Commit | Line | Data |
---|---|---|
d3409d5e | 1 | /* -*-c-*- |
d3409d5e | 2 | * |
3 | * Basic arithmetic on multiprecision integers | |
4 | * | |
5 | * (c) 1999 Straylight/Edgeware | |
6 | */ | |
7 | ||
45c0fd36 | 8 | /*----- Licensing notice --------------------------------------------------* |
d3409d5e | 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. | |
45c0fd36 | 16 | * |
d3409d5e | 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. | |
45c0fd36 | 21 | * |
d3409d5e | 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 | ||
d3409d5e | 28 | /*----- Header files ------------------------------------------------------*/ |
29 | ||
30 | #include "mp.h" | |
31 | ||
ef5f4810 | 32 | /*----- Macros ------------------------------------------------------------*/ |
33 | ||
34 | #define MAX(x, y) ((x) >= (y) ? (x) : (y)) | |
35 | ||
d3409d5e | 36 | /*----- Main code ---------------------------------------------------------*/ |
37 | ||
81578196 | 38 | /* --- @mp_lsl@, @mp_lslc@, @mp_lsr@ --- * |
d3409d5e | 39 | * |
f09e814a | 40 | * Arguments: @mp *d@ = destination |
41 | * @mp *a@ = source | |
42 | * @size_t n@ = number of bits to move | |
d3409d5e | 43 | * |
f09e814a | 44 | * Returns: Result, @a@ shifted left or right by @n@. |
81578196 | 45 | * |
46 | * Use: Bitwise shift operators. @mp_lslc@ fills the bits introduced | |
47 | * on the right with ones instead of zeroes: it's used | |
48 | * internally by @mp_lsl2c@, though it may be useful on its | |
49 | * own. | |
d3409d5e | 50 | */ |
51 | ||
f09e814a | 52 | mp *mp_lsl(mp *d, mp *a, size_t n) |
d3409d5e | 53 | { |
f09e814a | 54 | MP_DEST(d, MP_LEN(a) + (n + MPW_BITS - 1) / MPW_BITS, a->f); |
55 | mpx_lsl(d->v, d->vl, a->v, a->vl, n); | |
56 | d->f = a->f & (MP_NEG | MP_BURN); | |
57 | MP_SHRINK(d); | |
58 | return (d); | |
59 | } | |
d3409d5e | 60 | |
81578196 | 61 | mp *mp_lslc(mp *d, mp *a, size_t n) |
62 | { | |
63 | MP_DEST(d, MP_LEN(a) + (n + MPW_BITS - 1) / MPW_BITS, a->f); | |
64 | mpx_lslc(d->v, d->vl, a->v, a->vl, n); | |
65 | d->f = a->f & (MP_NEG | MP_BURN); | |
66 | MP_SHRINK(d); | |
67 | return (d); | |
68 | } | |
69 | ||
f09e814a | 70 | mp *mp_lsr(mp *d, mp *a, size_t n) |
71 | { | |
d34decd2 | 72 | MP_DEST(d, MP_LEN(a), a->f); |
f09e814a | 73 | mpx_lsr(d->v, d->vl, a->v, a->vl, n); |
74 | d->f = a->f & (MP_NEG | MP_BURN); | |
d3409d5e | 75 | MP_SHRINK(d); |
76 | return (d); | |
77 | } | |
78 | ||
f09e814a | 79 | /* --- @mp_lsl2c@, @mp_lsr2c@ --- * |
d3409d5e | 80 | * |
81 | * Arguments: @mp *d@ = destination | |
82 | * @mp *a@ = source | |
f09e814a | 83 | * @size_t n@ = number of bits to move |
d3409d5e | 84 | * |
f09e814a | 85 | * Returns: Result, @a@ shifted left or right by @n@. Handles the |
86 | * pretence of sign-extension for negative numbers. | |
d3409d5e | 87 | */ |
88 | ||
f09e814a | 89 | mp *mp_lsl2c(mp *d, mp *a, size_t n) |
d3409d5e | 90 | { |
a69a3efd | 91 | if (!MP_NEGP(a)) |
f09e814a | 92 | return (mp_lsl(d, a, n)); |
93 | d = mp_not2c(d, a); | |
81578196 | 94 | d = mp_lslc(d, d, n); |
f09e814a | 95 | d = mp_not2c(d, d); |
96 | return (d); | |
97 | } | |
d3409d5e | 98 | |
f09e814a | 99 | mp *mp_lsr2c(mp *d, mp *a, size_t n) |
100 | { | |
a69a3efd | 101 | if (!MP_NEGP(a)) |
f09e814a | 102 | return (mp_lsr(d, a, n)); |
103 | d = mp_not2c(d, a); | |
104 | d = mp_lsr(d, d, n); | |
105 | d = mp_not2c(d, d); | |
106 | return (d); | |
d3409d5e | 107 | } |
108 | ||
f09e814a | 109 | /* --- @mp_testbit@ --- * |
d3409d5e | 110 | * |
f09e814a | 111 | * Arguments: @mp *x@ = a large integer |
09d00c6b | 112 | * @unsigned long n@ = which bit to test |
d3409d5e | 113 | * |
f09e814a | 114 | * Returns: Nonzero if the bit is set, zero if not. |
d3409d5e | 115 | */ |
116 | ||
09d00c6b | 117 | int mp_testbit(mp *x, unsigned long n) |
d3409d5e | 118 | { |
814141d3 | 119 | if (n >= MPW_BITS * MP_LEN(x)) |
f09e814a | 120 | return (0); |
09d00c6b | 121 | return ((x->v[n/MPW_BITS] >> n%MPW_BITS) & 1u); |
d3409d5e | 122 | } |
123 | ||
f09e814a | 124 | /* --- @mp_testbit2c@ --- * |
d3409d5e | 125 | * |
f09e814a | 126 | * Arguments: @mp *x@ = a large integer |
