1 | |
---|
2 | |
---|
3 | #include <NTL/lzz_pX.h> |
---|
4 | |
---|
5 | #include <NTL/new.h> |
---|
6 | |
---|
7 | NTL_START_IMPL |
---|
8 | |
---|
9 | |
---|
10 | |
---|
11 | long divide(zz_pX& q, const zz_pX& a, const zz_pX& b) |
---|
12 | { |
---|
13 | if (IsZero(b)) { |
---|
14 | if (IsZero(a)) { |
---|
15 | clear(q); |
---|
16 | return 1; |
---|
17 | } |
---|
18 | else |
---|
19 | return 0; |
---|
20 | } |
---|
21 | |
---|
22 | zz_pX lq, r; |
---|
23 | DivRem(lq, r, a, b); |
---|
24 | if (!IsZero(r)) return 0; |
---|
25 | q = lq; |
---|
26 | return 1; |
---|
27 | } |
---|
28 | |
---|
29 | long divide(const zz_pX& a, const zz_pX& b) |
---|
30 | { |
---|
31 | if (IsZero(b)) return IsZero(a); |
---|
32 | zz_pX lq, r; |
---|
33 | DivRem(lq, r, a, b); |
---|
34 | if (!IsZero(r)) return 0; |
---|
35 | return 1; |
---|
36 | } |
---|
37 | |
---|
38 | |
---|
39 | |
---|
40 | void zz_pXMatrix::operator=(const zz_pXMatrix& M) |
---|
41 | { |
---|
42 | elts[0][0] = M.elts[0][0]; |
---|
43 | elts[0][1] = M.elts[0][1]; |
---|
44 | elts[1][0] = M.elts[1][0]; |
---|
45 | elts[1][1] = M.elts[1][1]; |
---|
46 | } |
---|
47 | |
---|
48 | |
---|
49 | void RightShift(zz_pX& x, const zz_pX& a, long n) |
---|
50 | { |
---|
51 | if (IsZero(a)) { |
---|
52 | clear(x); |
---|
53 | return; |
---|
54 | } |
---|
55 | |
---|
56 | if (n < 0) { |
---|
57 | if (n < -NTL_MAX_LONG) Error("overflow in RightShift"); |
---|
58 | LeftShift(x, a, -n); |
---|
59 | return; |
---|
60 | } |
---|
61 | |
---|
62 | long da = deg(a); |
---|
63 | long i; |
---|
64 | |
---|
65 | if (da < n) { |
---|
66 | clear(x); |
---|
67 | return; |
---|
68 | } |
---|
69 | |
---|
70 | if (&x != &a) |
---|
71 | x.rep.SetLength(da-n+1); |
---|
72 | |
---|
73 | for (i = 0; i <= da-n; i++) |
---|
74 | x.rep[i] = a.rep[i+n]; |
---|
75 | |
---|
76 | if (&x == &a) |
---|
77 | x.rep.SetLength(da-n+1); |
---|
78 | |
---|
79 | x.normalize(); |
---|
80 | } |
---|
81 | |
---|
82 | void LeftShift(zz_pX& x, const zz_pX& a, long n) |
---|
83 | { |
---|
84 | if (IsZero(a)) { |
---|
85 | clear(x); |
---|
86 | return; |
---|
87 | } |
---|
88 | |
---|
89 | if (n < 0) { |
---|
90 | if (n < -NTL_MAX_LONG) |
---|
91 | clear(x); |
---|
92 | else |
---|
93 | RightShift(x, a, -n); |
---|
94 | return; |
---|
95 | } |
---|
96 | |
---|
97 | if (NTL_OVERFLOW(n, 1, 0)) |
---|
98 | Error("overflow in LeftShift"); |
---|
99 | |
---|
100 | long m = a.rep.length(); |
---|
101 | |
---|
102 | x.rep.SetLength(m+n); |
---|
103 | |
---|
104 | long i; |
---|
105 | for (i = m-1; i >= 0; i--) |
---|
106 | x.rep[i+n] = a.rep[i]; |
---|
107 | |
---|
108 | for (i = 0; i < n; i++) |
---|
109 | clear(x.rep[i]); |
---|
110 | } |
---|
111 | |
---|
112 | |
---|
113 | void ShiftAdd(zz_pX& U, const zz_pX& V, long n) |
---|
114 | // assumes input does not alias output |
---|
115 | { |
---|
116 | if (IsZero(V)) |
---|
117 | return; |
---|
118 | |
---|
119 | long du = deg(U); |
---|
120 | long dv = deg(V); |
---|
121 | |
---|
122 | long d = max(du, n+dv); |
---|
123 | |
---|
124 | U.rep.SetLength(d+1); |
---|
125 | long i; |
---|
126 | |
---|
127 | for (i = du+1; i <= d; i++) |
---|
128 | clear(U.rep[i]); |
---|
129 | |
---|
130 | for (i = 0; i <= dv; i++) |
---|
131 | add(U.rep[i+n], U.rep[i+n], V.rep[i]); |
---|
132 | |
---|
133 | U.normalize(); |
---|
134 | } |
---|
135 | |
---|
136 | void ShiftSub(zz_pX& U, const zz_pX& V, long n) |
---|
137 | // assumes input does not alias output |
---|
138 | { |
---|
139 | if (IsZero(V)) |
---|
140 | return; |
---|
141 | |
---|
142 | long du = deg(U); |
---|
143 | long dv = deg(V); |
---|
144 | |
---|
145 | long d = max(du, n+dv); |
---|
146 | |
---|
147 | U.rep.SetLength(d+1); |
---|
148 | long i; |
---|
149 | |
---|
150 | for (i = du+1; i <= d; i++) |
---|
151 | clear(U.rep[i]); |
---|
152 | |
---|
153 | for (i = 0; i <= dv; i++) |
---|
154 | sub(U.rep[i+n], U.rep[i+n], V.rep[i]); |
---|
155 | |
---|
156 | U.normalize(); |
---|
157 | } |
---|
158 | |
---|
159 | void mul(zz_pX& U, zz_pX& V, const zz_pXMatrix& M) |
---|
160 | // (U, V)^T = M*(U, V)^T |
---|
161 | { |
---|
162 | long d = deg(U) - deg(M(1,1)); |
---|
163 | long k = NextPowerOfTwo(d - 1); |
---|
164 | |
---|
165 | // When the GCD algorithm is run on polynomials of degree n, n-1, |
---|
166 | // where n is a power of two, then d-1 is likely to be a power of two. |
---|
167 | // It would be more natural to set k = NextPowerOfTwo(d+1), but this |
---|
168 | // would be much less efficient in this case. |
---|
169 | |
---|
170 | long n = (1L << k); |
---|
171 | long xx; |
---|
172 | zz_p a0, a1, b0, b1, c0, d0, u0, u1, v0, v1, nu0, nu1, nv0; |
---|
173 | zz_p t1, t2; |
---|
174 | |
---|
175 | if (n == d-1) |
---|
176 | xx = 1; |
---|
177 | else if (n == d) |
---|
178 | xx = 2; |
---|
179 | else |
---|
180 | xx = 3; |
---|
181 | |
---|
182 | switch (xx) { |
---|
183 | case 1: |
---|
184 | GetCoeff(a0, M(0,0), 0); |
---|
185 | GetCoeff(a1, M(0,0), 1); |
---|
186 | GetCoeff(b0, M(0,1), 0); |
---|
187 | GetCoeff(b1, M(0,1), 1); |
---|
188 | GetCoeff(c0, M(1,0), 0); |
---|
189 | GetCoeff(d0, M(1,1), 0); |
---|
190 | |
---|
191 | GetCoeff(u0, U, 0); |
---|
192 | GetCoeff(u1, U, 1); |
---|
193 | GetCoeff(v0, V, 0); |
---|
194 | GetCoeff(v1, V, 1); |
---|
195 | |
---|
196 | mul(t1, (a0), (u0)); |
---|
197 | mul(t2, (b0), (v0)); |
---|
198 | add(t1, t1, t2); |
---|
199 | nu0 = t1; |
---|
200 | |
---|
201 | mul(t1, (a1), (u0)); |
---|
202 | mul(t2, (a0), (u1)); |
---|
203 | add(t1, t1, t2); |
---|
204 | mul(t2, (b1), (v0)); |
---|
205 | add(t1, t1, t2); |
---|
206 | mul(t2, (b0), (v1)); |
---|
207 | add(t1, t1, t2); |
---|
208 | nu1 = t1; |
---|
209 | |
---|
210 | mul(t1, (c0), (u0)); |
---|
211 | mul(t2, (d0), (v0)); |
---|
212 | add (t1, t1, t2); |
---|
213 | nv0 = t1; |
---|
214 | |
---|
215 | break; |
---|
216 | |
---|
217 | case 2: |
---|
218 | GetCoeff(a0, M(0,0), 0); |
---|
219 | GetCoeff(b0, M(0,1), 0); |
---|
220 | |
---|
221 | GetCoeff(u0, U, 0); |
---|
222 | GetCoeff(v0, V, 0); |
---|
223 | |
---|
224 | mul(t1, (a0), (u0)); |
---|
225 | mul(t2, (b0), (v0)); |
---|
226 | add(t1, t1, t2); |
---|
227 | nu0 = t1; |
---|
228 | |
---|
229 | break; |
---|
230 | |
---|
231 | case 3: |
---|
232 | break; |
---|
233 | |
---|
234 | } |
---|
235 | |
---|
236 | fftRep RU(INIT_SIZE, k), RV(INIT_SIZE, k), R1(INIT_SIZE, k), |
---|
237 | R2(INIT_SIZE, k); |
---|
238 | |
---|
239 | TofftRep(RU, U, k); |
---|
240 | TofftRep(RV, V, k); |
---|
241 | |
---|
242 | TofftRep(R1, M(0,0), k); |
---|
243 | mul(R1, R1, RU); |
---|
244 | TofftRep(R2, M(0,1), k); |
---|
245 | mul(R2, R2, RV); |
---|
246 | add(R1, R1, R2); |
---|
247 | FromfftRep(U, R1, 0, d); |
---|
248 | |
---|
249 | TofftRep(R1, M(1,0), k); |
---|
250 | mul(R1, R1, RU); |
---|
251 | TofftRep(R2, M(1,1), k); |
---|
252 | mul(R2, R2, RV); |
---|
253 | add(R1, R1, R2); |
---|
254 | FromfftRep(V, R1, 0, d-1); |
---|
255 | |
---|
256 | // now fix-up results |
