| 1 | /* -*-c-*- |
| 2 | * |
| 3 | * Efficient reduction modulo sparse binary polynomials |
| 4 | * |
| 5 | * (c) 2004 Straylight/Edgeware |
| 6 | */ |
| 7 | |
| 8 | /*----- Licensing notice --------------------------------------------------* |
| 9 | * |
| 10 | * This file is part of Catacomb. |
| 11 | * |
| 12 | * Catacomb is free software; you can redistribute it and/or modify |
| 13 | * it under the terms of the GNU Library General Public License as |
| 14 | * published by the Free Software Foundation; either version 2 of the |
| 15 | * License, or (at your option) any later version. |
| 16 | * |
| 17 | * Catacomb is distributed in the hope that it will be useful, |
| 18 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 19 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 20 | * GNU Library General Public License for more details. |
| 21 | * |
| 22 | * You should have received a copy of the GNU Library General Public |
| 23 | * License along with Catacomb; if not, write to the Free |
| 24 | * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, |
| 25 | * MA 02111-1307, USA. |
| 26 | */ |
| 27 | |
| 28 | /*----- Header files ------------------------------------------------------*/ |
| 29 | |
| 30 | #include <mLib/alloc.h> |
| 31 | #include <mLib/darray.h> |
| 32 | #include <mLib/macros.h> |
| 33 | |
| 34 | #include "gf.h" |
| 35 | #include "gfreduce.h" |
| 36 | #include "gfreduce-exp.h" |
| 37 | #include "fibrand.h" |
| 38 | #include "mprand.h" |
| 39 | |
| 40 | /*----- Data structures ---------------------------------------------------*/ |
| 41 | |
| 42 | DA_DECL(instr_v, gfreduce_instr); |
| 43 | |
| 44 | /*----- Main code ---------------------------------------------------------*/ |
| 45 | |
| 46 | /* --- What's going on here? --- * |
| 47 | * |
| 48 | * Let's face it, @gfx_div@ sucks. It works (I hope), but it's not in any |
| 49 | * sense fast. Here, we do efficient reduction modulo sparse polynomials. |
| 50 | * (It works for arbitrary polynomials, but isn't efficient for dense ones.) |
| 51 | * |
| 52 | * Suppose that %$p(x) = x^n + p'(x) = \sum_{0\le i<n} p_i x^i$%, hopefully |
| 53 | * with only a few other %$p_i \ne 0$%. We're going to compile %$p$% into a |
| 54 | * sequence of instructions which can be used to perform reduction modulo |
| 55 | * %$p$%. The important observation is that %$x^n \equiv p' \pmod p$%. |
| 56 | * |
| 57 | * Suppose we're working with %$w$%-bit words; let %$n = N w + n'$% with |
| 58 | * %$0 \le n' < w$%. Let %$u(x)$% be some arbitrary polynomial. Write |
| 59 | * %$u = z x^k + u'$% with %$\deg u' < k \ge n$%; then a reduction step uses |
| 60 | * that %$u \equiv u' + z p' x^{k-n} \pmod p$%: the right hand side has |
| 61 | * degree %$\max \{ \deg u', k + \deg p' - n + \deg z \} < \deg u$%, so this |
| 62 | * makes progress towards a complete reduction. |
| 63 | * |
| 64 | * The compiled instruction sequence computes |
| 65 | * %$u' + z p' x^{k-n} = u' + \sum_{0\le i<n} z x^{k-n+i}$%. |