09d00c6b | 127 | * @unsigned long n@ = which bit to test |
d3409d5e | 128 | * |
f09e814a | 129 | * Returns: Nonzero if the bit is set, zero if not. Fakes up two's |
130 | * complement representation. | |
d3409d5e | 131 | */ |
132 | ||
09d00c6b | 133 | int mp_testbit2c(mp *x, unsigned long n) |
d3409d5e | 134 | { |
f09e814a | 135 | int r; |
a69a3efd | 136 | if (!MP_NEGP(x)) |
f09e814a | 137 | return (mp_testbit(x, n)); |
138 | x = mp_not2c(MP_NEW, x); | |
139 | r = !mp_testbit(x, n); | |
140 | MP_DROP(x); | |
141 | return (r); | |
d3409d5e | 142 | } |
143 | ||
09d00c6b | 144 | /* --- @mp_setbit@, @mp_clearbit@ --- * |
145 | * | |
146 | * Arguments: @mp *d@ = a destination | |
147 | * @mp *x@ = a large integer | |
148 | * @unsigned long n@ = which bit to modify | |
149 | * | |
150 | * Returns: The argument @x@, with the appropriate bit set or cleared. | |
151 | */ | |
152 | ||
153 | mp *mp_setbit(mp *d, mp *x, unsigned long n) | |
154 | { | |
155 | size_t rq; | |
156 | ||
157 | rq = n + MPW_BITS; rq -= rq % MPW_BITS; | |
158 | if (d != x) { | |
159 | if (d) MP_DROP(d); | |
160 | d = MP_COPY(x); | |
161 | } | |
162 | MP_DEST(d, rq, x->f & (MP_NEG | MP_BURN)); | |
163 | d->v[n/MPW_BITS] |= 1 << n%MPW_BITS; | |
164 | return (d); | |
165 | } | |
166 | ||
167 | mp *mp_clearbit(mp *d, mp *x, unsigned long n) | |
168 | { | |
169 | size_t rq; | |
170 | ||
171 | rq = n + MPW_BITS; rq -= rq % MPW_BITS; | |
172 | if (d != x) { | |
173 | if (d) MP_DROP(d); | |
174 | d = MP_COPY(x); | |
175 | } | |
176 | MP_DEST(d, rq, x->f & (MP_NEG | MP_BURN)); | |
177 | d->v[n/MPW_BITS] &= ~(1 << n%MPW_BITS); | |
178 | return (d); | |
179 | } | |
180 | ||
181 | /* --- @mp_setbit2c@, @mp_clearbit2c@ --- * | |
182 | * | |
183 | * Arguments: @mp *d@ = a destination | |
184 | * @mp *x@ = a large integer | |
185 | * @unsigned long n@ = which bit to modify | |
186 | * | |
187 | * Returns: The argument @x@, with the appropriate bit set or cleared. | |
188 | * Fakes up two's complement representation. | |
189 | */ | |
190 | ||
191 | mp *mp_setbit2c(mp *d, mp *x, unsigned long n) | |
192 | { | |
a69a3efd | 193 | if (!MP_NEGP(x)) |
09d00c6b | 194 | return mp_setbit(d, x, n); |
195 | d = mp_not2c(d, x); | |
196 | d = mp_clearbit(d, d, n); | |
197 | d = mp_not2c(d, d); | |
198 | return (d); | |
199 | } | |
200 | ||
201 | mp *mp_clearbit2c(mp *d, mp *x, unsigned long n) | |
202 | { | |
a69a3efd | 203 | if (!MP_NEGP(x)) |
09d00c6b | 204 | return mp_clearbit(d, x, n); |
205 | d = mp_not2c(d, x); | |
206 | d = mp_setbit(d, d, n); | |
207 | d = mp_not2c(d, d); | |
208 | return (d); | |
209 | } | |
210 | ||
4b536f42 | 211 | /* --- @mp_eq@ --- * |
212 | * | |
213 | * Arguments: @const mp *a, *b@ = two numbers | |
214 | * | |
215 | * Returns: Nonzero if the numbers are equal. | |
216 | */ | |
217 | ||
218 | int mp_eq(const mp *a, const mp *b) { return (MP_EQ(a, b)); } | |
219 | ||
d3409d5e | 220 | /* --- @mp_cmp@ --- * |
221 | * | |
222 | * Arguments: @const mp *a, *b@ = two numbers | |
223 | * | |
224 | * Returns: Less than, equal to or greater than zero, according to | |
225 | * whether @a@ is less than, equal to or greater than @b@. | |
226 | */ | |
227 | ||
228 | int mp_cmp(const mp *a, const mp *b) | |
229 | { | |
4108c8d2 | 230 | if (!((a->f ^ b->f) & MP_NEG)) { |
231 | if (a->f & MP_NEG) | |
232 | return (-mpx_ucmp(a->v, a->vl, b->v, b->vl)); | |
233 | else | |
234 | return (mpx_ucmp(a->v, a->vl, b->v, b->vl)); | |
235 | } else if (a->f & MP_NEG) | |
d3409d5e | 236 | return (-1); |
237 | else | |
238 | return (+1); | |
239 | } | |
240 | ||
397041a9 | 241 | /* --- @mp_neg@ --- * |
242 | * | |
243 | * Arguments: @mp *d@ = destination | |
244 | * @mp *a@ = argument | |
245 | * | |
246 | * Returns: The negation of the argument. | |
247 | * | |
248 | * Use: Negates its argument. | |
249 | */ | |
250 | ||
251 | mp *mp_neg(mp *d, mp *a) | |
252 | { | |
253 | /* --- Surprising amounts of messing about required --- */ | |
254 | ||
255 | MP_SHRINK(a); | |
256 | MP_COPY(a); | |
81578196 | 257 | if (d) |
258 | MP_DROP(d); | |
259 | if (a->v == a->vl) | |
397041a9 | 260 | return (a); |
