| 1 | /*
|
|---|
| 2 | * Copyright 1995-2016 The OpenSSL Project Authors. All Rights Reserved.
|
|---|
| 3 | *
|
|---|
| 4 | * Licensed under the Apache License 2.0 (the "License"). You may not use
|
|---|
| 5 | * this file except in compliance with the License. You can obtain a copy
|
|---|
| 6 | * in the file LICENSE in the source distribution or at
|
|---|
| 7 | * https://www.openssl.org/source/license.html
|
|---|
| 8 | */
|
|---|
| 9 |
|
|---|
| 10 | #include <assert.h>
|
|---|
| 11 | #include <openssl/crypto.h>
|
|---|
| 12 | #include "internal/cryptlib.h"
|
|---|
| 13 | #include "bn_local.h"
|
|---|
| 14 |
|
|---|
| 15 | #if defined(BN_LLONG) || defined(BN_UMULT_HIGH)
|
|---|
| 16 |
|
|---|
| 17 | BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num,
|
|---|
| 18 | BN_ULONG w)
|
|---|
| 19 | {
|
|---|
| 20 | BN_ULONG c1 = 0;
|
|---|
| 21 |
|
|---|
| 22 | assert(num >= 0);
|
|---|
| 23 | if (num <= 0)
|
|---|
| 24 | return c1;
|
|---|
| 25 |
|
|---|
| 26 | # ifndef OPENSSL_SMALL_FOOTPRINT
|
|---|
| 27 | while (num & ~3) {
|
|---|
| 28 | mul_add(rp[0], ap[0], w, c1);
|
|---|
| 29 | mul_add(rp[1], ap[1], w, c1);
|
|---|
| 30 | mul_add(rp[2], ap[2], w, c1);
|
|---|
| 31 | mul_add(rp[3], ap[3], w, c1);
|
|---|
| 32 | ap += 4;
|
|---|
| 33 | rp += 4;
|
|---|
| 34 | num -= 4;
|
|---|
| 35 | }
|
|---|
| 36 | # endif
|
|---|
| 37 | while (num) {
|
|---|
| 38 | mul_add(rp[0], ap[0], w, c1);
|
|---|
| 39 | ap++;
|
|---|
| 40 | rp++;
|
|---|
| 41 | num--;
|
|---|
| 42 | }
|
|---|
| 43 |
|
|---|
| 44 | return c1;
|
|---|
| 45 | }
|
|---|
| 46 |
|
|---|
| 47 | BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
|
|---|
| 48 | {
|
|---|
| 49 | BN_ULONG c1 = 0;
|
|---|
| 50 |
|
|---|
| 51 | assert(num >= 0);
|
|---|
| 52 | if (num <= 0)
|
|---|
| 53 | return c1;
|
|---|
| 54 |
|
|---|
| 55 | # ifndef OPENSSL_SMALL_FOOTPRINT
|
|---|
| 56 | while (num & ~3) {
|
|---|
| 57 | mul(rp[0], ap[0], w, c1);
|
|---|
| 58 | mul(rp[1], ap[1], w, c1);
|
|---|
| 59 | mul(rp[2], ap[2], w, c1);
|
|---|
| 60 | mul(rp[3], ap[3], w, c1);
|
|---|
| 61 | ap += 4;
|
|---|
| 62 | rp += 4;
|
|---|
| 63 | num -= 4;
|
|---|
| 64 | }
|
|---|
| 65 | # endif
|
|---|
| 66 | while (num) {
|
|---|
| 67 | mul(rp[0], ap[0], w, c1);
|
|---|
| 68 | ap++;
|
|---|
| 69 | rp++;
|
|---|
| 70 | num--;
|
|---|
| 71 | }
|
|---|
| 72 | return c1;
|
|---|
| 73 | }
|
|---|
| 74 |
|
|---|
| 75 | void bn_sqr_words(BN_ULONG *r, const BN_ULONG *a, int n)
|
|---|
| 76 | {
|
|---|
| 77 | assert(n >= 0);
|
|---|
| 78 | if (n <= 0)
|
|---|
| 79 | return;
|
|---|
| 80 |
|
|---|
| 81 | # ifndef OPENSSL_SMALL_FOOTPRINT
|
|---|
| 82 | while (n & ~3) {
|
|---|
| 83 | sqr(r[0], r[1], a[0]);
|
|---|
| 84 | sqr(r[2], r[3], a[1]);
|
|---|