| 66 | */ |
| 67 | |
| 68 | /* --- @gfreduce_create@ --- * |
| 69 | * |
| 70 | * Arguments: @gfreduce *r@ = structure to fill in |
| 71 | * @mp *x@ = a (hopefully sparse) polynomial |
| 72 | * |
| 73 | * Returns: --- |
| 74 | * |
| 75 | * Use: Initializes a context structure for reduction. |
| 76 | */ |
| 77 | |
| 78 | struct gen { |
| 79 | unsigned f; /* Flags */ |
| 80 | #define f_lsr 1u /* Overflow from previous word */ |
| 81 | #define f_load 2u /* Outstanding @LOAD@ */ |
| 82 | instr_v iv; /* Instruction vector */ |
| 83 | size_t w; /* Currently loaded target word */ |
| 84 | size_t wi; /* Left-shifts for current word */ |
| 85 | }; |
| 86 | |
| 87 | #define INSTR(g_, op_, arg_) do { \ |
| 88 | struct gen *_g = (g_); \ |
| 89 | instr_v *_iv = &_g->iv; \ |
| 90 | size_t _i = DA_LEN(_iv); \ |
| 91 | \ |
| 92 | DA_ENSURE(_iv, 1); \ |
| 93 | DA(_iv)[_i].op = (op_); \ |
| 94 | DA(_iv)[_i].arg = (arg_); \ |
| 95 | DA_EXTEND(_iv, 1); \ |
| 96 | } while (0) |
| 97 | |
| 98 | static void emit_load(struct gen *g, size_t w) |
| 99 | { |
| 100 | INSTR(g, GFRI_LOAD, w); |
| 101 | g->f |= f_load; |
| 102 | g->w = w; |
| 103 | } |
| 104 | |
| 105 | static void emit_right_shifts(struct gen *g) |
| 106 | { |
| 107 | gfreduce_instr *ip; |
| 108 | size_t i, wl; |
| 109 | |
| 110 | /* --- Close off the current word --- * |
| 111 | * |
| 112 | * If we shifted into this current word with a nonzero bit offset, then |
| 113 | * we'll also need to arrange to perform a sequence of right shifts into |
| 114 | * the following word, which we might as well do by scanning the |
| 115 | * instruction sequence (which starts at @wi@). |
| 116 | * |
| 117 | * Either way, we leave a @LOAD@ unmatched if there was one before, in the |
| 118 | * hope that callers have an easier time; @g->w@ is updated to reflect the |
| 119 | * currently open word. |
| 120 | */ |
| 121 | |
| 122 | if (!(g->f & f_lsr)) |
| 123 | return; |
| 124 | |
| 125 | wl = DA_LEN(&g->iv); |
| 126 | INSTR(g, GFRI_STORE, g->w); |
| 127 | emit_load(g, g->w - 1); |
| 128 | for (i = g->wi; i < wl; i++) { |
| 129 | ip = &DA(&g->iv)[i]; |
| 130 | assert(ip->op == GFRI_LSL); |
| 131 | if (ip->arg) |
| 132 | INSTR(g, GFRI_LSR, MPW_BITS - ip->arg); |
| 133 | } |
| 134 | g->f &= ~f_lsr; |
| 135 | } |
| 136 | |
| 137 | static void ensure_loaded(struct gen *g, size_t w) |
| 138 | { |
| 139 | if (!(g->f & f_load)) { |
| 140 | emit_load(g, w); |
| 141 | g->wi = DA_LEN(&g->iv); |
| 142 | } else if (w != g->w) { |
| 143 | emit_right_shifts(g); |
| 144 | if (w != g->w) { |
| 145 | INSTR(g, GFRI_STORE, g->w); |
| 146 | emit_load(g, w); |
| 147 | } |
| 148 | g->wi = DA_LEN(&g->iv); |
| 149 | } |
| 150 | } |
| 151 | |
| 152 | void gfreduce_create(gfreduce *r, mp *p) |
| 153 | { |
| 154 | struct gen g = { 0, DA_INIT }; |
| 155 | unsigned long d; |
| 156 | unsigned dw; |
| 157 | mpscan sc; |
| 158 | unsigned long i; |
| 159 | size_t w, bb; |
| 160 | |
| 161 | /* --- Sort out the easy stuff --- */ |
| 162 | |