---|
257 | |
---|
258 | switch (xx) { |
---|
259 | case 1: |
---|
260 | GetCoeff(u0, U, 0); |
---|
261 | sub(u0, u0, nu0); |
---|
262 | SetCoeff(U, d-1, u0); |
---|
263 | SetCoeff(U, 0, nu0); |
---|
264 | |
---|
265 | GetCoeff(u1, U, 1); |
---|
266 | sub(u1, u1, nu1); |
---|
267 | SetCoeff(U, d, u1); |
---|
268 | SetCoeff(U, 1, nu1); |
---|
269 | |
---|
270 | GetCoeff(v0, V, 0); |
---|
271 | sub(v0, v0, nv0); |
---|
272 | SetCoeff(V, d-1, v0); |
---|
273 | SetCoeff(V, 0, nv0); |
---|
274 | |
---|
275 | break; |
---|
276 | |
---|
277 | |
---|
278 | case 2: |
---|
279 | GetCoeff(u0, U, 0); |
---|
280 | sub(u0, u0, nu0); |
---|
281 | SetCoeff(U, d, u0); |
---|
282 | SetCoeff(U, 0, nu0); |
---|
283 | |
---|
284 | break; |
---|
285 | |
---|
286 | } |
---|
287 | } |
---|
288 | |
---|
289 | |
---|
290 | void mul(zz_pXMatrix& A, zz_pXMatrix& B, zz_pXMatrix& C) |
---|
291 | // A = B*C, B and C are destroyed |
---|
292 | { |
---|
293 | long db = deg(B(1,1)); |
---|
294 | long dc = deg(C(1,1)); |
---|
295 | long da = db + dc; |
---|
296 | |
---|
297 | long k = NextPowerOfTwo(da+1); |
---|
298 | |
---|
299 | fftRep B00, B01, B10, B11, C0, C1, T1, T2; |
---|
300 | |
---|
301 | TofftRep(B00, B(0,0), k); B(0,0).kill(); |
---|
302 | TofftRep(B01, B(0,1), k); B(0,1).kill(); |
---|
303 | TofftRep(B10, B(1,0), k); B(1,0).kill(); |
---|
304 | TofftRep(B11, B(1,1), k); B(1,1).kill(); |
---|
305 | |
---|
306 | TofftRep(C0, C(0,0), k); C(0,0).kill(); |
---|
307 | TofftRep(C1, C(1,0), k); C(1,0).kill(); |
---|
308 | |
---|
309 | mul(T1, B00, C0); |
---|
310 | mul(T2, B01, C1); |
---|
311 | add(T1, T1, T2); |
---|
312 | FromfftRep(A(0,0), T1, 0, da); |
---|
313 | |
---|
314 | mul(T1, B10, C0); |
---|
315 | mul(T2, B11, C1); |
---|
316 | add(T1, T1, T2); |
---|
317 | FromfftRep(A(1,0), T1, 0, da); |
---|
318 | |
---|
319 | TofftRep(C0, C(0,1), k); C(0,1).kill(); |
---|
320 | TofftRep(C1, C(1,1), k); C(1,1).kill(); |
---|
321 | |
---|
322 | mul(T1, B00, C0); |
---|
323 | mul(T2, B01, C1); |
---|
324 | add(T1, T1, T2); |
---|
325 | FromfftRep(A(0,1), T1, 0, da); |
---|
326 | |
---|
327 | mul(T1, B10, C0); |
---|
328 | mul(T2, B11, C1); |
---|
329 | add(T1, T1, T2); |
---|
330 | FromfftRep(A(1,1), T1, 0, da); |
---|
331 | } |
---|
332 | |
---|
333 | void IterHalfGCD(zz_pXMatrix& M_out, zz_pX& U, zz_pX& V, long d_red) |
---|
334 | { |
---|
335 | M_out(0,0).SetMaxLength(d_red); |
---|
336 | M_out(0,1).SetMaxLength(d_red); |
---|
337 | M_out(1,0).SetMaxLength(d_red); |
---|
338 | M_out(1,1).SetMaxLength(d_red); |
---|
339 | |
---|
340 | set(M_out(0,0)); clear(M_out(0,1)); |
---|
341 | clear(M_out(1,0)); set(M_out(1,1)); |
---|
342 | |
---|
343 | long goal = deg(U) - d_red; |
---|
344 | |
---|
345 | if (deg(V) <= goal) |
---|
346 | return; |
---|
347 | |
---|
348 | zz_pX Q, t(INIT_SIZE, d_red); |
---|
349 | |
---|
350 | while (deg(V) > goal) { |
---|
351 | PlainDivRem(Q, U, U, V); |
---|
352 | swap(U, V); |
---|
353 | |
---|
354 | mul(t, Q, M_out(1,0)); |
---|
355 | sub(t, M_out(0,0), t); |
---|
356 | M_out(0,0) = M_out(1,0); |
---|
357 | M_out(1,0) = t; |
---|
358 | |
---|
359 | mul(t, Q, M_out(1,1)); |
---|
360 | sub(t, M_out(0,1), t); |
---|
361 | M_out(0,1) = M_out(1,1); |
---|
362 | M_out(1,1) = t; |
---|
363 | } |
---|
364 | } |
---|
365 | |
---|
366 | |
---|
367 | |
---|
368 | void HalfGCD(zz_pXMatrix& M_out, const zz_pX& U, const zz_pX& V, long d_red) |
---|
369 | { |
---|
370 | if (IsZero(V) || deg(V) <= deg(U) - d_red) { |
---|
371 | set(M_out(0,0)); clear(M_out(0,1)); |
---|
372 | clear(M_out(1,0)); set(M_out(1,1)); |
---|
373 | |
---|
374 | return; |
---|
375 | } |
---|
376 | |
---|
377 | |
---|
378 | long n = deg(U) - 2*d_red + 2; |
---|
379 | if (n < 0) n = 0; |
---|
380 | |
---|
381 | zz_pX U1, V1; |
---|
382 | |
---|
383 | RightShift(U1, U, n); |
---|
384 | RightShift(V1, V, n); |
---|
385 | |
---|
386 | if (d_red <= NTL_zz_pX_HalfGCD_CROSSOVER) { |
---|
387 | IterHalfGCD(M_out, U1, V1, d_red); |
---|
388 | return; |
---|
389 | } |
---|
390 | |
---|
391 | long d1 = (d_red + 1)/2; |
---|
392 | if (d1 < 1) d1 = 1; |
---|
393 | if (d1 >= d_red) d1 = d_red - 1; |
---|
394 | |
---|
395 | zz_pXMatrix M1; |
---|
396 | |
---|
397 | HalfGCD(M1, U1, V1, d1); |
---|
398 | mul(U1, V1, M1); |
---|
399 | |
---|
400 | long d2 = deg(V1) - deg(U) + n + d_red; |
---|
401 | |
---|
402 | if (IsZero(V1) || d2 <= 0) { |
---|
403 | M_out = M1; |
---|
404 | return; |
---|
405 | } |
---|
406 | |
---|
407 | |
---|
408 | zz_pX Q; |
---|
409 | zz_pXMatrix M2; |
---|
410 | |
---|
411 | DivRem(Q, U1, U1, V1); |
---|
412 | swap(U1, V1); |
---|
413 | |
---|
414 | HalfGCD(M2, U1, V1, d2); |
---|
415 | |
---|
416 | zz_pX t(INIT_SIZE, deg(M1(1,1))+deg(Q)+1); |
---|
417 | |
---|
418 | mul(t, Q, M1(1,0)); |
---|
419 | sub(t, M1(0,0), t); |
---|
420 | swap(M1(0,0), M1(1,0)); |
---|
421 | swap(M1(1,0), t); |
---|
422 | |
---|
423 | t.kill(); |
---|
424 | |
---|
425 | t.SetMaxLength(deg(M1(1,1))+deg(Q)+1); |
---|
426 | |
---|
427 | mul(t, Q, M1(1,1)); |
---|
428 | sub(t, M1(0,1), t); |
---|
429 | swap(M1(0,1), M1(1,1)); |
---|
430 | swap(M1(1,1), t); |
---|
431 | |
---|
432 | t.kill(); |
---|
433 | |
---|
434 | mul(M_out, M2, M1); |
---|
435 | } |
---|
436 | |
---|
437 | |
---|
438 | |
---|
439 | |
---|
440 | void XHalfGCD(zz_pXMatrix& M_out, zz_pX& U, zz_pX& V, long d_red) |
---|
441 | { |
---|
442 | if (IsZero(V) || deg(V) <= deg(U) - d_red) { |
---|
443 | set(M_out(0,0)); clear(M_out(0,1)); |
---|
444 | clear(M_out(1,0)); set(M_out(1,1)); |
---|
445 | |
---|
446 | return; |
---|
447 | } |
---|
448 | |
---|
449 | long du = deg(U); |
---|
450 | |
---|
451 | if (d_red <= NTL_zz_pX_HalfGCD_CROSSOVER) { |
---|
452 | IterHalfGCD(M_out, U, V, d_red); |
---|
453 | return; |
---|
454 | } |
---|
455 | |
---|
456 | long d1 = (d_red + 1)/2; |
---|
457 | if (d1 < 1) d1 = 1; |
---|
458 | if (d1 >= d_red) d1 = d_red - 1; |
---|
459 | |
---|
460 | zz_pXMatrix M1; |
---|
461 | |
---|
462 | HalfGCD(M1, U, V, d1); |
---|
463 | mul(U, V, M1); |
---|
464 | |
---|
465 | long d2 = deg(V) - du + d_red; |
---|
466 | |
---|
467 | if (IsZero(V) || d2 <= 0) { |
---|
468 | M_out = M1; |
---|
469 | return; |
---|
470 | } |
---|
471 | |
---|
472 | |
---|
473 | zz_pX Q; |
---|
474 | zz_pXMatrix M2; |
---|
475 | |
---|
476 | DivRem(Q, U, U, V); |
---|
477 | swap(U, V); |
---|
478 | |
---|
479 | XHalfGCD(M2, U, V, d2); |
---|
480 | |
---|
481 | zz_pX t(INIT_SIZE, deg(M1(1,1))+deg(Q)+1); |
---|
482 | |
---|
483 | mul(t, Q, M1(1,0)); |
---|
484 | sub(t, M1(0,0), t); |
---|
485 | swap(M1(0,0), M1(1,0)); |
---|
486 | swap(M1(1,0), t); |
---|
487 | |
---|
488 | t.kill(); |
---|
489 | |
---|
490 | t.SetMaxLength(deg(M1(1,1))+deg(Q)+1); |
---|
491 | |
---|
492 | mul(t, Q, M1(1,1)); |
---|
493 | sub(t, M1(0,1), t); |
---|
494 | swap(M1(0,1), M1(1,1)); |
---|
495 | swap(M1(1,1), t); |
---|
496 | |
---|
497 | t.kill(); |
---|
498 | |
---|
499 | mul(M_out, M2, M1); |