397041a9 | 261 | MP_DEST(a, MP_LEN(a), a->f); |
262 | a->f ^= MP_NEG; | |
263 | return (a); | |
264 | } | |
265 | ||
f09e814a | 266 | /* --- @mp_bitop@ --- * |
0f32e0f8 | 267 | * |
268 | * Arguments: @mp *d@ = destination | |
269 | * @mp *a, *b@ = sources | |
270 | * | |
f09e814a | 271 | * Returns: The result of the given bitwise operation. These functions |
272 | * don't handle negative numbers at all sensibly. For that, use | |
273 | * the @...2c@ variants. The functions are named after the | |
274 | * truth tables they generate: | |
275 | * | |
276 | * a: 0011 | |
277 | * b: 0101 | |
278 | * @mpx_bitXXXX@ | |
0f32e0f8 | 279 | */ |
280 | ||
f09e814a | 281 | #define MP_BITBINOP(string) \ |
0f32e0f8 | 282 | \ |
f09e814a | 283 | mp *mp_bit##string(mp *d, mp *a, mp *b) \ |
0f32e0f8 | 284 | { \ |
75263f25 | 285 | MP_DEST(d, MAX(MP_LEN(a), MP_LEN(b)), (a->f | b->f) & ~MP_NEG); \ |
f09e814a | 286 | mpx_bit##string(d->v, d->vl, a->v, a->vl, b->v, b->vl); \ |
0f32e0f8 | 287 | d->f = (a->f | b->f) & MP_BURN; \ |
288 | MP_SHRINK(d); \ | |
289 | return (d); \ | |
290 | } | |
291 | ||
f09e814a | 292 | MPX_DOBIN(MP_BITBINOP) |
293 | ||
294 | /* --- @mp_not@ --- * | |
295 | * | |
296 | * Arguments: @mp *d@ = destination | |
297 | * @mp *a@ = source | |
298 | * | |
299 | * Returns: The bitwise complement of the source. | |
45c0fd36 | 300 | */ |
0f32e0f8 | 301 | |
302 | mp *mp_not(mp *d, mp *a) | |
303 | { | |
304 | MP_DEST(d, MP_LEN(a), a->f); | |
305 | mpx_not(d->v, d->vl, a->v, a->vl); | |
306 | d->f = a->f & MP_BURN; | |
307 | MP_SHRINK(d); | |
308 | return (d); | |
309 | } | |
310 | ||
f09e814a | 311 | /* --- @mp_bitop2c@ --- * |
312 | * | |
313 | * Arguments: @mp *d@ = destination | |
314 | * @mp *a, *b@ = sources | |
315 | * | |
316 | * Returns: The result of the given bitwise operation. Negative numbers | |
317 | * are treated as two's complement, sign-extended infinitely to | |
318 | * the left. The functions are named after the truth tables | |
319 | * they generate: | |
320 | * | |
321 | * a: 0011 | |
322 | * b: 0101 | |
323 | * @mpx_bitXXXX@ | |
324 | */ | |
325 | ||
326 | /* --- How this actually works --- * | |
327 | * | |
328 | * The two arguments are inverted (with a sign-swap) if they're currently | |
329 | * negative. This means that we end up using a different function (one which | |
330 | * reinverts as we go) for the main operation. Also, if the sign would be | |
331 | * negative at the end, we preinvert the output and then invert again with a | |
332 | * sign-swap. | |
333 | * | |
45c0fd36 | 334 | * Start with: wxyz WXYZ |
f09e814a | 335 | * If @a@ negative: yzwx or YZWX |
45c0fd36 MW |
336 | * If @b@ negative: xwzy XWZY |
337 | * If both negative: zyxw ZYXW | |
f09e814a | 338 | */ |
339 | ||
340 | #define MP_BIT2CBINOP(n, base, an, bn, abn, p_base, p_an, p_bn, p_abn) \ | |
341 | \ | |
342 | mp *mp_bit##n##2c(mp *d, mp *a, mp *b) \ | |
343 | { \ | |
344 | if (!((a->f | b->f) & MP_NEG)) { /* Both positive */ \ | |
345 | d = mp_bit##base(d, a, b); \ | |
346 | p_base \ | |
347 | } else if (!(b->f & MP_NEG)) { /* Only @b@ positive */ \ | |
348 | MP_COPY(b); \ | |
349 | d = mp_not2c(d, a); \ | |
350 | d = mp_bit##an(d, d, b); \ | |
351 | MP_DROP(b); \ | |
352 | p_an \ | |
353 | } else if (!(a->f & MP_NEG)) { /* Only @a@ positive */ \ | |
354 | MP_COPY(a); \ | |
355 | d = mp_not2c(d, b); \ | |
356 | d = mp_bit##bn(d, a, d); \ | |
357 | MP_DROP(a); \ | |
358 | p_bn \ | |
359 | } else { /* Both negative */ \ | |
360 | mp *t = mp_not2c(MP_NEW, a); \ | |
15f21155 | 361 | d = mp_not2c(d, b); \ |
f09e814a | 362 | d = mp_bit##abn(d, t, d); \ |
363 | MP_DROP(t); \ | |
364 | p_abn \ | |
365 | } \ | |
366 | return (d); \ | |
367 | } \ | |
368 | ||
369 | #define NEG d = mp_not2c(d, d); | |
370 | #define POS | |
371 | MP_BIT2CBINOP(0000, 0000, 0000, 0000, 0000, POS, POS, POS, POS) | |
372 | MP_BIT2CBINOP(0001, 0001, 0100, 0010, 0111, POS, POS, POS, NEG) | |
373 | MP_BIT2CBINOP(0010, 0010, 0111, 0001, 0100, POS, NEG, POS, POS) | |
374 | MP_BIT2CBINOP(0011, 0011, 0011, 0011, 0011, POS, NEG, POS, NEG) | |