| 163 | d = mp_bits(p); assert(d); d--; |
| 164 | r->lim = d/MPW_BITS; |
| 165 | dw = d%MPW_BITS; |
| 166 | if (!dw) |
| 167 | r->mask = 0; |
| 168 | else { |
| 169 | r->mask = MPW(((mpw)-1) << dw); |
| 170 | r->lim++; |
| 171 | } |
| 172 | r->p = mp_copy(p); |
| 173 | |
| 174 | /* --- How this works --- * |
| 175 | * |
| 176 | * The instruction sequence is run with two ambient parameters: a pointer |
| 177 | * (usually) just past the most significant word of the polynomial to be |
| 178 | * reduced; and a word %$z$% which is the multiple of %$p'$% we are meant |
| 179 | * to add. |
| 180 | * |
| 181 | * The sequence visits each word of the polynomial at most once. Suppose |
| 182 | * %$u = z x^{w N} + u'$%; our pointer points just past the end of %$u'$%. |
| 183 | * Word %$I$% of %$u'$% will be affected by modulus bits %$p_i$% where |
| 184 | * %$(N - I - 1) w + 1 \le i \le (N - I + 1) w - 1$%, so %$p_i$% affects |
| 185 | * word %$I = \lceil (n - i + 1)/w \rceil$% and (if %$i$% is not a multiple |
| 186 | * of %$w$%) also word %$I - 1$%. |
| 187 | * |
| 188 | * We have four instructions: @LOAD@ reads a specified word of %$u$% into an |
| 189 | * accumulator, and @STORE@ stores it back (we'll always store back to the |
| 190 | * same word we most recently read, but this isn't a requirement); and |
| 191 | * @LSL@ and @LSR@, which XOR in appropriately shifted copies of %$z$% into |
| 192 | * the accumulator. So a typical program will contain sequences of @LSR@ |
| 193 | * and @LSL@ instructions sandwiched between @LOAD@/@STORE@ pairs. |
| 194 | * |
| 195 | * We do a single right-to-left pass across %$p$%. |
| 196 | */ |
| 197 | |
| 198 | bb = MPW_BITS - dw; |
| 199 | |
| 200 | for (i = 0, mp_scan(&sc, p); mp_step(&sc) && i < d; i++) { |
| 201 | if (!mp_bit(&sc)) |
| 202 | continue; |
| 203 | |
| 204 | /* --- We've found a set bit, so work out which word it affects --- * |
| 205 | * |
| 206 | * In general, a bit affects two words: it needs to be shifted left into |
| 207 | * one, and shifted right into the next. We find the former here. |
| 208 | */ |
| 209 | |
| 210 | w = (d - i + MPW_BITS - 1)/MPW_BITS; |
| 211 | |
| 212 | /* --- Concentrate on the appropriate word --- */ |
| 213 | |
| 214 | ensure_loaded(&g, w); |
| 215 | |
| 216 | /* --- Accumulate a new @LSL@ instruction --- * |
| 217 | * |
| 218 | * If this was a nonzero shift, then we'll need to arrange to do right |
| 219 | * shifts into the following word. |
| 220 | */ |
| 221 | |
| 222 | INSTR(&g, GFRI_LSL, (bb + i)%MPW_BITS); |
| 223 | if ((bb + i)%MPW_BITS) |
| 224 | g.f |= f_lsr; |
| 225 | } |
| 226 | |
| 227 | /* --- Wrapping up --- * |
| 228 | * |
| 229 | * We probably need a final @STORE@, and maybe a sequence of right shifts. |
| 230 | */ |
| 231 | |
| 232 | if (g.f & f_load) { |
| 233 | emit_right_shifts(&g); |
| 234 | INSTR(&g, GFRI_STORE, g.w); |
| 235 | } |
| 236 | |
| 237 | r->in = DA_LEN(&g.iv); |
| 238 | r->iv = xmalloc(r->in * sizeof(gfreduce_instr)); |