---|
500 | } |
---|
501 | |
---|
502 | void HalfGCD(zz_pX& U, zz_pX& V) |
---|
503 | { |
---|
504 | long d_red = (deg(U)+1)/2; |
---|
505 | |
---|
506 | if (IsZero(V) || deg(V) <= deg(U) - d_red) { |
---|
507 | return; |
---|
508 | } |
---|
509 | |
---|
510 | long du = deg(U); |
---|
511 | |
---|
512 | |
---|
513 | long d1 = (d_red + 1)/2; |
---|
514 | if (d1 < 1) d1 = 1; |
---|
515 | if (d1 >= d_red) d1 = d_red - 1; |
---|
516 | |
---|
517 | zz_pXMatrix M1; |
---|
518 | |
---|
519 | HalfGCD(M1, U, V, d1); |
---|
520 | mul(U, V, M1); |
---|
521 | |
---|
522 | long d2 = deg(V) - du + d_red; |
---|
523 | |
---|
524 | if (IsZero(V) || d2 <= 0) { |
---|
525 | return; |
---|
526 | } |
---|
527 | |
---|
528 | M1(0,0).kill(); |
---|
529 | M1(0,1).kill(); |
---|
530 | M1(1,0).kill(); |
---|
531 | M1(1,1).kill(); |
---|
532 | |
---|
533 | |
---|
534 | zz_pX Q; |
---|
535 | |
---|
536 | DivRem(Q, U, U, V); |
---|
537 | swap(U, V); |
---|
538 | |
---|
539 | HalfGCD(M1, U, V, d2); |
---|
540 | |
---|
541 | mul(U, V, M1); |
---|
542 | } |
---|
543 | |
---|
544 | |
---|
545 | void GCD(zz_pX& d, const zz_pX& u, const zz_pX& v) |
---|
546 | { |
---|
547 | zz_pX u1, v1; |
---|
548 | |
---|
549 | u1 = u; |
---|
550 | v1 = v; |
---|
551 | |
---|
552 | if (deg(u1) == deg(v1)) { |
---|
553 | if (IsZero(u1)) { |
---|
554 | clear(d); |
---|
555 | return; |
---|
556 | } |
---|
557 | |
---|
558 | rem(v1, v1, u1); |
---|
559 | } |
---|
560 | else if (deg(u1) < deg(v1)) { |
---|
561 | swap(u1, v1); |
---|
562 | } |
---|
563 | |
---|
564 | // deg(u1) > deg(v1) |
---|
565 | |
---|
566 | while (deg(u1) > NTL_zz_pX_GCD_CROSSOVER && !IsZero(v1)) { |
---|
567 | HalfGCD(u1, v1); |
---|
568 | |
---|
569 | if (!IsZero(v1)) { |
---|
570 | rem(u1, u1, v1); |
---|
571 | swap(u1, v1); |
---|
572 | } |
---|
573 | } |
---|
574 | |
---|
575 | PlainGCD(d, u1, v1); |
---|
576 | } |
---|
577 | |
---|
578 | |
---|
579 | |
---|
580 | void XGCD(zz_pX& d, zz_pX& s, zz_pX& t, const zz_pX& a, const zz_pX& b) |
---|
581 | { |
---|
582 | zz_p w; |
---|
583 | |
---|
584 | if (IsZero(a) && IsZero(b)) { |
---|
585 | clear(d); |
---|
586 | set(s); |
---|
587 | clear(t); |
---|
588 | return; |
---|
589 | } |
---|
590 | |
---|
591 | zz_pX U, V, Q; |
---|
592 | |
---|
593 | U = a; |
---|
594 | V = b; |
---|
595 | |
---|
596 | long flag = 0; |
---|
597 | |
---|
598 | if (deg(U) == deg(V)) { |
---|
599 | DivRem(Q, U, U, V); |
---|
600 | swap(U, V); |
---|
601 | flag = 1; |
---|
602 | } |
---|
603 | else if (deg(U) < deg(V)) { |
---|
604 | swap(U, V); |
---|
605 | flag = 2; |
---|
606 | } |
---|
607 | |
---|
608 | zz_pXMatrix M; |
---|
609 | |
---|
610 | XHalfGCD(M, U, V, deg(U)+1); |
---|
611 | |
---|
612 | d = U; |
---|
613 | |
---|
614 | if (flag == 0) { |
---|
615 | s = M(0,0); |
---|
616 | t = M(0,1); |
---|
617 | } |
---|
618 | else if (flag == 1) { |
---|
619 | s = M(0,1); |
---|
620 | mul(t, Q, M(0,1)); |
---|
621 | sub(t, M(0,0), t); |
---|
622 | } |
---|
623 | else { /* flag == 2 */ |
---|
624 | s = M(0,1); |
---|
625 | t = M(0,0); |
---|
626 | } |
---|
627 | |
---|
628 | // normalize |
---|
629 | |
---|
630 | inv(w, LeadCoeff(d)); |
---|
631 | mul(d, d, w); |
---|
632 | mul(s, s, w); |
---|
633 | mul(t, t, w); |
---|
634 | } |
---|
635 | |
---|
636 | |
---|
637 | |
---|
638 | |
---|
639 | |
---|
640 | |
---|
641 | |
---|
642 | void IterBuild(zz_p* a, long n) |
---|
643 | { |
---|
644 | long i, k; |
---|
645 | zz_p b, t; |
---|
646 | |
---|
647 | if (n <= 0) return; |
---|
648 | |
---|
649 | negate(a[0], a[0]); |
---|
650 | |
---|
651 | for (k = 1; k <= n-1; k++) { |
---|
652 | negate(b, a[k]); |
---|
653 | add(a[k], b, a[k-1]); |
---|
654 | for (i = k-1; i >= 1; i--) { |
---|
655 | mul(t, a[i], b); |
---|
656 | add(a[i], t, a[i-1]); |
---|
657 | } |
---|
658 | mul(a[0], a[0], b); |
---|
659 | } |
---|
660 | } |
---|
661 | |
---|
662 | void mul(zz_p* x, const zz_p* a, const zz_p* b, long n) |
---|
663 | { |
---|
664 | zz_p t, accum; |
---|
665 | |
---|
666 | long i, j, jmin, jmax; |
---|
667 | |
---|
668 | long d = 2*n-1; |
---|
669 | |
---|
670 | for (i = 0; i <= d; i++) { |
---|
671 | jmin = max(0, i-(n-1)); |
---|
672 | jmax = min(n-1, i); |
---|
673 | clear(accum); |
---|
674 | for (j = jmin; j <= jmax; j++) { |
---|
675 | mul(t, (a[j]), (b[i-j])); |
---|
676 | add(accum, accum, t); |
---|
677 | } |
---|
678 | if (i >= n) { |
---|
679 | add(accum, accum, (a[i-n])); |
---|
680 | add(accum, accum, (b[i-n])); |
---|
681 | } |
---|
682 | |
---|
683 | x[i] = accum; |
---|
684 | } |
---|
685 | } |
---|
686 | |
---|
687 | |
---|
688 | void BuildFromRoots(zz_pX& x, const vec_zz_p& a) |
---|
689 | { |
---|
690 | long n = a.length(); |
---|
691 | |
---|
692 | if (n == 0) { |
---|
693 | set(x); |
---|
694 | return; |
---|
695 | } |
---|
696 | |
---|
697 | long k0 = NextPowerOfTwo(NTL_zz_pX_MUL_CROSSOVER)-1; |
---|
698 | long crossover = 1L << k0; |
---|
699 | |
---|
700 | if (n <= NTL_zz_pX_MUL_CROSSOVER) { |
---|
701 | x.rep.SetMaxLength(n+1); |
---|
702 | x.rep = a; |
---|
703 | IterBuild(&x.rep[0], n); |
---|
704 | x.rep.SetLength(n+1); |
---|
705 | SetCoeff(x, n); |
---|
706 | return; |
---|
707 | } |
---|
708 | |
---|
709 | long k = NextPowerOfTwo(n); |
---|
710 | |
---|
711 | long m = 1L << k; |
---|
712 | long i, j; |
---|
713 | long l, width; |
---|
714 | |
---|
715 | zz_pX b(INIT_SIZE, m+1); |
---|
716 | |
---|
717 | b.rep = a; |
---|
718 | b.rep.SetLength(m+1); |
---|
719 | for (i = n; i < m; i++) |
---|
720 | clear(b.rep[i]); |
---|
721 | |
---|
722 | set(b.rep[m]); |
---|
723 | |
---|
724 | fftRep R1(INIT_SIZE, k), R2(INIT_SIZE, k); |
---|
725 | |
---|
726 | |
---|
727 | zz_p t1, one; |
---|
728 | set(one); |
---|
729 | |
---|
730 | vec_zz_p G(INIT_SIZE, crossover), H(INIT_SIZE, crossover); |
---|
731 | zz_p *g = G.elts(); |
---|
732 | zz_p *h = H.elts(); |
---|
733 | zz_p *tmp; |
---|
734 | |
---|
735 | for (i = 0; i < m; i+= crossover) { |
---|
736 | for (j = 0; j < crossover; j++) |
---|
737 | negate(g[j], b.rep[i+j]); |
---|
738 | |
---|
739 | if (k0 > 0) { |
---|
740 | for (j = 0; j < crossover; j+=2) { |
---|
741 | mul(t1, g[j], g[j+1]); |
---|
742 | add(g[j+1], g[j], g[j+1]); |
---|
743 | g[j] = t1; |
---|
744 | } |
---|
745 | } |
---|
746 | |
---|
747 | for (l = 1; l < k0; l++) { |
---|
748 | width = 1L << l; |
---|
749 | |
---|
750 | for (j = 0; j < crossover; j += 2*width) |
---|
751 | mul(&h[j], &g[j], &g[j+width], width); |
---|
752 | |
---|
753 | tmp = g; g = h; h = tmp; |
---|
754 | } |
---|
755 | |
---|
756 | for (j = 0; j < crossover; j++) |
---|
757 | b.rep[i+j] = g[j]; |
---|
758 | } |
---|
759 | |
---|
760 | for (l = k0; l < k; l++) { |
---|
761 | width = 1L << l; |
---|
762 | for (i = 0; i < m; i += 2*width) { |
---|
763 | t1 = b.rep[i+width]; |
---|
764 | set(b.rep[i+width]); |
---|
765 | TofftRep(R1, b, l+1, i, i+width); |
---|
766 | b.rep[i+width] = t1; |
---|
767 | t1 = b.rep[i+2*width]; |
