5 * Efficient reduction modulo nice primes
7 * (c) 2004 Straylight/Edgeware
10 /*----- Licensing notice --------------------------------------------------*
12 * This file is part of Catacomb.
14 * Catacomb is free software; you can redistribute it and/or modify
15 * it under the terms of the GNU Library General Public License as
16 * published by the Free Software Foundation; either version 2 of the
17 * License, or (at your option) any later version.
19 * Catacomb is distributed in the hope that it will be useful,
20 * but WITHOUT ANY WARRANTY; without even the implied warranty of
21 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
22 * GNU Library General Public License for more details.
24 * You should have received a copy of the GNU Library General Public
25 * License along with Catacomb; if not, write to the Free
26 * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
30 /*----- Header files ------------------------------------------------------*/
32 #include <mLib/darray.h>
33 #include <mLib/macros.h>
37 #include "mpreduce-exp.h"
39 /*----- Data structures ---------------------------------------------------*/
41 DA_DECL(instr_v, mpreduce_instr);
43 /*----- Main code ---------------------------------------------------------*/
45 /* --- @mpreduce_create@ --- *
47 * Arguments: @gfreduce *r@ = structure to fill in
48 * @mp *x@ = an integer
50 * Returns: Zero if successful; nonzero on failure.
52 * Use: Initializes a context structure for reduction.
55 int mpreduce_create(mpreduce *r, mp *p)
58 enum { Z = 0, Z1 = 2, X = 4, X0 = 6 };
65 /* --- Fill in the easy stuff --- */
76 /* --- Stash a new instruction --- */
78 #define INSTR(op_, argx_, argy_) do { \
80 DA(&iv)[DA_LEN(&iv)].op = (op_); \
81 DA(&iv)[DA_LEN(&iv)].argx = (argx_); \
82 DA(&iv)[DA_LEN(&iv)].argy = (argy_); \
86 /* --- Main loop --- *
88 * A simple state machine decomposes @p@ conveniently into positive and
89 * negative powers of 2. The pure form of the state machine is left below
90 * for reference (and in case I need inspiration for a NAF exponentiator).
94 for (i = 0, mp_scan(&sc, p); mp_step(&sc); i++) {
95 switch (st | mp_bit(&sc)) {
96 case Z | 1: st = Z1; break;
97 case Z1 | 0: st = Z; printf("+ %lu\n", i - 1); break;
98 case Z1 | 1: st = X; printf("- %lu\n", i - 1); break;
99 case X | 0: st = X0; break;
100 case X0 | 1: st = X; printf("- %lu\n", i - 1); break;
101 case X0 | 0: st = Z; printf("+ %lu\n", i - 1); break;
104 if (st >= X) printf("+ %lu\n", i - 1);
108 for (i = 0, mp_scan(&sc, p); i < d && mp_step(&sc); i++) {
109 switch (st | mp_bit(&sc)) {
110 case Z | 1: st = Z1; break;
111 case Z1 | 0: st = Z; op = MPRI_SUB; goto instr;
112 case Z1 | 1: st = X; op = MPRI_ADD; goto instr;
113 case X | 0: st = X0; break;
114 case X0 | 1: st = X; op = MPRI_ADD; goto instr;
115 case X0 | 0: st = Z; op = MPRI_SUB; goto instr;
117 w = (d - i)/MPW_BITS + 1;
118 b = (MPW_BITS + i - d - 1)%MPW_BITS;
119 INSTR(op | !!b, w, b);
122 if (DA_LEN(&iv) && (DA(&iv)[DA_LEN(&iv) - 1].op & ~1u) == MPRI_SUB) {
130 /* --- Wrap up --- */
136 r->iv = xmalloc(r->in * sizeof(mpreduce_instr));
137 memcpy(r->iv, DA(&iv), r->in * sizeof(mpreduce_instr));
139 r->iv = xmalloc(r->in * 2 * sizeof(mpreduce_instr));
140 for (i = 0; i < r->in; i++) {
141 r->iv[i] = DA(&iv)[i];
142 op = r->iv[i].op & ~1u;
151 r->iv[i + r->in].op = op;
152 r->iv[i + r->in].argx = w;
153 r->iv[i + r->in].argy = b;
160 /* --- @mpreduce_destroy@ --- *
162 * Arguments: @mpreduce *r@ = structure to free
166 * Use: Reclaims the resources from a reduction context.
169 void mpreduce_destroy(mpreduce *r)
172 if (r->iv) xfree(r->iv);
175 /* --- @mpreduce_dump@ --- *
177 * Arguments: @mpreduce *r@ = structure to dump
178 * @FILE *fp@ = file to dump on
182 * Use: Dumps a reduction context.
185 void mpreduce_dump(mpreduce *r, FILE *fp)
188 static const char *opname[] = { "add", "addshift", "sub", "subshift" };
190 fprintf(fp, "mod = "); mp_writefile(r->p, fp, 16);
191 fprintf(fp, "\n lim = %lu; s = %d\n", (unsigned long)r->lim, r->s);
192 for (i = 0; i < r->in; i++) {
193 assert(r->iv[i].op < N(opname));
194 fprintf(fp, " %s %lu %lu\n",
196 (unsigned long)r->iv[i].argx,
197 (unsigned long)r->iv[i].argy);
200 fprintf(fp, "tail end charlie\n");
201 for (i = r->in; i < 2 * r->in; i++) {
202 assert(r->iv[i].op < N(opname));
203 fprintf(fp, " %s %lu %lu\n",
205 (unsigned long)r->iv[i].argx,
206 (unsigned long)r->iv[i].argy);
211 /* --- @mpreduce_do@ --- *
213 * Arguments: @mpreduce *r@ = reduction context
214 * @mp *d@ = destination
217 * Returns: Destination, @x@ reduced modulo the reduction poly.