| 239 | memcpy(r->iv, DA(&g.iv), r->in * sizeof(gfreduce_instr)); |
| 240 | DA_DESTROY(&g.iv); |
| 241 | } |
| 242 | |
| 243 | #undef INSTR |
| 244 | |
| 245 | #undef f_lsr |
| 246 | #undef f_load |
| 247 | |
| 248 | /* --- @gfreduce_destroy@ --- * |
| 249 | * |
| 250 | * Arguments: @gfreduce *r@ = structure to free |
| 251 | * |
| 252 | * Returns: --- |
| 253 | * |
| 254 | * Use: Reclaims the resources from a reduction context. |
| 255 | */ |
| 256 | |
| 257 | void gfreduce_destroy(gfreduce *r) |
| 258 | { |
| 259 | mp_drop(r->p); |
| 260 | xfree(r->iv); |
| 261 | } |
| 262 | |
| 263 | /* --- @gfreduce_dump@ --- * |
| 264 | * |
| 265 | * Arguments: @gfreduce *r@ = structure to dump |
| 266 | * @FILE *fp@ = file to dump on |
| 267 | * |
| 268 | * Returns: --- |
| 269 | * |
| 270 | * Use: Dumps a reduction context. |
| 271 | */ |
| 272 | |
| 273 | void gfreduce_dump(gfreduce *r, FILE *fp) |
| 274 | { |
| 275 | size_t i; |
| 276 | |
| 277 | fprintf(fp, "poly = "); mp_writefile(r->p, fp, 16); |
| 278 | fprintf(fp, "\n lim = %lu; mask = %lx\n", |
| 279 | (unsigned long)r->lim, (unsigned long)r->mask); |
| 280 | for (i = 0; i < r->in; i++) { |
| 281 | static const char *opname[] = { "load", "lsl", "lsr", "store" }; |
| 282 | assert(r->iv[i].op < N(opname)); |
| 283 | fprintf(fp, " %s %lu\n", |
| 284 | opname[r->iv[i].op], |
| 285 | (unsigned long)r->iv[i].arg); |
| 286 | } |
| 287 | } |
| 288 | |
| 289 | /* --- @gfreduce_do@ --- * |
| 290 | * |
| 291 | * Arguments: @gfreduce *r@ = reduction context |
| 292 | * @mp *d@ = destination |
| 293 | * @mp *x@ = source |
| 294 | * |
| 295 | * Returns: Destination, @x@ reduced modulo the reduction poly. |
| 296 | */ |
| 297 | |
| 298 | static void run(const gfreduce_instr *i, const gfreduce_instr *il, |
| 299 | mpw *v, mpw z) |
| 300 | { |
| 301 | mpw w = 0; |
| 302 | |
| 303 | for (; i < il; i++) { |
| 304 | switch (i->op) { |
| 305 | case GFRI_LOAD: w = *(v - i->arg); break; |
| 306 | case GFRI_LSL: w ^= z << i->arg; break; |
| 307 | case GFRI_LSR: w ^= z >> i->arg; break; |
| 308 | case GFRI_STORE: *(v - i->arg) = MPW(w); break; |
| 309 | default: abort(); |
| 310 | } |
| 311 | } |
| 312 | } |
| 313 | |
| 314 | mp *gfreduce_do(gfreduce *r, mp *d, mp *x) |
| 315 | { |
| 316 | mpw *v, *vl; |
| 317 | const gfreduce_instr *il; |
| 318 | mpw z; |
| 319 | |
| 320 | /* --- Try to reuse the source's space --- */ |
| 321 | |
| 322 | MP_COPY(x); |
| 323 | if (d) MP_DROP(d); |
| 324 | MP_DEST(x, MP_LEN(x), x->f); |
| 325 | |
| 326 | /* --- Do the reduction --- */ |
| 327 | |
| 328 | il = r->iv + r->in; |
| 329 | if (MP_LEN(x) >= r->lim) { |
| 330 | v = x->v + r->lim; |
| 331 | vl = x->vl; |
| 332 | while (vl-- > v) { |
| 333 | while (*vl) { |
| 334 | z = *vl; |
| 335 | *vl = 0; |
| 336 | run(r->iv, il, vl, z); |
| 337 | } |
| 338 | } |
| 339 | if (r->mask) { |
| 340 | while (*vl & r->mask) { |
| 341 | z = *vl & r->mask; |
| 342 | *vl &= ~r->mask; |
| 343 | run(r->iv, il, vl, z); |
| 344 | } |
| 345 | } |