---|
768 | set(b.rep[i+2*width]); |
---|
769 | TofftRep(R2, b, l+1, i+width, i+2*width); |
---|
770 | b.rep[i+2*width] = t1; |
---|
771 | mul(R1, R1, R2); |
---|
772 | FromfftRep(&b.rep[i], R1, 0, 2*width-1); |
---|
773 | sub(b.rep[i], b.rep[i], one); |
---|
774 | } |
---|
775 | } |
---|
776 | |
---|
777 | x.rep.SetLength(n+1); |
---|
778 | long delta = m-n; |
---|
779 | for (i = 0; i <= n; i++) |
---|
780 | x.rep[i] = b.rep[i+delta]; |
---|
781 | |
---|
782 | // no need to normalize |
---|
783 | } |
---|
784 | |
---|
785 | |
---|
786 | |
---|
787 | void eval(zz_p& b, const zz_pX& f, zz_p a) |
---|
788 | // does a Horner evaluation |
---|
789 | { |
---|
790 | zz_p acc; |
---|
791 | long i; |
---|
792 | |
---|
793 | clear(acc); |
---|
794 | for (i = deg(f); i >= 0; i--) { |
---|
795 | mul(acc, acc, a); |
---|
796 | add(acc, acc, f.rep[i]); |
---|
797 | } |
---|
798 | |
---|
799 | b = acc; |
---|
800 | } |
---|
801 | |
---|
802 | |
---|
803 | |
---|
804 | void eval(vec_zz_p& b, const zz_pX& f, const vec_zz_p& a) |
---|
805 | // naive algorithm: repeats Horner |
---|
806 | { |
---|
807 | if (&b == &f.rep) { |
---|
808 | vec_zz_p bb; |
---|
809 | eval(bb, f, a); |
---|
810 | b = bb; |
---|
811 | return; |
---|
812 | } |
---|
813 | |
---|
814 | long m = a.length(); |
---|
815 | b.SetLength(m); |
---|
816 | long i; |
---|
817 | for (i = 0; i < m; i++) |
---|
818 | eval(b[i], f, a[i]); |
---|
819 | } |
---|
820 | |
---|
821 | |
---|
822 | |
---|
823 | |
---|
824 | void interpolate(zz_pX& f, const vec_zz_p& a, const vec_zz_p& b) |
---|
825 | { |
---|
826 | long m = a.length(); |
---|
827 | if (b.length() != m) Error("interpolate: vector length mismatch"); |
---|
828 | |
---|
829 | if (m == 0) { |
---|
830 | clear(f); |
---|
831 | return; |
---|
832 | } |
---|
833 | |
---|
834 | vec_zz_p prod; |
---|
835 | prod = a; |
---|
836 | |
---|
837 | zz_p t1, t2; |
---|
838 | |
---|
839 | long k, i; |
---|
840 | |
---|
841 | vec_zz_p res; |
---|
842 | res.SetLength(m); |
---|
843 | |
---|
844 | for (k = 0; k < m; k++) { |
---|
845 | |
---|
846 | const zz_p& aa = a[k]; |
---|
847 | |
---|
848 | set(t1); |
---|
849 | for (i = k-1; i >= 0; i--) { |
---|
850 | mul(t1, t1, aa); |
---|
851 | add(t1, t1, prod[i]); |
---|
852 | } |
---|
853 | |
---|
854 | clear(t2); |
---|
855 | for (i = k-1; i >= 0; i--) { |
---|
856 | mul(t2, t2, aa); |
---|
857 | add(t2, t2, res[i]); |
---|
858 | } |
---|
859 | |
---|
860 | |
---|
861 | inv(t1, t1); |
---|
862 | sub(t2, b[k], t2); |
---|
863 | mul(t1, t1, t2); |
---|
864 | |
---|
865 | for (i = 0; i < k; i++) { |
---|
866 | mul(t2, prod[i], t1); |
---|
867 | add(res[i], res[i], t2); |
---|
868 | } |
---|
869 | |
---|
870 | res[k] = t1; |
---|
871 | |
---|
872 | if (k < m-1) { |
---|
873 | if (k == 0) |
---|
874 | negate(prod[0], prod[0]); |
---|
875 | else { |
---|
876 | negate(t1, a[k]); |
---|
877 | add(prod[k], t1, prod[k-1]); |
---|
878 | for (i = k-1; i >= 1; i--) { |
---|
879 | mul(t2, prod[i], t1); |
---|
880 | add(prod[i], t2, prod[i-1]); |
---|
881 | } |
---|
882 | mul(prod[0], prod[0], t1); |
---|
883 | } |
---|
884 | } |
---|
885 | } |
---|
886 | |
---|
887 | while (m > 0 && IsZero(res[m-1])) m--; |
---|
888 | res.SetLength(m); |
---|
889 | f.rep = res; |
---|
890 | } |
---|
891 | |
---|
892 | NTL_vector_impl(zz_pX,vec_zz_pX) |
---|
893 | |
---|
894 | NTL_eq_vector_impl(zz_pX,vec_zz_pX) |
---|
895 | |
---|
896 | |
---|
897 | |
---|
898 | |
---|
899 | void InnerProduct(zz_pX& x, const vec_zz_p& v, long low, long high, |
---|
900 | const vec_zz_pX& H, long n, vec_zz_p& t) |
---|
901 | { |
---|
902 | zz_p s; |
---|
903 | long i, j; |
---|
904 | |
---|
905 | zz_p *tp = t.elts(); |
---|
906 | |
---|
907 | for (j = 0; j < n; j++) |
---|
908 | clear(tp[j]); |
---|
909 | |
---|
910 | |
---|
911 | long p = zz_p::modulus(); |
---|
912 | double pinv = zz_p::ModulusInverse(); |
---|
913 | |
---|
914 | high = min(high, v.length()-1); |
---|
915 | for (i = low; i <= high; i++) { |
---|
916 | const vec_zz_p& h = H[i-low].rep; |
---|
917 | long m = h.length(); |
---|
918 | zz_p w = (v[i]); |
---|
919 | |
---|
920 | long W = rep(w); |
---|
921 | mulmod_precon_t Wpinv = PrepMulModPrecon(W, p, pinv); // ((double) W)*pinv; |
---|
922 | const zz_p *hp = h.elts(); |
---|
923 | |
---|
924 | for (j = 0; j < m; j++) { |
---|
925 | long S = MulModPrecon(rep(hp[j]), W, p, Wpinv); |
---|
926 | S = AddMod(S, rep(tp[j]), p); |
---|
927 | tp[j].LoopHole() = S; |
---|
928 | } |
---|
929 | } |
---|
930 | |
---|
931 | x.rep = t; |
---|
932 | x.normalize(); |
---|
933 | } |
---|
934 | |
---|
935 | |
---|
936 | void CompMod(zz_pX& x, const zz_pX& g, const zz_pXArgument& A, |
---|
937 | const zz_pXModulus& F) |
---|
938 | { |
---|
939 | if (deg(g) <= 0) { |
---|
940 | x = g; |
---|
941 | return; |
---|
942 | } |
---|
943 | |
---|
944 | |
---|
945 | zz_pX s, t; |
---|
946 | vec_zz_p scratch(INIT_SIZE, F.n); |
---|
947 | |
---|
948 | long m = A.H.length() - 1; |
---|
949 | long l = ((g.rep.length()+m-1)/m) - 1; |
---|
950 | |
---|
951 | zz_pXMultiplier M; |
---|
952 | build(M, A.H[m], F); |
---|
953 | |
---|
954 | InnerProduct(t, g.rep, l*m, l*m + m - 1, A.H, F.n, scratch); |
---|
955 | for (long i = l-1; i >= 0; i--) { |
---|
956 | InnerProduct(s, g.rep, i*m, i*m + m - 1, A.H, F.n, scratch); |
---|
957 | MulMod(t, t, M, F); |
---|
958 | add(t, t, s); |
---|
959 | } |
---|
960 | |
---|
961 | x = t; |
---|
962 | } |
---|
963 | |
---|
964 | |
---|
965 | void build(zz_pXArgument& A, const zz_pX& h, const zz_pXModulus& F, long m) |
---|
966 | { |
---|
967 | if (m <= 0 || deg(h) >= F.n) Error("build: bad args"); |
---|
968 | |
---|
969 | if (m > F.n) m = F.n; |
---|
970 | |
---|
971 | long i; |
---|
972 | |
---|
973 | if (zz_pXArgBound > 0) { |
---|
974 | double sz = 1; |
---|
975 | sz = sz*F.n; |
---|
976 | sz = sz+6; |
---|
977 | sz = sz*(sizeof (long)); |
---|
978 | sz = sz/1024; |
---|
979 | m = min(m, long(zz_pXArgBound/sz)); |
---|
980 | m = max(m, 1); |
---|
981 | } |
---|
982 | |
---|
983 | zz_pXMultiplier M; |
---|
984 | |
---|
985 | build(M, h, F); |
---|
986 | |
---|
987 | A.H.SetLength(m+1); |
---|
988 | |
---|
989 | set(A.H[0]); |
---|
990 | A.H[1] = h; |
---|
991 | for (i = 2; i <= m; i++) |
---|
992 | MulMod(A.H[i], A.H[i-1], M, F); |
---|
993 | } |
---|
994 | |
---|
995 | |
---|
996 | |
---|
997 | |
---|
998 | long zz_pXArgBound = 0; |
---|
999 | |
---|
1000 | |
---|
1001 | void CompMod(zz_pX& x, const zz_pX& g, const zz_pX& h, const zz_pXModulus& F) |
---|
1002 | // x = g(h) mod f |
---|
1003 | { |
---|
1004 | long m = SqrRoot(g.rep.length()); |
---|
1005 | |
---|
1006 | if (m == 0) { |
---|
1007 | clear(x); |
---|
1008 | return; |
---|
1009 | } |
---|
1010 | |
---|
1011 | zz_pXArgument A; |
---|
1012 | |
---|
1013 | build(A, h, F, m); |
---|
1014 | |
---|
1015 | CompMod(x, g, A, F); |
---|
1016 | } |
---|
1017 | |
---|
1018 | |
---|
1019 | |
---|
1020 | |
---|