375 | MP_BIT2CBINOP(0100, 0100, 0001, 0111, 0010, POS, POS, NEG, POS) | |
376 | MP_BIT2CBINOP(0101, 0101, 0101, 0101, 0101, POS, POS, NEG, NEG) | |
377 | MP_BIT2CBINOP(0110, 0110, 0110, 0110, 0110, POS, NEG, NEG, POS) | |
378 | MP_BIT2CBINOP(0111, 0111, 0010, 0100, 0001, POS, NEG, NEG, NEG) | |
379 | MP_BIT2CBINOP(1000, 0111, 0010, 0100, 0001, NEG, POS, POS, POS) | |
380 | MP_BIT2CBINOP(1001, 0110, 0110, 0110, 0110, NEG, POS, POS, NEG) | |
381 | MP_BIT2CBINOP(1010, 0101, 0101, 0101, 0101, NEG, NEG, POS, POS) | |
382 | MP_BIT2CBINOP(1011, 0100, 0001, 0111, 0010, NEG, NEG, POS, NEG) | |
383 | MP_BIT2CBINOP(1100, 0011, 0011, 0011, 0011, NEG, POS, NEG, POS) | |
384 | MP_BIT2CBINOP(1101, 0010, 0111, 0001, 0100, NEG, POS, NEG, NEG) | |
385 | MP_BIT2CBINOP(1110, 0001, 0100, 0010, 0111, NEG, NEG, NEG, POS) | |
386 | MP_BIT2CBINOP(1111, 0000, 0000, 0000, 0000, NEG, NEG, NEG, NEG) | |
387 | #undef NEG | |
388 | #undef POS | |
389 | ||
390 | /* --- @mp_not2c@ --- * | |
391 | * | |
392 | * Arguments: @mp *d@ = destination | |
393 | * @mp *a@ = source | |
394 | * | |
395 | * Returns: The sign-extended complement of the argument. | |
396 | */ | |
397 | ||
398 | mp *mp_not2c(mp *d, mp *a) | |
399 | { | |
400 | mpw one = 1; | |
401 | ||
402 | MP_DEST(d, MP_LEN(a) + 1, a->f); | |
403 | if (d == a) { | |
a69a3efd | 404 | if (MP_NEGP(a)) |
f09e814a | 405 | MPX_USUBN(d->v, d->vl, 1); |
406 | else | |
407 | MPX_UADDN(d->v, d->vl, 1); | |
408 | } else { | |
a69a3efd | 409 | if (MP_NEGP(a)) |
f09e814a | 410 | mpx_usub(d->v, d->vl, a->v, a->vl, &one, &one + 1); |
411 | else | |
412 | mpx_uadd(d->v, d->vl, a->v, a->vl, &one, &one + 1); | |
413 | } | |
414 | d->f = (a->f & (MP_NEG | MP_BURN)) ^ MP_NEG; | |
415 | MP_SHRINK(d); | |
416 | return (d); | |
417 | } | |
418 | ||
d3409d5e | 419 | /* --- @mp_add@ --- * |
420 | * | |
421 | * Arguments: @mp *d@ = destination | |
ef5f4810 | 422 | * @mp *a, *b@ = sources |
d3409d5e | 423 | * |
424 | * Returns: Result, @a@ added to @b@. | |
425 | */ | |
426 | ||
ef5f4810 | 427 | mp *mp_add(mp *d, mp *a, mp *b) |
d3409d5e | 428 | { |
d34decd2 | 429 | MP_DEST(d, MAX(MP_LEN(a), MP_LEN(b)) + 1, a->f | b->f); |
d3409d5e | 430 | if (!((a->f ^ b->f) & MP_NEG)) |
431 | mpx_uadd(d->v, d->vl, a->v, a->vl, b->v, b->vl); | |
432 | else { | |
433 | if (MPX_UCMP(a->v, a->vl, <, b->v, b->vl)) { | |
ef5f4810 | 434 | mp *t = a; a = b; b = t; |
d3409d5e | 435 | } |
436 | mpx_usub(d->v, d->vl, a->v, a->vl, b->v, b->vl); | |
437 | } | |
438 | d->f = ((a->f | b->f) & MP_BURN) | (a->f & MP_NEG); | |
439 | MP_SHRINK(d); | |
440 | return (d); | |
441 | } | |
442 | ||
443 | /* --- @mp_sub@ --- * | |
444 | * | |
445 | * Arguments: @mp *d@ = destination | |
ef5f4810 | 446 | * @mp *a, *b@ = sources |
d3409d5e | 447 | * |
448 | * Returns: Result, @b@ subtracted from @a@. | |
449 | */ | |
450 | ||
ef5f4810 | 451 | mp *mp_sub(mp *d, mp *a, mp *b) |
d3409d5e | 452 | { |
453 | unsigned sgn = 0; | |
d34decd2 | 454 | MP_DEST(d, MAX(MP_LEN(a), MP_LEN(b)) + 1, a->f | b->f); |
d3409d5e | 455 | if ((a->f ^ b->f) & MP_NEG) |
456 | mpx_uadd(d->v, d->vl, a->v, a->vl, b->v, b->vl); | |
457 | else { | |
458 | if (MPX_UCMP(a->v, a->vl, <, b->v, b->vl)) { | |
ef5f4810 | 459 | mp *t = a; a = b; b = t; |
d3409d5e | 460 | sgn = MP_NEG; |
461 | } | |
462 | mpx_usub(d->v, d->vl, a->v, a->vl, b->v, b->vl); | |
463 | } | |
464 | d->f = ((a->f | b->f) & MP_BURN) | ((a->f ^ sgn) & MP_NEG); | |
465 | MP_SHRINK(d); | |
466 | return (d); | |
467 | } | |
468 | ||
469 | /* --- @mp_mul@ --- * | |
470 | * | |
471 | * Arguments: @mp *d@ = destination | |
ef5f4810 | 472 | * @mp *a, *b@ = sources |
d3409d5e | 473 | * |
474 | * Returns: Result, @a@ multiplied by @b@. | |
475 | */ | |
476 | ||
ef5f4810 | 477 | mp *mp_mul(mp *d, mp *a, mp *b) |
d3409d5e | 478 | { |
ef5f4810 | 479 | a = MP_COPY(a); |
480 | b = MP_COPY(b); | |
481 | ||
52cdaca9 | 482 | if (MP_LEN(a) <= MPK_THRESH || MP_LEN(b) <= MPK_THRESH) { |
d34decd2 | 483 | MP_DEST(d, MP_LEN(a) + MP_LEN(b), a->f | b->f | MP_UNDEF); |
ef5f4810 | 484 | mpx_umul(d->v, d->vl, a->v, a->vl, b->v, b->vl); |
8017495b | 485 | } else { |