| 346 | } |
| 347 | |
| 348 | /* --- Done --- */ |
| 349 | |
| 350 | MP_SHRINK(x); |
| 351 | return (x); |
| 352 | } |
| 353 | |
| 354 | /* --- @gfreduce_sqrt@ --- * |
| 355 | * |
| 356 | * Arguments: @gfreduce *r@ = pointer to reduction context |
| 357 | * @mp *d@ = destination |
| 358 | * @mp *x@ = some polynomial |
| 359 | * |
| 360 | * Returns: The square root of @x@ modulo @r->p@, or null. |
| 361 | */ |
| 362 | |
| 363 | mp *gfreduce_sqrt(gfreduce *r, mp *d, mp *x) |
| 364 | { |
| 365 | mp *y = MP_COPY(x); |
| 366 | mp *z, *spare = MP_NEW; |
| 367 | unsigned long m = mp_bits(r->p) - 1; |
| 368 | unsigned long i; |
| 369 | |
| 370 | for (i = 0; i < m - 1; i++) { |
| 371 | mp *t = gf_sqr(spare, y); |
| 372 | spare = y; |
| 373 | y = gfreduce_do(r, t, t); |
| 374 | } |
| 375 | z = gf_sqr(spare, y); |
| 376 | z = gfreduce_do(r, z, z); |
| 377 | if (!MP_EQ(x, z)) { |
| 378 | mp_drop(y); |
| 379 | y = 0; |
| 380 | } |
| 381 | mp_drop(z); |
| 382 | mp_drop(d); |
| 383 | return (y); |
| 384 | } |
| 385 | |
| 386 | /* --- @gfreduce_trace@ --- * |
| 387 | * |
| 388 | * Arguments: @gfreduce *r@ = pointer to reduction context |
| 389 | * @mp *x@ = some polynomial |
| 390 | * |
| 391 | * Returns: The trace of @x@. (%$\Tr(x)=x + x^2 + \cdots + x^{2^{m-1}}$% |
| 392 | * if %$x \in \gf{2^m}$%). |
| 393 | */ |
| 394 | |
| 395 | int gfreduce_trace(gfreduce *r, mp *x) |
| 396 | { |
| 397 | mp *y = MP_COPY(x); |
| 398 | mp *spare = MP_NEW; |
| 399 | unsigned long m = mp_bits(r->p) - 1; |
| 400 | unsigned long i; |
| 401 | int rc; |
| 402 | |
| 403 | for (i = 0; i < m - 1; i++) { |
| 404 | mp *t = gf_sqr(spare, y); |
| 405 | spare = y; |
| 406 | y = gfreduce_do(r, t, t); |
| 407 | y = gf_add(y, y, x); |
| 408 | } |
| 409 | rc = !MP_ZEROP(y); |
| 410 | mp_drop(spare); |
| 411 | mp_drop(y); |
| 412 | return (rc); |
| 413 | } |
| 414 | |
| 415 | /* --- @gfreduce_halftrace@ --- * |
| 416 | * |
| 417 | * Arguments: @gfreduce *r@ = pointer to reduction context |
| 418 | * @mp *d@ = destination |
| 419 | * @mp *x@ = some polynomial |
| 420 | * |
| 421 | * Returns: The half-trace of @x@. |
| 422 | * (%$\HfTr(x)= x + x^{2^2} + \cdots + x^{2^{m-1}}$% |
| 423 | * if %$x \in \gf{2^m}$% with %$m$% odd). |
| 424 | */ |
| 425 | |
| 426 | mp *gfreduce_halftrace(gfreduce *r, mp *d, mp *x) |
| 427 | { |
| 428 | mp *y = MP_COPY(x); |
| 429 | mp *spare = MP_NEW; |
| 430 | unsigned long m = mp_bits(r->p) - 1; |
| 431 | unsigned long i; |
| 432 | |
| 433 | mp_drop(d); |
| 434 | for (i = 0; i < m - 1; i += 2) { |
| 435 | mp *t = gf_sqr(spare, y); |
| 436 | spare = y; |
| 437 | y = gfreduce_do(r, t, t); |
| 438 | t = gf_sqr(spare, y); |
| 439 | spare = y; |
| 440 | y = gfreduce_do(r, t, t); |
| 441 | y = gf_add(y, y, x); |
| 442 | } |
| 443 | mp_drop(spare); |
| 444 | return (y); |
| 445 | } |
| 446 | |
| 447 | /* --- @gfreduce_quadsolve@ --- * |
| 448 | * |
| 449 | * Arguments: @gfreduce *r@ = pointer to reduction context |
| 450 | * @mp *d@ = destination |