1021 | void Comp2Mod(zz_pX& x1, zz_pX& x2, const zz_pX& g1, const zz_pX& g2, |
---|
1022 | const zz_pX& h, const zz_pXModulus& F) |
---|
1023 | |
---|
1024 | { |
---|
1025 | long m = SqrRoot(g1.rep.length() + g2.rep.length()); |
---|
1026 | |
---|
1027 | if (m == 0) { |
---|
1028 | clear(x1); |
---|
1029 | clear(x2); |
---|
1030 | return; |
---|
1031 | } |
---|
1032 | |
---|
1033 | zz_pXArgument A; |
---|
1034 | |
---|
1035 | build(A, h, F, m); |
---|
1036 | |
---|
1037 | zz_pX xx1, xx2; |
---|
1038 | |
---|
1039 | CompMod(xx1, g1, A, F); |
---|
1040 | CompMod(xx2, g2, A, F); |
---|
1041 | |
---|
1042 | x1 = xx1; |
---|
1043 | x2 = xx2; |
---|
1044 | } |
---|
1045 | |
---|
1046 | void Comp3Mod(zz_pX& x1, zz_pX& x2, zz_pX& x3, |
---|
1047 | const zz_pX& g1, const zz_pX& g2, const zz_pX& g3, |
---|
1048 | const zz_pX& h, const zz_pXModulus& F) |
---|
1049 | |
---|
1050 | { |
---|
1051 | long m = SqrRoot(g1.rep.length() + g2.rep.length() + g3.rep.length()); |
---|
1052 | |
---|
1053 | if (m == 0) { |
---|
1054 | clear(x1); |
---|
1055 | clear(x2); |
---|
1056 | clear(x3); |
---|
1057 | return; |
---|
1058 | } |
---|
1059 | |
---|
1060 | zz_pXArgument A; |
---|
1061 | |
---|
1062 | build(A, h, F, m); |
---|
1063 | |
---|
1064 | zz_pX xx1, xx2, xx3; |
---|
1065 | |
---|
1066 | CompMod(xx1, g1, A, F); |
---|
1067 | CompMod(xx2, g2, A, F); |
---|
1068 | CompMod(xx3, g3, A, F); |
---|
1069 | |
---|
1070 | x1 = xx1; |
---|
1071 | x2 = xx2; |
---|
1072 | x3 = xx3; |
---|
1073 | } |
---|
1074 | |
---|
1075 | static void StripZeroes(vec_zz_p& x) |
---|
1076 | { |
---|
1077 | long n = x.length(); |
---|
1078 | while (n > 0 && IsZero(x[n-1])) |
---|
1079 | n--; |
---|
1080 | x.SetLength(n); |
---|
1081 | } |
---|
1082 | |
---|
1083 | |
---|
1084 | void PlainUpdateMap(vec_zz_p& xx, const vec_zz_p& a, |
---|
1085 | const zz_pX& b, const zz_pX& f) |
---|
1086 | { |
---|
1087 | long n = deg(f); |
---|
1088 | long i, m; |
---|
1089 | |
---|
1090 | if (IsZero(b)) { |
---|
1091 | xx.SetLength(0); |
---|
1092 | return; |
---|
1093 | } |
---|
1094 | |
---|
1095 | m = n-1 - deg(b); |
---|
1096 | |
---|
1097 | vec_zz_p x(INIT_SIZE, n); |
---|
1098 | |
---|
1099 | for (i = 0; i <= m; i++) |
---|
1100 | InnerProduct(x[i], a, b.rep, i); |
---|
1101 | |
---|
1102 | if (deg(b) != 0) { |
---|
1103 | zz_pX c(INIT_SIZE, n); |
---|
1104 | LeftShift(c, b, m); |
---|
1105 | |
---|
1106 | for (i = m+1; i < n; i++) { |
---|
1107 | MulByXMod(c, c, f); |
---|
1108 | InnerProduct(x[i], a, c.rep); |
---|
1109 | } |
---|
1110 | } |
---|
1111 | |
---|
1112 | xx = x; |
---|
1113 | } |
---|
1114 | |
---|
1115 | |
---|
1116 | |
---|
1117 | |
---|
1118 | void UpdateMap(vec_zz_p& x, const vec_zz_p& aa, |
---|
1119 | const zz_pXMultiplier& B, const zz_pXModulus& F) |
---|
1120 | { |
---|
1121 | long n = F.n; |
---|
1122 | |
---|
1123 | vec_zz_p a; |
---|
1124 | a = aa; |
---|
1125 | StripZeroes(a); |
---|
1126 | |
---|
1127 | if (a.length() > n) Error("UpdateMap: bad args"); |
---|
1128 | long i; |
---|
1129 | |
---|
1130 | if (!B.UseFFT) { |
---|
1131 | PlainUpdateMap(x, a, B.b, F.f); |
---|
1132 | StripZeroes(x); |
---|
1133 | return; |
---|
1134 | } |
---|
1135 | |
---|
1136 | fftRep R1(INIT_SIZE, F.k), R2(INIT_SIZE, F.l); |
---|
1137 | vec_zz_p V1(INIT_SIZE, n); |
---|
1138 | |
---|
1139 | |
---|
1140 | RevTofftRep(R1, a, F.k, 0, a.length()-1, 0); |
---|
1141 | mul(R2, R1, F.FRep); |
---|
1142 | RevFromfftRep(V1, R2, 0, n-2); |
---|
1143 | for (i = 0; i <= n-2; i++) negate(V1[i], V1[i]); |
---|
1144 | RevTofftRep(R2, V1, F.l, 0, n-2, n-1); |
---|
1145 | mul(R2, R2, B.B1); |
---|
1146 | mul(R1, R1, B.B2); |
---|
1147 | |
---|
1148 | AddExpand(R2, R1); |
---|
1149 | RevFromfftRep(x, R2, 0, n-1); |
---|
1150 | StripZeroes(x); |
---|
1151 | } |
---|
1152 | |
---|
1153 | |
---|
1154 | |
---|
1155 | void ProjectPowers(vec_zz_p& x, const vec_zz_p& a, long k, |
---|
1156 | const zz_pXArgument& H, const zz_pXModulus& F) |
---|
1157 | |
---|
1158 | { |
---|
1159 | long n = F.n; |
---|
1160 | |
---|
1161 | if (a.length() > n || k < 0 || NTL_OVERFLOW(k, 1, 0)) |
---|
1162 | Error("ProjectPowers: bad args"); |
---|
1163 | |
---|
1164 | long m = H.H.length()-1; |
---|
1165 | long l = (k+m-1)/m - 1; |
---|
1166 | |
---|
1167 | zz_pXMultiplier M; |
---|
1168 | build(M, H.H[m], F); |
---|
1169 | |
---|
1170 | vec_zz_p s(INIT_SIZE, n); |
---|
1171 | s = a; |
---|
1172 | StripZeroes(s); |
---|
1173 | |
---|
1174 | x.SetLength(k); |
---|
1175 | |
---|
1176 | for (long i = 0; i <= l; i++) { |
---|
1177 | long m1 = min(m, k-i*m); |
---|
1178 | zz_p* w = &x[i*m]; |
---|
1179 | for (long j = 0; j < m1; j++) |
---|
1180 | InnerProduct(w[j], H.H[j].rep, s); |
---|
1181 | if (i < l) |
---|
1182 | UpdateMap(s, s, M, F); |
---|
1183 | } |
---|
1184 | } |
---|
1185 | |
---|
1186 | |
---|
1187 | |
---|
1188 | void ProjectPowers(vec_zz_p& x, const vec_zz_p& a, long k, |
---|
1189 | const zz_pX& h, const zz_pXModulus& F) |
---|
1190 | |
---|
1191 | { |
---|
1192 | if (a.length() > F.n || k < 0) Error("ProjectPowers: bad args"); |
---|
1193 | |
---|
1194 | if (k == 0) { |
---|
1195 | x.SetLength(0); |
---|
1196 | return; |
---|
1197 | } |
---|
1198 | |
---|
1199 | long m = SqrRoot(k); |
---|
1200 | |
---|
1201 | zz_pXArgument H; |
---|
1202 | |
---|
1203 | build(H, h, F, m); |
---|
1204 | ProjectPowers(x, a, k, H, F); |
---|
1205 | } |
---|
1206 | |
---|
1207 | |
---|
1208 | void BerlekampMassey(zz_pX& h, const vec_zz_p& a, long m) |
---|
1209 | { |
---|
1210 | zz_pX Lambda, Sigma, Temp; |
---|
1211 | long L; |
---|
1212 | zz_p Delta, Delta1, t1; |
---|
1213 | long shamt; |
---|
1214 | |
---|
1215 | // cerr << "*** " << m << "\n"; |
---|
1216 | |
---|
1217 | Lambda.SetMaxLength(m+1); |
---|
1218 | Sigma.SetMaxLength(m+1); |
---|
1219 | Temp.SetMaxLength(m+1); |
---|
1220 | |
---|
1221 | L = 0; |
---|
1222 | set(Lambda); |
---|
1223 | clear(Sigma); |
---|
1224 | set(Delta); |
---|
1225 | shamt = 0; |
---|
1226 | |
---|
1227 | long i, r, dl; |
---|
1228 | |
---|
1229 | for (r = 1; r <= 2*m; r++) { |
---|
1230 | // cerr << r << "--"; |
---|
1231 | clear(Delta1); |
---|
1232 | dl = deg(Lambda); |
---|
1233 | for (i = 0; i <= dl; i++) { |
---|
1234 | mul(t1, Lambda.rep[i], a[r-i-1]); |
---|
1235 | add(Delta1, Delta1, t1); |
---|
1236 | } |
---|
1237 | |
---|
1238 | if (IsZero(Delta1)) { |
---|
1239 | shamt++; |
---|
1240 | // cerr << "case 1: " << deg(Lambda) << " " << deg(Sigma) << " " << shamt << "\n"; |
---|
1241 | } |
---|
1242 | else if (2*L < r) { |
---|
1243 | div(t1, Delta1, Delta); |
---|
1244 | mul(Temp, Sigma, t1); |
---|
1245 | Sigma = Lambda; |
---|
1246 | ShiftSub(Lambda, Temp, shamt+1); |
---|
1247 | shamt = 0; |
---|
1248 | L = r-L; |
---|
1249 | Delta = Delta1; |
---|
1250 | // cerr << "case 2: " << deg(Lambda) << " " << deg(Sigma) << " " << shamt << "\n"; |
---|
1251 | } |
---|
1252 | else { |
---|
1253 | shamt++; |
---|
1254 | div(t1, Delta1, Delta); |