220 static void run(const mpreduce_instr *i, const mpreduce_instr *il,
223 for (; i < il; i++) {
226 mp_build(&vv, v - i->argx, v + 1);
227 printf(" 0x"); mp_writefile(&vv, stdout, 16);
228 printf(" %c (0x%lx << %u) == 0x",
229 (i->op & ~1u) == MPRI_ADD ? '+' : '-',
234 case MPRI_ADD: MPX_UADDN(v - i->argx, v + 1, z); break;
235 case MPRI_ADDLSL: mpx_uaddnlsl(v - i->argx, v + 1, z, i->argy); break;
236 case MPRI_SUB: MPX_USUBN(v - i->argx, v + 1, z); break;
237 case MPRI_SUBLSL: mpx_usubnlsl(v - i->argx, v + 1, z, i->argy); break;
242 mp_build(&vv, v - i->argx, v + 1);
243 mp_writefile(&vv, stdout, 16);
249 mp *mpreduce_do(mpreduce *r, mp *d, mp *x)
252 const mpreduce_instr *il;
256 mp *_r = 0, *_rr = 0;
259 /* --- If source is negative, divide --- */
262 mp_div(0, &d, x, r->p);
266 /* --- Try to reuse the source's space --- */
270 MP_DEST(x, MP_LEN(x), x->f);
272 /* --- Do the reduction --- */
276 mp_div(0, &_r, x, r->p);
282 if (MP_LEN(x) >= r->lim) {
289 run(r->iv, il, vl, z);
292 mp_div(0, &_rr, x, r->p);
293 assert(MP_EQ(_r, _rr));
298 while (*vl >> r->s) {
300 *vl &= ((1 << r->s) - 1);
301 run(r->iv + r->in, il + r->in, vl, z);
304 mp_div(0, &_rr, x, r->p);
305 assert(MP_EQ(_r, _rr));
311 /* --- Finishing touches --- */
314 if (MP_CMP(x, >=, r->p))
315 x = mp_sub(x, x, r->p);
320 assert(MP_EQ(_r, x));
327 /* --- @mpreduce_exp@ --- *
329 * Arguments: @mpreduce *mr@ = pointer to reduction context
330 * @mp *d@ = fake destination
334 * Returns: Result, %$a^e \bmod m$%.
337 mp *mpreduce_exp(mpreduce *mr, mp *d, mp *a, mp *e)
340 mp *spare = (e->f & MP_BURN) ? MP_NEWSEC : MP_NEW;
348 a = mp_modinv(a, a, mr->p);
349 if (MP_LEN(e) < EXP_THRESH)
360 /*----- Test rig ----------------------------------------------------------*/
365 #define MP(x) mp_readstring(MP_NEW, #x, 0, 0)
367 static int vreduce(dstr *v)
369 mp *d = *(mp **)v[0].buf;
370 mp *n = *(mp **)v[1].buf;
371 mp *r = *(mp **)v[2].buf;
376 mpreduce_create(&rr, d);
377 c = mpreduce_do(&rr, MP_NEW, n);
379 fprintf(stderr, "\n*** reduction failed\n*** ");
380 mpreduce_dump(&rr, stderr);
381 fprintf(stderr, "\n*** n = "); mp_writefile(n, stderr, 10);
382 fprintf(stderr, "\n*** r = "); mp_writefile(r, stderr, 10);
383 fprintf(stderr, "\n*** c = "); mp_writefile(c, stderr, 10);
384 fprintf(stderr, "\n");
387 mpreduce_destroy(&rr);
388 mp_drop(n); mp_drop(d); mp_drop(r); mp_drop(c);
389 assert(mparena_count(MPARENA_GLOBAL) == 0);
393 static int vmodexp(dstr *v)
395 mp *p = *(mp **)v[0].buf;
396 mp *g = *(mp **)v[1].buf;
397 mp *x = *(mp **)v[2].buf;
398 mp *r = *(mp **)v[3].buf;
403 mpreduce_create(&rr, p);
404 c = mpreduce_exp(&rr, MP_NEW, g, x);
406 fprintf(stderr, "\n*** modexp failed\n*** ");
407 fprintf(stderr, "\n*** p = "); mp_writefile(p, stderr, 10);
408 fprintf(stderr, "\n*** g = "); mp_writefile(g, stderr, 10);
409 fprintf(stderr, "\n*** x = "); mp_writefile(x, stderr, 10);
410 fprintf(stderr, "\n*** c = "); mp_writefile(c, stderr, 10);
411 fprintf(stderr, "\n*** r = "); mp_writefile(r, stderr, 10);
412 fprintf(stderr, "\n");
415 mpreduce_destroy(&rr);
416 mp_drop(p); mp_drop(g); mp_drop(r); mp_drop(x); mp_drop(c);
417 assert(mparena_count(MPARENA_GLOBAL) == 0);
421 static test_chunk defs[] = {
422 { "reduce", vreduce, { &type_mp, &type_mp, &type_mp, 0 } },
423 { "modexp", vmodexp, { &type_mp, &type_mp, &type_mp, &type_mp, 0 } },
427 int main(int argc, char *argv[])
429 test_run(argc, argv, defs, SRCDIR"/tests/mpreduce");
435 /*----- That's all, folks -------------------------------------------------*/