| 451 | * @mp *x@ = some polynomial |
| 452 | * |
| 453 | * Returns: A polynomial @y@ such that %$y^2 + y = x$%, or null. |
| 454 | */ |
| 455 | |
| 456 | mp *gfreduce_quadsolve(gfreduce *r, mp *d, mp *x) |
| 457 | { |
| 458 | unsigned long m = mp_bits(r->p) - 1; |
| 459 | mp *t; |
| 460 | |
| 461 | MP_COPY(x); |
| 462 | if (m & 1) |
| 463 | d = gfreduce_halftrace(r, d, x); |
| 464 | else { |
| 465 | mp *z, *w, *rho = MP_NEW; |
| 466 | mp *spare = MP_NEW; |
| 467 | grand *fr = fibrand_create(0); |
| 468 | unsigned long i; |
| 469 | |
| 470 | for (;;) { |
| 471 | rho = mprand(rho, m, fr, 0); |
| 472 | z = MP_ZERO; |
| 473 | w = MP_COPY(rho); |
| 474 | for (i = 0; i < m - 1; i++) { |
| 475 | t = gf_sqr(spare, z); spare = z; z = gfreduce_do(r, t, t); |
| 476 | t = gf_sqr(spare, w); spare = w; w = gfreduce_do(r, t, t); |
| 477 | t = gf_mul(spare, w, x); t = gfreduce_do(r, t, t); spare = t; |
| 478 | z = gf_add(z, z, t); |
| 479 | w = gf_add(w, w, rho); |
| 480 | } |
| 481 | if (!MP_ZEROP(w)) |
| 482 | break; |
| 483 | MP_DROP(z); |
| 484 | MP_DROP(w); |
| 485 | } |
| 486 | if (d) MP_DROP(d); |
| 487 | MP_DROP(w); |
| 488 | MP_DROP(spare); |
| 489 | MP_DROP(rho); |
| 490 | fr->ops->destroy(fr); |
| 491 | d = z; |
| 492 | } |
| 493 | |
| 494 | t = gf_sqr(MP_NEW, d); t = gfreduce_do(r, t, t); t = gf_add(t, t, d); |
| 495 | if (!MP_EQ(t, x)) { |
| 496 | MP_DROP(d); |
| 497 | d = 0; |
| 498 | } |
| 499 | MP_DROP(t); |
| 500 | MP_DROP(x); |
| 501 | if (d) d->v[0] &= ~(mpw)1; |
| 502 | return (d); |
| 503 | } |
| 504 | |
| 505 | /* --- @gfreduce_exp@ --- * |
| 506 | * |
| 507 | * Arguments: @gfreduce *gr@ = pointer to reduction context |
| 508 | * @mp *d@ = fake destination |
| 509 | * @mp *a@ = base |
| 510 | * @mp *e@ = exponent |
| 511 | * |
| 512 | * Returns: Result, %$a^e \bmod m$%. |
| 513 | */ |
| 514 | |
| 515 | mp *gfreduce_exp(gfreduce *gr, mp *d, mp *a, mp *e) |
| 516 | { |
| 517 | mp *x = MP_ONE; |
| 518 | mp *spare = (e->f & MP_BURN) ? MP_NEWSEC : MP_NEW; |
| 519 | |
| 520 | MP_SHRINK(e); |
| 521 | MP_COPY(a); |
| 522 | if (MP_ZEROP(e)) |
| 523 | ; |
| 524 | else { |
| 525 | if (MP_NEGP(e)) |
| 526 | a = gf_modinv(a, a, gr->p); |
| 527 | if (MP_LEN(e) < EXP_THRESH) |
| 528 | EXP_SIMPLE(x, a, e); |
| 529 | else |
| 530 | EXP_WINDOW(x, a, e); |
| 531 | } |
| 532 | mp_drop(d); |
| 533 | mp_drop(a); |
| 534 | mp_drop(spare); |
| 535 | return (x); |
| 536 | } |
| 537 | |
| 538 | /*----- Test rig ----------------------------------------------------------*/ |
| 539 | |
| 540 | #ifdef TEST_RIG |
| 541 | |
| 542 | #define MP(x) mp_readstring(MP_NEW, #x, 0, 0) |
| 543 | |
| 544 | static int vreduce(dstr *v) |
| 545 | { |
| 546 | mp *d = *(mp **)v[0].buf; |
| 547 | mp *n = *(mp **)v[1].buf; |
| 548 | mp *r = *(mp **)v[2].buf; |
| 549 | mp *c; |
| 550 | int ok = 1; |
| 551 | gfreduce rr; |
| 552 | |
| 553 | gfreduce_create(&rr, d); |
| 554 | c = gfreduce_do(&rr, MP_NEW, n); |
| 555 | if (!MP_EQ(c, r)) { |
| 556 | fprintf(stderr, "\n*** reduction failed\n*** "); |