dd22938e | 486 | size_t m = MAX(MP_LEN(a), MP_LEN(b)); |
ef5f4810 | 487 | mpw *s; |
dd22938e | 488 | MP_DEST(d, 3 * m, a->f | b->f | MP_UNDEF); |
489 | s = mpalloc(d->a, 5 * m); | |
490 | mpx_kmul(d->v, d->vl, a->v, a->vl, b->v, b->vl, s, s + 5 * m); | |
d34decd2 | 491 | mpfree(d->a, s); |
ef5f4810 | 492 | } |
493 | ||
d3409d5e | 494 | d->f = ((a->f | b->f) & MP_BURN) | ((a->f ^ b->f) & MP_NEG); |
495 | MP_SHRINK(d); | |
ef5f4810 | 496 | MP_DROP(a); |
497 | MP_DROP(b); | |
d3409d5e | 498 | return (d); |
499 | } | |
500 | ||
501 | /* --- @mp_sqr@ --- * | |
502 | * | |
503 | * Arguments: @mp *d@ = destination | |
ef5f4810 | 504 | * @mp *a@ = source |
d3409d5e | 505 | * |
506 | * Returns: Result, @a@ squared. | |
507 | */ | |
508 | ||
ef5f4810 | 509 | mp *mp_sqr(mp *d, mp *a) |
d3409d5e | 510 | { |
ef5f4810 | 511 | size_t m = MP_LEN(a); |
512 | ||
513 | a = MP_COPY(a); | |
52cdaca9 | 514 | if (m > MPK_THRESH) { |
ef5f4810 | 515 | mpw *s; |
dd22938e | 516 | MP_DEST(d, 3 * m, a->f | MP_UNDEF); |
517 | s = mpalloc(d->a, 5 * m); | |
518 | mpx_ksqr(d->v, d->vl, a->v, a->vl, s, s + 5 * m); | |
d34decd2 | 519 | mpfree(d->a, s); |
dd22938e | 520 | } else { |
521 | MP_DEST(d, 2 * m + 2, a->f | MP_UNDEF); | |
ef5f4810 | 522 | mpx_usqr(d->v, d->vl, a->v, a->vl); |
dd22938e | 523 | } |
d3409d5e | 524 | d->f = a->f & MP_BURN; |
525 | MP_SHRINK(d); | |
ef5f4810 | 526 | MP_DROP(a); |
d3409d5e | 527 | return (d); |
528 | } | |
529 | ||
530 | /* --- @mp_div@ --- * | |
531 | * | |
532 | * Arguments: @mp **qq, **rr@ = destination, quotient and remainder | |
ef5f4810 | 533 | * @mp *a, *b@ = sources |
d3409d5e | 534 | * |
535 | * Use: Calculates the quotient and remainder when @a@ is divided by | |
536 | * @b@. The destinations @*qq@ and @*rr@ must be distinct. | |
537 | * Either of @qq@ or @rr@ may be null to indicate that the | |
538 | * result is irrelevant. (Discarding both results is silly.) | |
539 | * There is a performance advantage if @a == *rr@. | |
540 | * | |
541 | * The behaviour when @a@ and @b@ have the same sign is | |
542 | * straightforward. When the signs differ, this implementation | |
543 | * chooses @r@ to have the same sign as @b@, rather than the | |
544 | * more normal choice that the remainder has the same sign as | |
545 | * the dividend. This makes modular arithmetic a little more | |
546 | * straightforward. | |
547 | */ | |
548 | ||
ef5f4810 | 549 | void mp_div(mp **qq, mp **rr, mp *a, mp *b) |
d3409d5e | 550 | { |
551 | mp *r = rr ? *rr : MP_NEW; | |
552 | mp *q = qq ? *qq : MP_NEW; | |
553 | mpw *sv, *svl; | |
554 | ||
d3409d5e | 555 | /* --- Set the remainder up right --- * |
556 | * | |
557 | * Just in case the divisor is larger, be able to cope with this. It's not | |
558 | * important in @mpx_udiv@, but it is here because of the sign correction. | |
559 | */ | |
560 | ||
d34decd2 | 561 | b = MP_COPY(b); |
562 | a = MP_COPY(a); | |
563 | if (r) | |
564 | MP_DROP(r); | |
565 | r = a; | |
a9c8f3d6 | 566 | MP_DEST(r, MAX(MP_LEN(a), MP_LEN(b)) + 2, a->f | b->f); |
d3409d5e | 567 | |
568 | /* --- Fix up the quotient too --- */ | |
569 | ||
d34decd2 | 570 | r = MP_COPY(r); |
571 | MP_DEST(q, MP_LEN(r), r->f | MP_UNDEF); | |
572 | MP_DROP(r); | |
573 | ||
574 | /* --- Set up some temporary workspace --- */ | |
575 | ||
576 | { | |
577 | size_t rq = MP_LEN(b) + 1; | |
578 | sv = mpalloc(r->a, rq); | |
579 | svl = sv + rq; | |
580 | } | |
d3409d5e | 581 | |
582 | /* --- Perform the calculation --- */ | |
583 | ||
584 | mpx_udiv(q->v, q->vl, r->v, r->vl, b->v, b->vl, sv, svl); | |
585 | ||
586 | /* --- Sort out the sign of the results --- * | |
587 | * | |
588 | * If the signs of the arguments differ, and the remainder is nonzero, I | |
589 | * must add one to the absolute value of the quotient and subtract the | |
590 | * remainder from @b@. | |
591 | */ | |
592 | ||
d34decd2 | 593 | q->f = ((r->f | b->f) & MP_BURN) | ((r->f ^ b->f) & MP_NEG); |
a69a3efd | 594 | if (MP_NEGP(q)) { |
ef5f4810 | 595 | mpw *v; |
596 | for (v = r->v; v < r->vl; v++) { | |
d3409d5e | 597 | if (*v) { |
598 | MPX_UADDN(q->v, q->vl, 1); | |