---|
1255 | mul(Temp, Sigma, t1); |
---|
1256 | ShiftSub(Lambda, Temp, shamt); |
---|
1257 | // cerr << "case 3: " << deg(Lambda) << " " << deg(Sigma) << " " << shamt << "\n"; |
---|
1258 | } |
---|
1259 | } |
---|
1260 | |
---|
1261 | // cerr << "finished: " << L << " " << deg(Lambda) << "\n"; |
---|
1262 | |
---|
1263 | dl = deg(Lambda); |
---|
1264 | h.rep.SetLength(L + 1); |
---|
1265 | |
---|
1266 | for (i = 0; i < L - dl; i++) |
---|
1267 | clear(h.rep[i]); |
---|
1268 | |
---|
1269 | for (i = L - dl; i <= L; i++) |
---|
1270 | h.rep[i] = Lambda.rep[L - i]; |
---|
1271 | } |
---|
1272 | |
---|
1273 | |
---|
1274 | void GCDMinPolySeq(zz_pX& h, const vec_zz_p& x, long m) |
---|
1275 | { |
---|
1276 | long i; |
---|
1277 | zz_pX a, b; |
---|
1278 | zz_pXMatrix M; |
---|
1279 | zz_p t; |
---|
1280 | |
---|
1281 | a.rep.SetLength(2*m); |
---|
1282 | for (i = 0; i < 2*m; i++) a.rep[i] = x[2*m-1-i]; |
---|
1283 | a.normalize(); |
---|
1284 | |
---|
1285 | SetCoeff(b, 2*m); |
---|
1286 | |
---|
1287 | HalfGCD(M, b, a, m+1); |
---|
1288 | |
---|
1289 | /* make monic */ |
---|
1290 | |
---|
1291 | inv(t, LeadCoeff(M(1,1))); |
---|
1292 | mul(h, M(1,1), t); |
---|
1293 | } |
---|
1294 | |
---|
1295 | |
---|
1296 | void MinPolySeq(zz_pX& h, const vec_zz_p& a, long m) |
---|
1297 | { |
---|
1298 | if (m < 0 || NTL_OVERFLOW(m, 1, 0)) Error("MinPoly: bad args"); |
---|
1299 | if (a.length() < 2*m) Error("MinPoly: sequence too short"); |
---|
1300 | |
---|
1301 | if (m > NTL_zz_pX_BERMASS_CROSSOVER) |
---|
1302 | GCDMinPolySeq(h, a, m); |
---|
1303 | else |
---|
1304 | BerlekampMassey(h, a, m); |
---|
1305 | } |
---|
1306 | |
---|
1307 | |
---|
1308 | void DoMinPolyMod(zz_pX& h, const zz_pX& g, const zz_pXModulus& F, long m, |
---|
1309 | const vec_zz_p& R) |
---|
1310 | { |
---|
1311 | vec_zz_p x; |
---|
1312 | |
---|
1313 | ProjectPowers(x, R, 2*m, g, F); |
---|
1314 | MinPolySeq(h, x, m); |
---|
1315 | } |
---|
1316 | |
---|
1317 | |
---|
1318 | void ProbMinPolyMod(zz_pX& h, const zz_pX& g, const zz_pXModulus& F, long m) |
---|
1319 | { |
---|
1320 | long n = F.n; |
---|
1321 | if (m < 1 || m > n) Error("ProbMinPoly: bad args"); |
---|
1322 | |
---|
1323 | long i; |
---|
1324 | vec_zz_p R(INIT_SIZE, n); |
---|
1325 | |
---|
1326 | for (i = 0; i < n; i++) random(R[i]); |
---|
1327 | DoMinPolyMod(h, g, F, m, R); |
---|
1328 | } |
---|
1329 | |
---|
1330 | void MinPolyMod(zz_pX& hh, const zz_pX& g, const zz_pXModulus& F, long m) |
---|
1331 | { |
---|
1332 | zz_pX h, h1; |
---|
1333 | long n = F.n; |
---|
1334 | if (m < 1 || m > n) Error("MinPoly: bad args"); |
---|
1335 | |
---|
1336 | /* probabilistically compute min-poly */ |
---|
1337 | |
---|
1338 | ProbMinPolyMod(h, g, F, m); |
---|
1339 | if (deg(h) == m) { hh = h; return; } |
---|
1340 | CompMod(h1, h, g, F); |
---|
1341 | if (IsZero(h1)) { hh = h; return; } |
---|
1342 | |
---|
1343 | /* not completely successful...must iterate */ |
---|
1344 | |
---|
1345 | long i; |
---|
1346 | |
---|
1347 | zz_pX h2, h3; |
---|
1348 | zz_pXMultiplier H1; |
---|
1349 | vec_zz_p R(INIT_SIZE, n); |
---|
1350 | |
---|
1351 | for (;;) { |
---|
1352 | R.SetLength(n); |
---|
1353 | for (i = 0; i < n; i++) random(R[i]); |
---|
1354 | build(H1, h1, F); |
---|
1355 | UpdateMap(R, R, H1, F); |
---|
1356 | DoMinPolyMod(h2, g, F, m-deg(h), R); |
---|
1357 | |
---|
1358 | mul(h, h, h2); |
---|
1359 | if (deg(h) == m) { hh = h; return; } |
---|
1360 | CompMod(h3, h2, g, F); |
---|
1361 | MulMod(h1, h3, H1, F); |
---|
1362 | if (IsZero(h1)) { hh = h; return; } |
---|
1363 | } |
---|
1364 | } |
---|
1365 | |
---|
1366 | void IrredPolyMod(zz_pX& h, const zz_pX& g, const zz_pXModulus& F, long m) |
---|
1367 | { |
---|
1368 | vec_zz_p R(INIT_SIZE, 1); |
---|
1369 | if (m < 1 || m > F.n) Error("IrredPoly: bad args"); |
---|
1370 | |
---|
1371 | set(R[0]); |
---|
1372 | DoMinPolyMod(h, g, F, m, R); |
---|
1373 | } |
---|
1374 | |
---|
1375 | |
---|
1376 | |
---|
1377 | void diff(zz_pX& x, const zz_pX& a) |
---|
1378 | { |
---|
1379 | long n = deg(a); |
---|
1380 | long i; |
---|
1381 | |
---|
1382 | if (n <= 0) { |
---|
1383 | clear(x); |
---|
1384 | return; |
---|
1385 | } |
---|
1386 | |
---|
1387 | if (&x != &a) |
---|
1388 | x.rep.SetLength(n); |
---|
1389 | |
---|
1390 | for (i = 0; i <= n-1; i++) { |
---|
1391 | mul(x.rep[i], a.rep[i+1], i+1); |
---|
1392 | } |
---|
1393 | |
---|
1394 | if (&x == &a) |
---|
1395 | x.rep.SetLength(n); |
---|
1396 | |
---|
1397 | x.normalize(); |
---|
1398 | } |
---|
1399 | |
---|
1400 | void MakeMonic(zz_pX& x) |
---|
1401 | { |
---|
1402 | if (IsZero(x)) |
---|
1403 | return; |
---|
1404 | |
---|
1405 | if (IsOne(LeadCoeff(x))) |
---|
1406 | return; |
---|
1407 | |
---|
1408 | zz_p t; |
---|
1409 | |
---|
1410 | inv(t, LeadCoeff(x)); |
---|
1411 | mul(x, x, t); |
---|
1412 | } |
---|
1413 | |
---|
1414 | |
---|
1415 | |
---|
1416 | |
---|
1417 | |
---|
1418 | void PlainMulTrunc(zz_pX& x, const zz_pX& a, const zz_pX& b, long n) |
---|
1419 | { |
---|
1420 | zz_pX y; |
---|
1421 | mul(y, a, b); |
---|
1422 | trunc(x, y, n); |
---|
1423 | } |
---|
1424 | |
---|
1425 | |
---|
1426 | void FFTMulTrunc(zz_pX& x, const zz_pX& a, const zz_pX& b, long n) |
---|
1427 | { |
---|
1428 | if (IsZero(a) || IsZero(b)) { |
---|
1429 | clear(x); |
---|
1430 | return; |
---|
1431 | } |
---|
1432 | |
---|
1433 | long d = deg(a) + deg(b); |
---|
1434 | if (n > d + 1) |
---|
1435 | n = d + 1; |
---|
1436 | |
---|
1437 | long k = NextPowerOfTwo(d + 1); |
---|
1438 | fftRep R1(INIT_SIZE, k), R2(INIT_SIZE, k); |
---|
1439 | |
---|
1440 | TofftRep(R1, a, k); |
---|
1441 | TofftRep(R2, b, k); |
---|
1442 | mul(R1, R1, R2); |
---|
1443 | FromfftRep(x, R1, 0, n-1); |
---|
1444 | } |
---|
1445 | |
---|
1446 | void MulTrunc(zz_pX& x, const zz_pX& a, const zz_pX& b, long n) |
---|
1447 | { |
---|
1448 | if (n < 0) Error("MulTrunc: bad args"); |
---|
1449 | |
---|
1450 | if (deg(a) <= NTL_zz_pX_MUL_CROSSOVER || deg(b) <= NTL_zz_pX_MUL_CROSSOVER) |
---|
1451 | PlainMulTrunc(x, a, b, n); |
---|
1452 | else |
---|
1453 | FFTMulTrunc(x, a, b, n); |
---|
1454 | } |
---|
1455 | |
---|
1456 | void PlainSqrTrunc(zz_pX& x, const zz_pX& a, long n) |
---|
1457 | { |
---|
1458 | zz_pX y; |
---|
1459 | sqr(y, a); |
---|
1460 | trunc(x, y, n); |
---|
1461 | } |
---|
1462 | |
---|
1463 | |
---|
1464 | void FFTSqrTrunc(zz_pX& x, const zz_pX& a, long n) |
---|
1465 | { |
---|
1466 | if (IsZero(a)) { |
---|
1467 | clear(x); |
---|
1468 | return; |
---|
1469 | } |
---|
1470 | |
---|
1471 | long d = 2*deg(a); |
---|
1472 | if (n > d + 1) |
---|
1473 | n = d + 1; |
---|
1474 | |
---|
1475 | long k = NextPowerOfTwo(d + 1); |
---|
1476 | fftRep R1(INIT_SIZE, k); |
---|
1477 | |
---|
1478 | TofftRep(R1, a, k); |
---|
1479 | mul(R1, R1, R1); |
---|
1480 | FromfftRep(x, R1, 0, n-1); |
---|
1481 | } |
---|
1482 | |
---|
1483 | void SqrTrunc(zz_pX& x, const zz_pX& a, long n) |