| 557 | gfreduce_dump(&rr, stderr); |
| 558 | fprintf(stderr, "\n*** n = "); mp_writefile(n, stderr, 16); |
| 559 | fprintf(stderr, "\n*** r = "); mp_writefile(r, stderr, 16); |
| 560 | fprintf(stderr, "\n*** c = "); mp_writefile(c, stderr, 16); |
| 561 | fprintf(stderr, "\n"); |
| 562 | ok = 0; |
| 563 | } |
| 564 | gfreduce_destroy(&rr); |
| 565 | mp_drop(n); mp_drop(d); mp_drop(r); mp_drop(c); |
| 566 | assert(mparena_count(MPARENA_GLOBAL) == 0); |
| 567 | return (ok); |
| 568 | } |
| 569 | |
| 570 | static int vmodexp(dstr *v) |
| 571 | { |
| 572 | mp *p = *(mp **)v[0].buf; |
| 573 | mp *g = *(mp **)v[1].buf; |
| 574 | mp *x = *(mp **)v[2].buf; |
| 575 | mp *r = *(mp **)v[3].buf; |
| 576 | mp *c; |
| 577 | int ok = 1; |
| 578 | gfreduce rr; |
| 579 | |
| 580 | gfreduce_create(&rr, p); |
| 581 | c = gfreduce_exp(&rr, MP_NEW, g, x); |
| 582 | if (!MP_EQ(c, r)) { |
| 583 | fprintf(stderr, "\n*** modexp failed\n*** "); |
| 584 | fprintf(stderr, "\n*** p = "); mp_writefile(p, stderr, 16); |
| 585 | fprintf(stderr, "\n*** g = "); mp_writefile(g, stderr, 16); |
| 586 | fprintf(stderr, "\n*** x = "); mp_writefile(x, stderr, 16); |
| 587 | fprintf(stderr, "\n*** c = "); mp_writefile(c, stderr, 16); |
| 588 | fprintf(stderr, "\n*** r = "); mp_writefile(r, stderr, 16); |
| 589 | fprintf(stderr, "\n"); |
| 590 | ok = 0; |
| 591 | } |
| 592 | gfreduce_destroy(&rr); |
| 593 | mp_drop(p); mp_drop(g); mp_drop(r); mp_drop(x); mp_drop(c); |
| 594 | assert(mparena_count(MPARENA_GLOBAL) == 0); |
| 595 | return (ok); |
| 596 | } |
| 597 | |
| 598 | static int vsqrt(dstr *v) |
| 599 | { |
| 600 | mp *p = *(mp **)v[0].buf; |
| 601 | mp *x = *(mp **)v[1].buf; |
| 602 | mp *r = *(mp **)v[2].buf; |
| 603 | mp *c; |
| 604 | int ok = 1; |
| 605 | gfreduce rr; |
| 606 | |
| 607 | gfreduce_create(&rr, p); |
| 608 | c = gfreduce_sqrt(&rr, MP_NEW, x); |
| 609 | if (!MP_EQ(c, r)) { |
| 610 | fprintf(stderr, "\n*** sqrt failed\n*** "); |
| 611 | fprintf(stderr, "\n*** p = "); mp_writefile(p, stderr, 16); |
| 612 | fprintf(stderr, "\n*** x = "); mp_writefile(x, stderr, 16); |
| 613 | fprintf(stderr, "\n*** c = "); mp_writefile(c, stderr, 16); |
| 614 | fprintf(stderr, "\n*** r = "); mp_writefile(r, stderr, 16); |
| 615 | fprintf(stderr, "\n"); |
| 616 | ok = 0; |
| 617 | } |
| 618 | gfreduce_destroy(&rr); |
| 619 | mp_drop(p); mp_drop(r); mp_drop(x); mp_drop(c); |
| 620 | assert(mparena_count(MPARENA_GLOBAL) == 0); |
| 621 | return (ok); |
| 622 | } |
| 623 | |
| 624 | static int vtr(dstr *v) |
| 625 | { |
| 626 | mp *p = *(mp **)v[0].buf; |
| 627 | mp *x = *(mp **)v[1].buf; |
| 628 | int r = *(int *)v[2].buf, c; |
| 629 | int ok = 1; |
| 630 | gfreduce rr; |
| 631 | |
| 632 | gfreduce_create(&rr, p); |
| 633 | c = gfreduce_trace(&rr, x); |
| 634 | if (c != r) { |
| 635 | fprintf(stderr, "\n*** trace failed\n*** "); |
| 636 | fprintf(stderr, "\n*** p = "); mp_writefile(p, stderr, 16); |
| 637 | fprintf(stderr, "\n*** x = "); mp_writefile(x, stderr, 16); |