| 85 | sqr(r[4], r[5], a[2]);
|
|---|
| 86 | sqr(r[6], r[7], a[3]);
|
|---|
| 87 | a += 4;
|
|---|
| 88 | r += 8;
|
|---|
| 89 | n -= 4;
|
|---|
| 90 | }
|
|---|
| 91 | # endif
|
|---|
| 92 | while (n) {
|
|---|
| 93 | sqr(r[0], r[1], a[0]);
|
|---|
| 94 | a++;
|
|---|
| 95 | r += 2;
|
|---|
| 96 | n--;
|
|---|
| 97 | }
|
|---|
| 98 | }
|
|---|
| 99 |
|
|---|
| 100 | #else /* !(defined(BN_LLONG) ||
|
|---|
| 101 | * defined(BN_UMULT_HIGH)) */
|
|---|
| 102 |
|
|---|
| 103 | BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num,
|
|---|
| 104 | BN_ULONG w)
|
|---|
| 105 | {
|
|---|
| 106 | BN_ULONG c = 0;
|
|---|
| 107 | BN_ULONG bl, bh;
|
|---|
| 108 |
|
|---|
| 109 | assert(num >= 0);
|
|---|
| 110 | if (num <= 0)
|
|---|
| 111 | return (BN_ULONG)0;
|
|---|
| 112 |
|
|---|
| 113 | bl = LBITS(w);
|
|---|
| 114 | bh = HBITS(w);
|
|---|
| 115 |
|
|---|
| 116 | # ifndef OPENSSL_SMALL_FOOTPRINT
|
|---|
| 117 | while (num & ~3) {
|
|---|
| 118 | mul_add(rp[0], ap[0], bl, bh, c);
|
|---|
| 119 | mul_add(rp[1], ap[1], bl, bh, c);
|
|---|
| 120 | mul_add(rp[2], ap[2], bl, bh, c);
|
|---|
| 121 | mul_add(rp[3], ap[3], bl, bh, c);
|
|---|
| 122 | ap += 4;
|
|---|
| 123 | rp += 4;
|
|---|
| 124 | num -= 4;
|
|---|
| 125 | }
|
|---|
| 126 | # endif
|
|---|
| 127 | while (num) {
|
|---|
| 128 | mul_add(rp[0], ap[0], bl, bh, c);
|
|---|
| 129 | ap++;
|
|---|
| 130 | rp++;
|
|---|
| 131 | num--;
|
|---|
| 132 | }
|
|---|
| 133 | return c;
|
|---|
| 134 | }
|
|---|
| 135 |
|
|---|
| 136 | BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
|
|---|
| 137 | {
|
|---|
| 138 | BN_ULONG carry = 0;
|
|---|
| 139 | BN_ULONG bl, bh;
|
|---|
| 140 |
|
|---|
| 141 | assert(num >= 0);
|
|---|
| 142 | if (num <= 0)
|
|---|
| 143 | return (BN_ULONG)0;
|
|---|
| 144 |
|
|---|
| 145 | bl = LBITS(w);
|
|---|
| 146 | bh = HBITS(w);
|
|---|
| 147 |
|
|---|
| 148 | # ifndef OPENSSL_SMALL_FOOTPRINT
|
|---|
| 149 | while (num & ~3) {
|
|---|
| 150 | mul(rp[0], ap[0], bl, bh, carry);
|
|---|
| 151 | mul(rp[1], ap[1], bl, bh, carry);
|
|---|
| 152 | mul(rp[2], ap[2], bl, bh, carry);
|
|---|
| 153 | mul(rp[3], ap[3], bl, bh, carry);
|
|---|
| 154 | ap += 4;
|
|---|
| 155 | rp += 4;
|
|---|
| 156 | num -= 4;
|
|---|
| 157 | }
|
|---|
| 158 | # endif
|
|---|
| 159 | while (num) {
|
|---|
| 160 | mul(rp[0], ap[0], bl, bh, carry);
|
|---|
| 161 | ap++;
|
|---|
| 162 | rp++;
|
|---|
| 163 | num--;
|
|---|
| 164 | }
|
|---|
| 165 | return carry;
|
|---|
| 166 | }
|
|---|
| 167 |
|
|---|
| 168 | void bn_sqr_words(BN_ULONG *r, const BN_ULONG *a, int n)
|
|---|
| 169 | {
|
|---|
| 170 | assert(n >= 0);
|
|---|
| 171 | if (n <= 0)
|
|---|
| 172 | return;
|
|---|
| 173 |
|
|---|
| 174 | # ifndef OPENSSL_SMALL_FOOTPRINT