---|
1484 | { |
---|
1485 | if (n < 0) Error("SqrTrunc: bad args"); |
---|
1486 | |
---|
1487 | if (deg(a) <= NTL_zz_pX_MUL_CROSSOVER) |
---|
1488 | PlainSqrTrunc(x, a, n); |
---|
1489 | else |
---|
1490 | FFTSqrTrunc(x, a, n); |
---|
1491 | } |
---|
1492 | |
---|
1493 | |
---|
1494 | |
---|
1495 | void FastTraceVec(vec_zz_p& S, const zz_pX& f) |
---|
1496 | { |
---|
1497 | long n = deg(f); |
---|
1498 | |
---|
1499 | if (n <= 0) |
---|
1500 | Error("FastTraceVec: bad args"); |
---|
1501 | |
---|
1502 | if (n == 0) { |
---|
1503 | S.SetLength(0); |
---|
1504 | return; |
---|
1505 | } |
---|
1506 | |
---|
1507 | if (n == 1) { |
---|
1508 | S.SetLength(1); |
---|
1509 | set(S[0]); |
---|
1510 | return; |
---|
1511 | } |
---|
1512 | |
---|
1513 | long i; |
---|
1514 | zz_pX f1; |
---|
1515 | |
---|
1516 | f1.rep.SetLength(n-1); |
---|
1517 | for (i = 0; i <= n-2; i++) |
---|
1518 | f1.rep[i] = f.rep[n-i]; |
---|
1519 | f1.normalize(); |
---|
1520 | |
---|
1521 | zz_pX f2; |
---|
1522 | f2.rep.SetLength(n-1); |
---|
1523 | for (i = 0; i <= n-2; i++) |
---|
1524 | mul(f2.rep[i], f.rep[n-1-i], i+1); |
---|
1525 | f2.normalize(); |
---|
1526 | |
---|
1527 | zz_pX f3; |
---|
1528 | InvTrunc(f3, f1, n-1); |
---|
1529 | MulTrunc(f3, f3, f2, n-1); |
---|
1530 | |
---|
1531 | S.SetLength(n); |
---|
1532 | |
---|
1533 | S[0] = n; |
---|
1534 | for (i = 1; i < n; i++) |
---|
1535 | negate(S[i], coeff(f3, i-1)); |
---|
1536 | } |
---|
1537 | |
---|
1538 | |
---|
1539 | void PlainTraceVec(vec_zz_p& S, const zz_pX& ff) |
---|
1540 | { |
---|
1541 | if (deg(ff) <= 0) |
---|
1542 | Error("TraceVec: bad args"); |
---|
1543 | |
---|
1544 | zz_pX f; |
---|
1545 | f = ff; |
---|
1546 | |
---|
1547 | MakeMonic(f); |
---|
1548 | |
---|
1549 | long n = deg(f); |
---|
1550 | |
---|
1551 | S.SetLength(n); |
---|
1552 | |
---|
1553 | if (n == 0) |
---|
1554 | return; |
---|
1555 | |
---|
1556 | long k, i; |
---|
1557 | zz_p acc, t; |
---|
1558 | |
---|
1559 | const zz_p *fp = f.rep.elts();; |
---|
1560 | zz_p *sp = S.elts(); |
---|
1561 | |
---|
1562 | sp[0] = n; |
---|
1563 | |
---|
1564 | for (k = 1; k < n; k++) { |
---|
1565 | mul(acc, fp[n-k], k); |
---|
1566 | |
---|
1567 | for (i = 1; i < k; i++) { |
---|
1568 | mul(t, fp[n-i], rep(sp[k-i])); |
---|
1569 | add(acc, acc, t); |
---|
1570 | } |
---|
1571 | |
---|
1572 | negate(sp[k], acc); |
---|
1573 | } |
---|
1574 | } |
---|
1575 | |
---|
1576 | void TraceVec(vec_zz_p& S, const zz_pX& f) |
---|
1577 | { |
---|
1578 | if (deg(f) <= NTL_zz_pX_TRACE_CROSSOVER) |
---|
1579 | PlainTraceVec(S, f); |
---|
1580 | else |
---|
1581 | FastTraceVec(S, f); |
---|
1582 | } |
---|
1583 | |
---|
1584 | void ComputeTraceVec(const zz_pXModulus& F) |
---|
1585 | { |
---|
1586 | vec_zz_p& S = *((vec_zz_p *) &F.tracevec); |
---|
1587 | |
---|
1588 | if (S.length() > 0) |
---|
1589 | return; |
---|
1590 | |
---|
1591 | if (!F.UseFFT) { |
---|
1592 | PlainTraceVec(S, F.f); |
---|
1593 | return; |
---|
1594 | } |
---|
1595 | |
---|
1596 | long i; |
---|
1597 | long n = F.n; |
---|
1598 | |
---|
1599 | fftRep R; |
---|
1600 | zz_pX P, g; |
---|
1601 | |
---|
1602 | g.rep.SetLength(n-1); |
---|
1603 | for (i = 1; i < n; i++) |
---|
1604 | mul(g.rep[n-i-1], F.f.rep[n-i], i); |
---|
1605 | g.normalize(); |
---|
1606 | |
---|
1607 | TofftRep(R, g, F.l); |
---|
1608 | mul(R, R, F.HRep); |
---|
1609 | FromfftRep(P, R, n-2, 2*n-4); |
---|
1610 | |
---|
1611 | S.SetLength(n); |
---|
1612 | |
---|
1613 | S[0] = n; |
---|
1614 | for (i = 1; i < n; i++) |
---|
1615 | negate(S[i], coeff(P, n-1-i)); |
---|
1616 | } |
---|
1617 | |
---|
1618 | void TraceMod(zz_p& x, const zz_pX& a, const zz_pXModulus& F) |
---|
1619 | { |
---|
1620 | long n = F.n; |
---|
1621 | |
---|
1622 | if (deg(a) >= n) |
---|
1623 | Error("trace: bad args"); |
---|
1624 | |
---|
1625 | if (F.tracevec.length() == 0) |
---|
1626 | TraceVec(F); |
---|
1627 | |
---|
1628 | InnerProduct(x, a.rep, F.tracevec); |
---|
1629 | } |
---|
1630 | |
---|
1631 | |
---|
1632 | void TraceMod(zz_p& x, const zz_pX& a, const zz_pX& f) |
---|
1633 | { |
---|
1634 | if (deg(a) >= deg(f) || deg(f) <= 0) |
---|
1635 | Error("trace: bad args"); |
---|
1636 | |
---|
1637 | project(x, TraceVec(f), a); |
---|
1638 | } |
---|
1639 | |
---|
1640 | |
---|
1641 | void PlainResultant(zz_p& rres, const zz_pX& a, const zz_pX& b) |
---|
1642 | { |
---|
1643 | zz_p res; |
---|
1644 | |
---|
1645 | if (IsZero(a) || IsZero(b)) |
---|
1646 | clear(res); |
---|
1647 | else if (deg(a) == 0 && deg(b) == 0) |
---|
1648 | set(res); |
---|
1649 | else { |
---|
1650 | long d0, d1, d2; |
---|
1651 | zz_p lc; |
---|
1652 | set(res); |
---|
1653 | |
---|
1654 | long n = max(deg(a),deg(b)) + 1; |
---|
1655 | zz_pX u(INIT_SIZE, n), v(INIT_SIZE, n); |
---|
1656 | |
---|
1657 | u = a; |
---|
1658 | v = b; |
---|
1659 | |
---|
1660 | for (;;) { |
---|
1661 | d0 = deg(u); |
---|
1662 | d1 = deg(v); |
---|
1663 | lc = LeadCoeff(v); |
---|
1664 | |
---|
1665 | PlainRem(u, u, v); |
---|
1666 | swap(u, v); |
---|
1667 | |
---|
1668 | d2 = deg(v); |
---|
1669 | if (d2 >= 0) { |
---|
1670 | power(lc, lc, d0-d2); |
---|
1671 | mul(res, res, lc); |
---|
1672 | if (d0 & d1 & 1) negate(res, res); |
---|
1673 | } |
---|
1674 | else { |
---|
1675 | if (d1 == 0) { |
---|
1676 | power(lc, lc, d0); |
---|
1677 | mul(res, res, lc); |
---|
1678 | } |
---|
1679 | else |
---|
1680 | clear(res); |
---|
1681 | |
---|
1682 | break; |
---|
1683 | } |
---|
1684 | } |
---|
1685 | |
---|
1686 | rres = res; |
---|
1687 | } |
---|
1688 | } |
---|
1689 | |
---|
1690 | |
---|
1691 | void ResIterHalfGCD(zz_pXMatrix& M_out, zz_pX& U, zz_pX& V, long d_red, |
---|
1692 | vec_zz_p& cvec, vec_long& dvec) |
---|
1693 | { |
---|
1694 | M_out(0,0).SetMaxLength(d_red); |
---|
1695 | M_out(0,1).SetMaxLength(d_red); |
---|
1696 | M_out(1,0).SetMaxLength(d_red); |
---|
1697 | M_out(1,1).SetMaxLength(d_red); |
---|
1698 | |
---|
1699 | set(M_out(0,0)); clear(M_out(0,1)); |
---|
1700 | clear(M_out(1,0)); set(M_out(1,1)); |
---|
1701 | |
---|
1702 | long goal = deg(U) - d_red; |
---|
1703 | |
---|
1704 | if (deg(V) <= goal) |
---|
1705 | return; |
---|
1706 | |
---|
1707 | zz_pX Q, t(INIT_SIZE, d_red); |
---|
1708 | |
---|
1709 | |
---|
1710 | while (deg(V) > goal) { |
---|
1711 | append(cvec, LeadCoeff(V)); |
---|
1712 | append(dvec, dvec[dvec.length()-1]-deg(U)+deg(V)); |
---|
1713 | PlainDivRem(Q, U, U, V); |
---|
1714 | swap(U, V); |
---|
1715 | |
---|
1716 | mul(t, Q, M_out(1,0)); |
---|
1717 | sub(t, M_out(0,0), t); |
---|
1718 | M_out(0,0) = M_out(1,0); |
---|
1719 | M_out(1,0) = t; |
---|
1720 | |
---|
1721 | mul(t, Q, M_out(1,1)); |
---|
1722 | sub(t, M_out(0,1), t); |
---|
1723 | M_out(0,1) = M_out(1,1); |
---|
1724 | M_out(1,1) = t; |
---|
1725 | } |
---|
1726 | } |
---|
1727 | |
---|
1728 | |
---|
1729 | |
---|
1730 | void ResHalfGCD(zz_pXMatrix& M_out, const zz_pX& U, const zz_pX& V, long d_red, |