599 | mpx_usub(r->v, r->vl, b->v, b->vl, r->v, r->vl); | |
600 | break; | |
601 | } | |
602 | } | |
603 | } | |
604 | ||
d34decd2 | 605 | r->f = ((r->f | b->f) & MP_BURN) | (b->f & MP_NEG); |
d3409d5e | 606 | |
607 | /* --- Store the return values --- */ | |
608 | ||
d34decd2 | 609 | mpfree(r->a, sv); |
610 | MP_DROP(b); | |
611 | ||
d3409d5e | 612 | if (!qq) |
613 | MP_DROP(q); | |
614 | else { | |
615 | MP_SHRINK(q); | |
616 | *qq = q; | |
617 | } | |
618 | ||
619 | if (!rr) | |
620 | MP_DROP(r); | |
621 | else { | |
622 | MP_SHRINK(r); | |
623 | *rr = r; | |
624 | } | |
d3409d5e | 625 | } |
626 | ||
f1713c63 | 627 | /* --- @mp_odd@ --- * |
628 | * | |
629 | * Arguments: @mp *d@ = pointer to destination integer | |
630 | * @mp *m@ = pointer to source integer | |
631 | * @size_t *s@ = where to store the power of 2 | |
632 | * | |
633 | * Returns: An odd integer integer %$t$% such that %$m = 2^s t$%. | |
634 | * | |
635 | * Use: Computes a power of two and an odd integer which, when | |
636 | * multiplied, give a specified result. This sort of thing is | |
637 | * useful in number theory quite often. | |
638 | */ | |
639 | ||
640 | mp *mp_odd(mp *d, mp *m, size_t *s) | |
641 | { | |
642 | size_t ss = 0; | |
643 | const mpw *v, *vl; | |
644 | ||
645 | v = m->v; | |
646 | vl = m->vl; | |
647 | for (; !*v && v < vl; v++) | |
648 | ss += MPW_BITS; | |
649 | if (v >= vl) | |
650 | ss = 0; | |
651 | else { | |
652 | mpw x = *v; | |
c29970a7 MW |
653 | unsigned z = MPW_P2; |
654 | mpw mask = ((mpw)1 << z) - 1; | |
f1713c63 | 655 | |
656 | while (z) { | |
f1713c63 | 657 | if (!(x & mask)) { |
658 | x >>= z; | |
659 | ss += z; | |
660 | } | |
661 | z >>= 1; | |
c29970a7 | 662 | mask >>= z; |
f1713c63 | 663 | } |
664 | } | |
665 | ||
666 | *s = ss; | |
667 | return (mp_lsr(d, m, ss)); | |
668 | } | |
669 | ||
0c4f06c0 MW |
670 | /* --- @mp_leastcongruent@ --- * |
671 | * | |
672 | * Arguments: @mp *d@ = pointer to destination | |
673 | * @mp *b@ = lower bound | |
674 | * @mp *r@ = representative | |
675 | * @mp *m@ = modulus | |
676 | * | |
677 | * Returns: The smallest integer %$x \equiv r \pmod{m}$% such that | |
678 | * %$x \ge b$%. | |
679 | */ | |
680 | ||
681 | mp *mp_leastcongruent(mp *d, mp *b, mp *r, mp *m) | |
682 | { | |
683 | /* --- Strategy --- * | |
684 | * | |
685 | * Start by finding %$u \equiv b - r \pmod{m}$% with %$0 < u \le m$%. Then | |
686 | * %$b \le x = b + m - u < b + m$%, and %$x \equiv r \pmod{m}$% as | |
687 | * required. | |
688 | */ | |
689 | ||
690 | MP_COPY(b); MP_COPY(m); | |
691 | d = mp_sub(d, b, r); | |
692 | mp_div(0, &d, d, m); | |
693 | if (MP_ZEROP(d)) { MP_DROP(d); d = MP_COPY(b); } | |
694 | else { d = mp_sub(d, b, d); d = mp_add(d, d, m); } | |
695 | MP_DROP(b); MP_DROP(m); | |
696 | return (d); | |
697 | } | |
698 | ||
d3409d5e | 699 | /*----- Test rig ----------------------------------------------------------*/ |
700 | ||
701 | #ifdef TEST_RIG | |
702 | ||
703 | static int verify(const char *op, mp *expect, mp *result, mp *a, mp *b) | |
704 | { | |
4b536f42 | 705 | if (!MP_EQ(expect, result)) { |
d3409d5e | 706 | fprintf(stderr, "\n*** %s failed", op); |
45c0fd36 MW |
707 | fputs("\n*** a = ", stderr); mp_writefile(a, stderr, 10); |
708 | fputs("\n*** b = ", stderr); mp_writefile(b, stderr, 10); | |
d3409d5e | 709 | fputs("\n*** result = ", stderr); mp_writefile(result, stderr, 10); |
710 | fputs("\n*** expect = ", stderr); mp_writefile(expect, stderr, 10); | |
711 | fputc('\n', stderr); | |
712 | return (0); | |
713 | } | |
714 | return (1); | |
715 | } | |
716 | ||
717 | #define RIG(name, op) \ | |
ef5f4810 | 718 | static int t##name(dstr *v) \ |
d3409d5e | 719 | { \ |
720 | mp *a = *(mp **)v[0].buf; \ | |
721 | mpw n = *(int *)v[1].buf; \ | |
722 | mp b; \ | |
723 | mp *r = *(mp **)v[2].buf; \ | |
724 | mp *c = op(MP_NEW, a, n); \ | |
725 | int ok; \ | |
726 | mp_build(&b, &n, &n + 1); \ | |
727 | ok = verify(#name, r, c, a, &b); \ | |
728 | mp_drop(a); mp_drop(c); mp_drop(r); \ | |
ef5f4810 | 729 | assert(mparena_count(MPARENA_GLOBAL) == 0); \ |
d3409d5e | 730 | return (ok); \ |
731 | } | |