|
|---|
| 175 | while (n & ~3) {
|
|---|
| 176 | sqr64(r[0], r[1], a[0]);
|
|---|
| 177 | sqr64(r[2], r[3], a[1]);
|
|---|
| 178 | sqr64(r[4], r[5], a[2]);
|
|---|
| 179 | sqr64(r[6], r[7], a[3]);
|
|---|
| 180 | a += 4;
|
|---|
| 181 | r += 8;
|
|---|
| 182 | n -= 4;
|
|---|
| 183 | }
|
|---|
| 184 | # endif
|
|---|
| 185 | while (n) {
|
|---|
| 186 | sqr64(r[0], r[1], a[0]);
|
|---|
| 187 | a++;
|
|---|
| 188 | r += 2;
|
|---|
| 189 | n--;
|
|---|
| 190 | }
|
|---|
| 191 | }
|
|---|
| 192 |
|
|---|
| 193 | #endif /* !(defined(BN_LLONG) ||
|
|---|
| 194 | * defined(BN_UMULT_HIGH)) */
|
|---|
| 195 |
|
|---|
| 196 | #if defined(BN_LLONG) && defined(BN_DIV2W)
|
|---|
| 197 |
|
|---|
| 198 | BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
|
|---|
| 199 | {
|
|---|
| 200 | return ((BN_ULONG)(((((BN_ULLONG) h) << BN_BITS2) | l) / (BN_ULLONG) d));
|
|---|
| 201 | }
|
|---|
| 202 |
|
|---|
| 203 | #else
|
|---|
| 204 |
|
|---|
| 205 | /* Divide h,l by d and return the result. */
|
|---|
| 206 | /* I need to test this some more :-( */
|
|---|
| 207 | BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
|
|---|
| 208 | {
|
|---|
| 209 | BN_ULONG dh, dl, q, ret = 0, th, tl, t;
|
|---|
| 210 | int i, count = 2;
|
|---|
| 211 |
|
|---|
| 212 | if (d == 0)
|
|---|
| 213 | return BN_MASK2;
|
|---|
| 214 |
|
|---|
| 215 | i = BN_num_bits_word(d);
|
|---|
| 216 | assert((i == BN_BITS2) || (h <= (BN_ULONG)1 << i));
|
|---|
| 217 |
|
|---|
| 218 | i = BN_BITS2 - i;
|
|---|
| 219 | if (h >= d)
|
|---|
| 220 | h -= d;
|
|---|
| 221 |
|
|---|
| 222 | if (i) {
|
|---|
| 223 | d <<= i;
|
|---|
| 224 | h = (h << i) | (l >> (BN_BITS2 - i));
|
|---|
| 225 | l <<= i;
|
|---|
| 226 | }
|
|---|
| 227 | dh = (d & BN_MASK2h) >> BN_BITS4;
|
|---|
| 228 | dl = (d & BN_MASK2l);
|
|---|
| 229 | for (;;) {
|
|---|
| 230 | if ((h >> BN_BITS4) == dh)
|
|---|
| 231 | q = BN_MASK2l;
|
|---|
| 232 | else
|
|---|
| 233 | q = h / dh;
|
|---|
| 234 |
|
|---|
| 235 | th = q * dh;
|
|---|
| 236 | tl = dl * q;
|
|---|
| 237 | for (;;) {
|
|---|
| 238 | t = h - th;
|
|---|
| 239 | if ((t & BN_MASK2h) ||
|
|---|
| 240 | ((tl) <= ((t << BN_BITS4) | ((l & BN_MASK2h) >> BN_BITS4))))
|
|---|
| 241 | break;
|
|---|
| 242 | q--;
|
|---|
| 243 | th -= dh;
|
|---|
| 244 | tl -= dl;
|
|---|
| 245 | }
|
|---|
| 246 | t = (tl >> BN_BITS4);
|
|---|
| 247 | tl = (tl << BN_BITS4) & BN_MASK2h;
|
|---|
| 248 | th += t;
|
|---|
| 249 |
|
|---|
| 250 | if (l < tl)
|
|---|
| 251 | th++;
|
|---|
| 252 | l -= tl;
|
|---|
| 253 | if (h < th) {
|
|---|
| 254 | h += d;
|
|---|
| 255 | q--;
|
|---|
| 256 | }
|
|---|
| 257 | h -= th;
|
|---|
| 258 |
|
|---|
| 259 | if (--count == 0)
|
|---|
| 260 | break;
|
|---|
| 261 |
|
|---|