---|
1731 | vec_zz_p& cvec, vec_long& dvec) |
---|
1732 | { |
---|
1733 | if (IsZero(V) || deg(V) <= deg(U) - d_red) { |
---|
1734 | set(M_out(0,0)); clear(M_out(0,1)); |
---|
1735 | clear(M_out(1,0)); set(M_out(1,1)); |
---|
1736 | |
---|
1737 | return; |
---|
1738 | } |
---|
1739 | |
---|
1740 | |
---|
1741 | long n = deg(U) - 2*d_red + 2; |
---|
1742 | if (n < 0) n = 0; |
---|
1743 | |
---|
1744 | zz_pX U1, V1; |
---|
1745 | |
---|
1746 | RightShift(U1, U, n); |
---|
1747 | RightShift(V1, V, n); |
---|
1748 | |
---|
1749 | if (d_red <= NTL_zz_pX_HalfGCD_CROSSOVER) { |
---|
1750 | ResIterHalfGCD(M_out, U1, V1, d_red, cvec, dvec); |
---|
1751 | return; |
---|
1752 | } |
---|
1753 | |
---|
1754 | long d1 = (d_red + 1)/2; |
---|
1755 | if (d1 < 1) d1 = 1; |
---|
1756 | if (d1 >= d_red) d1 = d_red - 1; |
---|
1757 | |
---|
1758 | zz_pXMatrix M1; |
---|
1759 | |
---|
1760 | ResHalfGCD(M1, U1, V1, d1, cvec, dvec); |
---|
1761 | mul(U1, V1, M1); |
---|
1762 | |
---|
1763 | long d2 = deg(V1) - deg(U) + n + d_red; |
---|
1764 | |
---|
1765 | if (IsZero(V1) || d2 <= 0) { |
---|
1766 | M_out = M1; |
---|
1767 | return; |
---|
1768 | } |
---|
1769 | |
---|
1770 | |
---|
1771 | zz_pX Q; |
---|
1772 | zz_pXMatrix M2; |
---|
1773 | |
---|
1774 | append(cvec, LeadCoeff(V1)); |
---|
1775 | append(dvec, dvec[dvec.length()-1]-deg(U1)+deg(V1)); |
---|
1776 | DivRem(Q, U1, U1, V1); |
---|
1777 | swap(U1, V1); |
---|
1778 | |
---|
1779 | ResHalfGCD(M2, U1, V1, d2, cvec, dvec); |
---|
1780 | |
---|
1781 | zz_pX t(INIT_SIZE, deg(M1(1,1))+deg(Q)+1); |
---|
1782 | |
---|
1783 | mul(t, Q, M1(1,0)); |
---|
1784 | sub(t, M1(0,0), t); |
---|
1785 | swap(M1(0,0), M1(1,0)); |
---|
1786 | swap(M1(1,0), t); |
---|
1787 | |
---|
1788 | t.kill(); |
---|
1789 | |
---|
1790 | t.SetMaxLength(deg(M1(1,1))+deg(Q)+1); |
---|
1791 | |
---|
1792 | mul(t, Q, M1(1,1)); |
---|
1793 | sub(t, M1(0,1), t); |
---|
1794 | swap(M1(0,1), M1(1,1)); |
---|
1795 | swap(M1(1,1), t); |
---|
1796 | |
---|
1797 | t.kill(); |
---|
1798 | |
---|
1799 | mul(M_out, M2, M1); |
---|
1800 | } |
---|
1801 | |
---|
1802 | void ResHalfGCD(zz_pX& U, zz_pX& V, vec_zz_p& cvec, vec_long& dvec) |
---|
1803 | { |
---|
1804 | long d_red = (deg(U)+1)/2; |
---|
1805 | |
---|
1806 | if (IsZero(V) || deg(V) <= deg(U) - d_red) { |
---|
1807 | return; |
---|
1808 | } |
---|
1809 | |
---|
1810 | long du = deg(U); |
---|
1811 | |
---|
1812 | |
---|
1813 | long d1 = (d_red + 1)/2; |
---|
1814 | if (d1 < 1) d1 = 1; |
---|
1815 | if (d1 >= d_red) d1 = d_red - 1; |
---|
1816 | |
---|
1817 | zz_pXMatrix M1; |
---|
1818 | |
---|
1819 | ResHalfGCD(M1, U, V, d1, cvec, dvec); |
---|
1820 | mul(U, V, M1); |
---|
1821 | |
---|
1822 | long d2 = deg(V) - du + d_red; |
---|
1823 | |
---|
1824 | if (IsZero(V) || d2 <= 0) { |
---|
1825 | return; |
---|
1826 | } |
---|
1827 | |
---|
1828 | M1(0,0).kill(); |
---|
1829 | M1(0,1).kill(); |
---|
1830 | M1(1,0).kill(); |
---|
1831 | M1(1,1).kill(); |
---|
1832 | |
---|
1833 | |
---|
1834 | zz_pX Q; |
---|
1835 | |
---|
1836 | append(cvec, LeadCoeff(V)); |
---|
1837 | append(dvec, dvec[dvec.length()-1]-deg(U)+deg(V)); |
---|
1838 | DivRem(Q, U, U, V); |
---|
1839 | swap(U, V); |
---|
1840 | |
---|
1841 | ResHalfGCD(M1, U, V, d2, cvec, dvec); |
---|
1842 | |
---|
1843 | mul(U, V, M1); |
---|
1844 | } |
---|
1845 | |
---|
1846 | |
---|
1847 | void resultant(zz_p& rres, const zz_pX& u, const zz_pX& v) |
---|
1848 | { |
---|
1849 | if (deg(u) <= NTL_zz_pX_GCD_CROSSOVER || deg(v) <= NTL_zz_pX_GCD_CROSSOVER) { |
---|
1850 | PlainResultant(rres, u, v); |
---|
1851 | return; |
---|
1852 | } |
---|
1853 | |
---|
1854 | zz_pX u1, v1; |
---|
1855 | |
---|
1856 | u1 = u; |
---|
1857 | v1 = v; |
---|
1858 | |
---|
1859 | zz_p res, t; |
---|
1860 | set(res); |
---|
1861 | |
---|
1862 | if (deg(u1) == deg(v1)) { |
---|
1863 | rem(u1, u1, v1); |
---|
1864 | swap(u1, v1); |
---|
1865 | |
---|
1866 | if (IsZero(v1)) { |
---|
1867 | clear(rres); |
---|
1868 | return; |
---|
1869 | } |
---|
1870 | |
---|
1871 | power(t, LeadCoeff(u1), deg(u1) - deg(v1)); |
---|
1872 | mul(res, res, t); |
---|
1873 | if (deg(u1) & 1) |
---|
1874 | negate(res, res); |
---|
1875 | } |
---|
1876 | else if (deg(u1) < deg(v1)) { |
---|
1877 | swap(u1, v1); |
---|
1878 | if (deg(u1) & deg(v1) & 1) |
---|
1879 | negate(res, res); |
---|
1880 | } |
---|
1881 | |
---|
1882 | // deg(u1) > deg(v1) && v1 != 0 |
---|
1883 | |
---|
1884 | vec_zz_p cvec; |
---|
1885 | vec_long dvec; |
---|
1886 | |
---|
1887 | cvec.SetMaxLength(deg(v1)+2); |
---|
1888 | dvec.SetMaxLength(deg(v1)+2); |
---|
1889 | |
---|
1890 | append(cvec, LeadCoeff(u1)); |
---|
1891 | append(dvec, deg(u1)); |
---|
1892 | |
---|
1893 | |
---|
1894 | while (deg(u1) > NTL_zz_pX_GCD_CROSSOVER && !IsZero(v1)) { |
---|
1895 | ResHalfGCD(u1, v1, cvec, dvec); |
---|
1896 | |
---|
1897 | if (!IsZero(v1)) { |
---|
1898 | append(cvec, LeadCoeff(v1)); |
---|
1899 | append(dvec, deg(v1)); |
---|
1900 | rem(u1, u1, v1); |
---|
1901 | swap(u1, v1); |
---|
1902 | } |
---|
1903 | } |
---|
1904 | |
---|
1905 | if (IsZero(v1) && deg(u1) > 0) { |
---|
1906 | clear(rres); |
---|
1907 | return; |
---|
1908 | } |
---|
1909 | |
---|
1910 | long i, l; |
---|
1911 | l = dvec.length(); |
---|
1912 | |
---|
1913 | if (deg(u1) == 0) { |
---|
1914 | // we went all the way... |
---|
1915 | |
---|
1916 | for (i = 0; i <= l-3; i++) { |
---|
1917 | power(t, cvec[i+1], dvec[i]-dvec[i+2]); |
---|
1918 | mul(res, res, t); |
---|
1919 | if (dvec[i] & dvec[i+1] & 1) |
---|
1920 | negate(res, res); |
---|
1921 | } |
---|
1922 | |
---|
1923 | power(t, cvec[l-1], dvec[l-2]); |
---|
1924 | mul(res, res, t); |
---|
1925 | } |
---|
1926 | else { |
---|
1927 | for (i = 0; i <= l-3; i++) { |
---|
1928 | power(t, cvec[i+1], dvec[i]-dvec[i+2]); |
---|
1929 | mul(res, res, t); |
---|
1930 | if (dvec[i] & dvec[i+1] & 1) |
---|
1931 | negate(res, res); |
---|
1932 | } |
---|
1933 | |
---|
1934 | power(t, cvec[l-1], dvec[l-2]-deg(v1)); |
---|
1935 | mul(res, res, t); |
---|
1936 | if (dvec[l-2] & dvec[l-1] & 1) |
---|
1937 | negate(res, res); |
---|
1938 | |
---|
1939 | PlainResultant(t, u1, v1); |
---|
1940 | mul(res, res, t); |
---|
1941 | } |
---|
1942 | |
---|
1943 | rres = res; |
---|
1944 | } |
---|
1945 | |
---|
1946 | void NormMod(zz_p& x, const zz_pX& a, const zz_pX& f) |
---|
1947 | { |
---|
1948 | if (deg(f) <= 0 || deg(a) >= deg(f)) |
---|
1949 | Error("norm: bad args"); |
---|
1950 | |
---|
1951 | if (IsZero(a)) { |
---|
1952 | clear(x); |
---|
1953 | return; |
---|
1954 | } |
---|
1955 | |
---|
1956 | zz_p t; |
---|
1957 | resultant(t, f, a); |
---|
1958 | if (!IsOne(LeadCoeff(f))) { |
---|
1959 | zz_p t1; |
---|
1960 | power(t1, LeadCoeff(f), deg(a)); |
---|
1961 | inv(t1, t1); |
---|
1962 | mul(t, t, t1); |
---|
1963 | } |
---|
1964 | |
---|
1965 | x = t; |
---|
1966 | } |
---|
1967 | |
---|
1968 | |
---|
1969 | NTL_END_IMPL |
---|