on the columns. */
159 SETMODULUS(modnum);
160 _mpd_init_w3table(w3table, -1, modnum);
161 for (p0=a, p1=p0+C, p2=p0+2*C; p0<a+C; p0++,p1++,p2++) {
162
163 SIZE3_NTT(p0, p1, p2, w3table);
164 }
165
166 /* Multiply each matrix element (addressed by i*C+k) by r**(i*k). */
167 kernel = _mpd_getkernel(n, -1, modnum);
168 for (i = 1; i < R; i++) {
169 w0 = 1; /* r**(i*0): initial value for k=0 */
170 w1 = POWMOD(kernel, i); /* r**(i*1): initial value for k=1 */
171 wstep = MULMOD(w1, w1); /* r**(2*i) */
172 for (k = 0; k < C-1; k += 2) {
173 mpd_uint_t x0 = a[i*C+k];
174 mpd_uint_t x1 = a[i*C+k+1];
175 MULMOD2(&x0, w0, &x1, w1);
176 MULMOD2C(&w0, &w1, wstep); /* r**(i*(k+2)) = r**(i*k) * r**(2*i) */
177 a[i*C+k] = x0;
178 a[i*C+k+1] = x1;
179 }
180 }
181
182 /* Length C transform on the rows. */
183 for (s = a; s < a+n; s += C) {
184 if (!six_step_fnt(s, C, modnum)) {
185 return 0;
186 }
187 }
188
189 #if 0
190 /* An unordered transform is sufficient for convolution. */
191 /* Transpose the matrix. */
192 #include "transpose.h"
193 transpose_3xpow2(a, R, C);
194 #endif
195
196 return 1;
197 }
198
199 /* backward transform, sign = 1; transform length = 3 * 2**n */
200 int
inv_four_step_fnt(mpd_uint_t * a,mpd_size_t n,int modnum)201 inv_four_step_fnt(mpd_uint_t *a, mpd_size_t n, int modnum)
202 {
203 mpd_size_t R = 3; /* number of rows */
204 mpd_size_t C = n / 3; /* number of columns */
205 mpd_uint_t w3table[3];
206 mpd_uint_t kernel, w0, w1, wstep;
207 mpd_uint_t *s, *p0, *p1, *p2;
208 mpd_uint_t umod;
209 #ifdef PPRO
210 double dmod;
211 uint32_t dinvmod[3];
212 #endif
213 mpd_size_t i, k;
214
215
216 assert(n >= 48);
217 assert(n <= 3*MPD_MAXTRANSFORM_2N);
218
219
220 #if 0
221 /* An unordered transform is sufficient for convolution. */
222 /* Transpose the matrix, producing an R*C matrix. */
223 #include "transpose.h"
224 transpose_3xpow2(a, C, R);
225 #endif
226
227 /* Length C transform on the rows. */
228 for (s = a; s < a+n; s += C) {
229 if (!inv_six_step_fnt(s, C, modnum)) {
230 return 0;
231 }
232 }
233
234 /* Multiply each matrix element (addressed by i*C+k) by r**(i*k). */
235 SETMODULUS(modnum);
236 kernel = _mpd_getkernel(n, 1, modnum);
237 for (i = 1; i < R; i++) {
238 w0 = 1;
239 w1 = POWMOD(kernel, i);
240 wstep = MULMOD(w1, w1);
241 for (k = 0; k < C; k += 2) {
242 mpd_uint_t x0 = a[i*C+k];
243 mpd_uint_t x1 = a[i*C+k+1];
244 MULMOD2(&x0, w0, &x1, w1);
245 MULMOD2C(&w0, &w1, wstep);
246 a[i*C+k] = x0;
247 a[i*C+k+1] = x1;
248 }
249 }
250
251 /* Length R transform on the columns. */
252 _mpd_init_w3table(w3table, 1, modnum);
253 for (p0=a, p1=p0+C, p2=p0+2*C; p0<a+C; p0++,p1++,p2++) {
254
255 SIZE3_NTT(p0, p1, p2, w3table);
256 }
257
258 return 1;
259 }
260