| 638 | fprintf(stderr, "\n*** c = %d", c); |
| 639 | fprintf(stderr, "\n*** r = %d", r); |
| 640 | fprintf(stderr, "\n"); |
| 641 | ok = 0; |
| 642 | } |
| 643 | gfreduce_destroy(&rr); |
| 644 | mp_drop(p); mp_drop(x); |
| 645 | assert(mparena_count(MPARENA_GLOBAL) == 0); |
| 646 | return (ok); |
| 647 | } |
| 648 | |
| 649 | static int vhftr(dstr *v) |
| 650 | { |
| 651 | mp *p = *(mp **)v[0].buf; |
| 652 | mp *x = *(mp **)v[1].buf; |
| 653 | mp *r = *(mp **)v[2].buf; |
| 654 | mp *c; |
| 655 | int ok = 1; |
| 656 | gfreduce rr; |
| 657 | |
| 658 | gfreduce_create(&rr, p); |
| 659 | c = gfreduce_halftrace(&rr, MP_NEW, x); |
| 660 | if (!MP_EQ(c, r)) { |
| 661 | fprintf(stderr, "\n*** halftrace failed\n*** "); |
| 662 | fprintf(stderr, "\n*** p = "); mp_writefile(p, stderr, 16); |
| 663 | fprintf(stderr, "\n*** x = "); mp_writefile(x, stderr, 16); |
| 664 | fprintf(stderr, "\n*** c = "); mp_writefile(c, stderr, 16); |
| 665 | fprintf(stderr, "\n*** r = "); mp_writefile(r, stderr, 16); |
| 666 | fprintf(stderr, "\n"); |
| 667 | ok = 0; |
| 668 | } |
| 669 | gfreduce_destroy(&rr); |
| 670 | mp_drop(p); mp_drop(r); mp_drop(x); mp_drop(c); |
| 671 | assert(mparena_count(MPARENA_GLOBAL) == 0); |
| 672 | return (ok); |
| 673 | } |
| 674 | |
| 675 | static int vquad(dstr *v) |
| 676 | { |
| 677 | mp *p = *(mp **)v[0].buf; |
| 678 | mp *x = *(mp **)v[1].buf; |
| 679 | mp *r = *(mp **)v[2].buf; |
| 680 | mp *c; |
| 681 | int ok = 1; |
| 682 | gfreduce rr; |
| 683 | |
| 684 | gfreduce_create(&rr, p); |
| 685 | c = gfreduce_quadsolve(&rr, MP_NEW, x); |
| 686 | if (!MP_EQ(c, r)) { |
| 687 | fprintf(stderr, "\n*** quadsolve failed\n*** "); |
| 688 | fprintf(stderr, "\n*** p = "); mp_writefile(p, stderr, 16); |
| 689 | fprintf(stderr, "\n*** x = "); mp_writefile(x, stderr, 16); |
| 690 | fprintf(stderr, "\n*** c = "); mp_writefile(c, stderr, 16); |
| 691 | fprintf(stderr, "\n*** r = "); mp_writefile(r, stderr, 16); |
| 692 | fprintf(stderr, "\n"); |
| 693 | ok = 0; |
| 694 | } |
| 695 | gfreduce_destroy(&rr); |
| 696 | mp_drop(p); mp_drop(r); mp_drop(x); mp_drop(c); |
| 697 | assert(mparena_count(MPARENA_GLOBAL) == 0); |
| 698 | return (ok); |
| 699 | } |
| 700 | |
| 701 | static test_chunk defs[] = { |
| 702 | { "reduce", vreduce, { &type_mp, &type_mp, &type_mp, 0 } }, |
| 703 | { "modexp", vmodexp, { &type_mp, &type_mp, &type_mp, &type_mp, 0 } }, |
| 704 | { "sqrt", vsqrt, { &type_mp, &type_mp, &type_mp, 0 } }, |
| 705 | { "trace", vtr, { &type_mp, &type_mp, &type_int, 0 } }, |
| 706 | { "halftrace", vhftr, { &type_mp, &type_mp, &type_mp, 0 } }, |
| 707 | { "quadsolve", vquad, { &type_mp, &type_mp, &type_mp, 0 } }, |
| 708 | { 0, 0, { 0 } } |
| 709 | }; |
| 710 | |
| 711 | int main(int argc, char *argv[]) |
| 712 | { |
| 713 | test_run(argc, argv, defs, SRCDIR"/t/gfreduce"); |
| 714 | return (0); |
| 715 | } |
| 716 | |
| 717 | #endif |
| 718 | |
| 719 | /*----- That's all, folks -------------------------------------------------*/ |