| 262 | ret = q << BN_BITS4;
|
|---|
| 263 | h = ((h << BN_BITS4) | (l >> BN_BITS4)) & BN_MASK2;
|
|---|
| 264 | l = (l & BN_MASK2l) << BN_BITS4;
|
|---|
| 265 | }
|
|---|
| 266 | ret |= q;
|
|---|
| 267 | return ret;
|
|---|
| 268 | }
|
|---|
| 269 | #endif /* !defined(BN_LLONG) && defined(BN_DIV2W) */
|
|---|
| 270 |
|
|---|
| 271 | #ifdef BN_LLONG
|
|---|
| 272 | BN_ULONG bn_add_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
|
|---|
| 273 | int n)
|
|---|
| 274 | {
|
|---|
| 275 | BN_ULLONG ll = 0;
|
|---|
| 276 |
|
|---|
| 277 | assert(n >= 0);
|
|---|
| 278 | if (n <= 0)
|
|---|
| 279 | return (BN_ULONG)0;
|
|---|
| 280 |
|
|---|
| 281 | # ifndef OPENSSL_SMALL_FOOTPRINT
|
|---|
| 282 | while (n & ~3) {
|
|---|
| 283 | ll += (BN_ULLONG) a[0] + b[0];
|
|---|
| 284 | r[0] = (BN_ULONG)ll & BN_MASK2;
|
|---|
| 285 | ll >>= BN_BITS2;
|
|---|
| 286 | ll += (BN_ULLONG) a[1] + b[1];
|
|---|
| 287 | r[1] = (BN_ULONG)ll & BN_MASK2;
|
|---|
| 288 | ll >>= BN_BITS2;
|
|---|
| 289 | ll += (BN_ULLONG) a[2] + b[2];
|
|---|
| 290 | r[2] = (BN_ULONG)ll & BN_MASK2;
|
|---|
| 291 | ll >>= BN_BITS2;
|
|---|
| 292 | ll += (BN_ULLONG) a[3] + b[3];
|
|---|
| 293 | r[3] = (BN_ULONG)ll & BN_MASK2;
|
|---|
| 294 | ll >>= BN_BITS2;
|
|---|
| 295 | a += 4;
|
|---|
| 296 | b += 4;
|
|---|
| 297 | r += 4;
|
|---|
| 298 | n -= 4;
|
|---|
| 299 | }
|
|---|
| 300 | # endif
|
|---|
| 301 | while (n) {
|
|---|
| 302 | ll += (BN_ULLONG) a[0] + b[0];
|
|---|
| 303 | r[0] = (BN_ULONG)ll & BN_MASK2;
|
|---|
| 304 | ll >>= BN_BITS2;
|
|---|
| 305 | a++;
|
|---|
| 306 | b++;
|
|---|
| 307 | r++;
|
|---|
| 308 | n--;
|
|---|
| 309 | }
|
|---|
| 310 | return (BN_ULONG)ll;
|
|---|
| 311 | }
|
|---|
| 312 | #else /* !BN_LLONG */
|
|---|
| 313 | BN_ULONG bn_add_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
|
|---|
| 314 | int n)
|
|---|
| 315 | {
|
|---|
| 316 | BN_ULONG c, l, t;
|
|---|
| 317 |
|
|---|
| 318 | assert(n >= 0);
|
|---|
| 319 | if (n <= 0)
|
|---|
| 320 | return (BN_ULONG)0;
|
|---|
| 321 |
|
|---|
| 322 | c = 0;
|
|---|
| 323 | # ifndef OPENSSL_SMALL_FOOTPRINT
|
|---|
| 324 | while (n & ~3) {
|
|---|
| 325 | t = a[0];
|
|---|
| 326 | t = (t + c) & BN_MASK2;
|
|---|
| 327 | c = (t < c);
|
|---|
| 328 | l = (t + b[0]) & BN_MASK2;
|
|---|
| 329 | c += (l < t);
|
|---|
| 330 | r[0] = l;
|
|---|
| 331 | t = a[1];
|
|---|
| 332 | t = (t + c) & BN_MASK2;
|
|---|
| 333 | c = (t < c);
|
|---|
| 334 | l = (t + b[1]) & BN_MASK2;
|
|---|
| 335 | c += (l < t);
|
|---|
| 336 | r[1] = l;
|
|---|
| 337 | t = a[2];
|
|---|
| 338 | t = (t + c) & BN_MASK2;
|
|---|
| 339 | c = (t < c);
|
|---|
| 340 | l = (t + b[2]) & BN_MASK2;
|
|---|
| 341 | c += (l < t);
|
|---|
| 342 | r[2] = l;
|
|---|
| 343 | t = a[3];
|