732 | ||
733 | RIG(lsl, mp_lsl) | |
734 | RIG(lsr, mp_lsr) | |
f09e814a | 735 | RIG(lsl2c, mp_lsl2c) |
736 | RIG(lsr2c, mp_lsr2c) | |
d3409d5e | 737 | |
738 | #undef RIG | |
739 | ||
740 | #define RIG(name, op) \ | |
ef5f4810 | 741 | static int t##name(dstr *v) \ |
d3409d5e | 742 | { \ |
743 | mp *a = *(mp **)v[0].buf; \ | |
744 | mp *b = *(mp **)v[1].buf; \ | |
745 | mp *r = *(mp **)v[2].buf; \ | |
746 | mp *c = op(MP_NEW, a, b); \ | |
747 | int ok = verify(#name, r, c, a, b); \ | |
748 | mp_drop(a); mp_drop(b); mp_drop(c); mp_drop(r); \ | |
ef5f4810 | 749 | assert(mparena_count(MPARENA_GLOBAL) == 0); \ |
d3409d5e | 750 | return (ok); \ |
751 | } | |
752 | ||
753 | RIG(add, mp_add) | |
754 | RIG(sub, mp_sub) | |
755 | RIG(mul, mp_mul) | |
f4535c64 | 756 | RIG(exp, mp_exp) |
d3409d5e | 757 | |
758 | #undef RIG | |
759 | ||
760 | static int tdiv(dstr *v) | |
761 | { | |
762 | mp *a = *(mp **)v[0].buf; | |
763 | mp *b = *(mp **)v[1].buf; | |
764 | mp *q = *(mp **)v[2].buf; | |
765 | mp *r = *(mp **)v[3].buf; | |
766 | mp *c = MP_NEW, *d = MP_NEW; | |
767 | int ok = 1; | |
768 | mp_div(&c, &d, a, b); | |
769 | ok &= verify("div(quotient)", q, c, a, b); | |
770 | ok &= verify("div(remainder)", r, d, a, b); | |
771 | mp_drop(a); mp_drop(b); mp_drop(c); mp_drop(d); mp_drop(r); mp_drop(q); | |
ef5f4810 | 772 | assert(mparena_count(MPARENA_GLOBAL) == 0); |
d3409d5e | 773 | return (ok); |
774 | } | |
775 | ||
f09e814a | 776 | static int tbin(dstr *v) |
777 | { | |
778 | static mp *(*fn[])(mp *, mp *, mp *) = { | |
45c0fd36 | 779 | #define DO(string) mp_bit##string##2c, |
f09e814a | 780 | MPX_DOBIN(DO) |
781 | #undef DO | |
782 | }; | |
783 | int ok = 1; | |
784 | unsigned op = 0; | |
785 | mp *a = *(mp **)v[1].buf; | |
786 | mp *b = *(mp **)v[2].buf; | |
787 | mp *r = *(mp **)v[3].buf; | |
788 | mp *c; | |
45c0fd36 | 789 | |
f09e814a | 790 | if (strcmp(v[0].buf, "and") == 0) op = 1; |
791 | else if (strcmp(v[0].buf, "or") == 0) op = 7; | |
792 | else if (strcmp(v[0].buf, "nand") == 0) op = 14; | |
793 | else if (strcmp(v[0].buf, "nor") == 0) op = 8; | |
794 | else if (strcmp(v[0].buf, "xor") == 0) op = 6; | |
795 | else { | |
796 | char *p = v[0].buf; | |
797 | while (*p) { | |
798 | op <<= 1; | |
799 | if (*p++ == '1') | |
800 | op |= 1; | |
801 | } | |
802 | } | |
803 | ||
804 | c = fn[op](MP_NEW, a, b); | |
805 | ok = verify(v[0].buf, r, c, a, b); | |
806 | mp_drop(a); mp_drop(b); mp_drop(r); mp_drop(c); | |
807 | assert(mparena_count(MPARENA_GLOBAL) == 0); | |
808 | return (ok); | |
809 | } | |
810 | ||
09d00c6b | 811 | static int tset(dstr *v) |
812 | { | |
813 | mp *a = *(mp **)v[0].buf; | |
814 | unsigned long n = *(unsigned long *)v[1].buf; | |
815 | mp *r = *(mp **)v[2].buf; | |
816 | mp *c; | |
817 | int ok = 1; | |
818 | ||
819 | c = mp_setbit2c(MP_NEW, a, n); | |
820 | if (!MP_EQ(c, r)) { | |
821 | ok = 0; | |
822 | fprintf(stderr, "\n***setbit (set) failed"); | |
823 | fputs("\n*** a = ", stderr); mp_writefile(a, stderr, 16); | |
824 | fprintf(stderr, "\n*** n = %lu", n); | |
825 | fputs("\n*** r = ", stderr); mp_writefile(r, stderr, 16); | |
826 | fputs("\n*** c = ", stderr); mp_writefile(c, stderr, 16); | |
827 | fputc('\n', stderr); | |
828 | } | |
829 | if (!mp_testbit2c(r, n)) { | |
830 | ok = 0; | |
831 | fprintf(stderr, "\n***setbit (test) failed"); | |
832 | fprintf(stderr, "\n*** n = %lu", n); | |
833 | fputs("\n*** r = ", stderr); mp_writefile(r, stderr, 16); | |
834 | fputc('\n', stderr); | |
835 | } | |
836 | mp_drop(a); | |
837 | mp_drop(r); | |
838 | mp_drop(c); | |
839 | assert(mparena_count(MPARENA_GLOBAL) == 0); | |
840 | return (ok); | |
841 | } | |
842 | ||
843 | static int tclr(dstr *v) | |
844 | { | |
845 | mp *a = *(mp **)v[0].buf; | |
846 | unsigned long n = *(unsigned long *)v[1].buf; | |
847 | mp *r = *(mp **)v[2].buf; | |
848 | mp *c; | |
849 | int ok = 1; | |
850 | ||
851 | c = mp_clearbit2c(MP_NEW, a, n); | |
852 | if (!MP_EQ(c, r)) { | |
853 | ok = 0; | |
854 | fprintf(stderr, "\n***clrbit (set) failed"); | |
855 | fputs("\n*** a = ", stderr); mp_writefile(a, stderr, 16); | |
856 | fprintf(stderr, "\n*** n = %lu", n); | |
857 | fputs("\n*** r = ", stderr); mp_writefile(r, stderr, 16); | |
858 | fputs("\n*** c = ", stderr); mp_writefile(c, stderr, 16); | |
859 | fputc('\n', stderr); | |
860 | } | |
861 | if (mp_testbit2c(r, n)) { | |
862 | ok = 0; | |
863 | fprintf(stderr, "\n***clrbit (test) failed"); | |
864 | fprintf(stderr, "\n*** n = %lu", n); | |
865 | fputs("\n*** r = ", stderr); mp_writefile(r, stderr, 16); | |
866 | fputc('\n', stderr); | |
867 | } | |
868 | mp_drop(a); | |
869 | mp_drop(c); | |
870 | mp_drop(r); | |
871 | assert(mparena_count(MPARENA_GLOBAL) == 0); | |
872 | return (ok); | |
873 | } | |
874 | ||
397041a9 | 875 | static int tneg(dstr *v) |
876 | { | |
877 | mp *a = *(mp **)v[0].buf; | |
878 | mp *r = *(mp **)v[1].buf; | |
879 | int ok = 1; | |
880 | mp *n = mp_neg(MP_NEW, a); | |
881 | if (!MP_EQ(r, n)) { | |
882 | ok = 0; | |
883 | fprintf(stderr, "\n*** neg failed\n"); | |
884 | fputs("\n*** a = ", stderr); mp_writefile(a, stderr, 10); | |
885 | fputs("\n*** r = ", stderr); mp_writefile(r, stderr, 10); | |
886 | fputs("\n*** n = ", stderr); mp_writefile(n, stderr, 10); | |
887 | fputc('\n', stderr); | |
888 | } | |
889 | mp_drop(n); | |
890 | n = mp_neg(a, a); | |
891 | if (!MP_EQ(r, n)) { | |
892 | ok = 0; | |
893 | fprintf(stderr, "\n*** neg failed\n"); | |
894 | fputs("\n*** a* = ", stderr); mp_writefile(a, stderr, 10); | |
895 | fputs("\n*** r = ", stderr); mp_writefile(r, stderr, 10); | |
896 | fputs("\n*** n = ", stderr); mp_writefile(n, stderr, 10); | |
897 | fputc('\n', stderr); | |
898 | } | |
899 | mp_drop(a); | |
900 | mp_drop(r); | |
901 | assert(mparena_count(MPARENA_GLOBAL) == 0); | |
45c0fd36 | 902 | return (ok); |
397041a9 | 903 | } |
904 | ||
f1713c63 | 905 | static int todd(dstr *v) |
906 | { | |
907 | mp *a = *(mp **)v[0].buf; | |
908 | size_t rs = *(uint32 *)v[1].buf; | |
909 | mp *rt = *(mp **)v[2].buf; | |
910 | int ok = 1; | |
911 | mp *t; | |
912 | size_t s; | |
913 | t = mp_odd(MP_NEW, a, &s); | |
4b536f42 | 914 | if (s != rs || !MP_EQ(t, rt)) { |
f1713c63 | 915 | ok = 0; |
916 | fprintf(stderr, "\n*** odd failed"); | |
917 | fputs("\n*** a = ", stderr); mp_writefile(a, stderr, 10); | |
918 | fprintf(stderr, "\n*** s = %lu", (unsigned long)s); | |
919 | fputs("\n*** t = ", stderr); mp_writefile(t, stderr, 10); | |
920 | fprintf(stderr, "\n*** rs = %lu", (unsigned long)rs); | |
921 | fputs("\n*** rt = ", stderr); mp_writefile(rt, stderr, 10); | |
922 | fputc('\n', stderr); | |
923 | } | |
924 | mp_drop(a); | |
925 | mp_drop(rt); | |
926 | mp_drop(t); | |
09d00c6b | 927 | assert(mparena_count(MPARENA_GLOBAL) == 0); |
f1713c63 | 928 | return (ok); |
929 | } | |
930 | ||
d3409d5e | 931 | static test_chunk tests[] = { |
f09e814a | 932 | { "lsl", tlsl, { &type_mp, &type_int, &type_mp, 0 } }, |
933 | { "lsr", tlsr, { &type_mp, &type_int, &type_mp, 0 } }, | |
934 | { "lsl2c", tlsl2c, { &type_mp, &type_int, &type_mp, 0 } }, | |
935 | { "lsr2c", tlsr2c, { &type_mp, &type_int, &type_mp, 0 } }, | |
09d00c6b | 936 | { "setbit", tset, { &type_mp, &type_ulong, &type_mp, 0 } }, |
937 | { "clrbit", tclr, { &type_mp, &type_ulong, &type_mp, 0 } }, | |
d3409d5e | 938 | { "add", tadd, { &type_mp, &type_mp, &type_mp, 0 } }, |
939 | { "sub", tsub, { &type_mp, &type_mp, &type_mp, 0 } }, | |
940 | { "mul", tmul, { &type_mp, &type_mp, &type_mp, 0 } }, | |
941 | { "div", tdiv, { &type_mp, &type_mp, &type_mp, &type_mp, 0 } }, | |
f4535c64 | 942 | { "exp", texp, { &type_mp, &type_mp, &type_mp, 0 } }, |
f09e814a | 943 | { "bin2c", tbin, { &type_string, &type_mp, &type_mp, &type_mp, 0 } }, |
f1713c63 | 944 | { "odd", todd, { &type_mp, &type_uint32, &type_mp, 0 } }, |
397041a9 | 945 | { "neg", tneg, { &type_mp, &type_mp, 0 } }, |
d3409d5e | 946 | { 0, 0, { 0 } }, |
947 | }; | |
948 | ||
949 | int main(int argc, char *argv[]) | |
950 | { | |
951 | sub_init(); | |
0f00dc4c | 952 | test_run(argc, argv, tests, SRCDIR "/t/mp"); |
d3409d5e | 953 | return (0); |
954 | } | |
955 | ||
956 | #endif | |
957 | ||
958 | /*----- That's all, folks -------------------------------------------------*/ |