|---|
| 344 | t = (t + c) & BN_MASK2;
|
|---|
| 345 | c = (t < c);
|
|---|
| 346 | l = (t + b[3]) & BN_MASK2;
|
|---|
| 347 | c += (l < t);
|
|---|
| 348 | r[3] = l;
|
|---|
| 349 | a += 4;
|
|---|
| 350 | b += 4;
|
|---|
| 351 | r += 4;
|
|---|
| 352 | n -= 4;
|
|---|
| 353 | }
|
|---|
| 354 | # endif
|
|---|
| 355 | while (n) {
|
|---|
| 356 | t = a[0];
|
|---|
| 357 | t = (t + c) & BN_MASK2;
|
|---|
| 358 | c = (t < c);
|
|---|
| 359 | l = (t + b[0]) & BN_MASK2;
|
|---|
| 360 | c += (l < t);
|
|---|
| 361 | r[0] = l;
|
|---|
| 362 | a++;
|
|---|
| 363 | b++;
|
|---|
| 364 | r++;
|
|---|
| 365 | n--;
|
|---|
| 366 | }
|
|---|
| 367 | return (BN_ULONG)c;
|
|---|
| 368 | }
|
|---|
| 369 | #endif /* !BN_LLONG */
|
|---|
| 370 |
|
|---|
| 371 | BN_ULONG bn_sub_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
|
|---|
| 372 | int n)
|
|---|
| 373 | {
|
|---|
| 374 | BN_ULONG t1, t2;
|
|---|
| 375 | int c = 0;
|
|---|
| 376 |
|
|---|
| 377 | assert(n >= 0);
|
|---|
| 378 | if (n <= 0)
|
|---|
| 379 | return (BN_ULONG)0;
|
|---|
| 380 |
|
|---|
| 381 | #ifndef OPENSSL_SMALL_FOOTPRINT
|
|---|
| 382 | while (n & ~3) {
|
|---|
| 383 | t1 = a[0];
|
|---|
| 384 | t2 = (t1 - c) & BN_MASK2;
|
|---|
| 385 | c = (t2 > t1);
|
|---|
| 386 | t1 = b[0];
|
|---|
| 387 | t1 = (t2 - t1) & BN_MASK2;
|
|---|
| 388 | r[0] = t1;
|
|---|
| 389 | c += (t1 > t2);
|
|---|
| 390 | t1 = a[1];
|
|---|
| 391 | t2 = (t1 - c) & BN_MASK2;
|
|---|
| 392 | c = (t2 > t1);
|
|---|
| 393 | t1 = b[1];
|
|---|
| 394 | t1 = (t2 - t1) & BN_MASK2;
|
|---|
| 395 | r[1] = t1;
|
|---|
| 396 | c += (t1 > t2);
|
|---|
| 397 | t1 = a[2];
|
|---|
| 398 | t2 = (t1 - c) & BN_MASK2;
|
|---|
| 399 | c = (t2 > t1);
|
|---|
| 400 | t1 = b[2];
|
|---|
| 401 | t1 = (t2 - t1) & BN_MASK2;
|
|---|
| 402 | r[2] = t1;
|
|---|
| 403 | c += (t1 > t2);
|
|---|
| 404 | t1 = a[3];
|
|---|
| 405 | t2 = (t1 - c) & BN_MASK2;
|
|---|
| 406 | c = (t2 > t1);
|
|---|
| 407 | t1 = b[3];
|
|---|
| 408 | t1 = (t2 - t1) & BN_MASK2;
|
|---|
| 409 | r[3] = t1;
|
|---|
| 410 | c += (t1 > t2);
|
|---|
| 411 | a += 4;
|
|---|
| 412 | b += 4;
|
|---|
| 413 | r += 4;
|
|---|
| 414 | n -= 4;
|
|---|
| 415 | }
|
|---|
| 416 | #endif
|
|---|
| 417 | while (n) {
|
|---|
| 418 | t1 = a[0];
|
|---|
| 419 | t2 = (t1 - c) & BN_MASK2;
|
|---|
| 420 | c = (t2 > t1);
|
|---|
| 421 | t1 = b[0];
|
|---|
| 422 | t1 = (t2 - t1) & BN_MASK2;
|
|---|
| 423 | r[0] = t1;
|
|---|
| 424 | c += (t1 > t2);
|
|---|
| 425 | a++;
|
|---|
| 426 | b++;
|
|---|
| 427 | r++;
|
|---|
| 428 | n--;
|
|---|
| 429 | }
|
|---|
| 430 | return c;
|
|---|
| 431 | }
|
|---|
| 432 |
|
|---|
| 433 | #if defined(BN_MUL_COMBA) && !defined(OPENSSL_SMALL_FOOTPRINT)
|
|---|
| 434 |
|
|---|
| 435 | #ifndef ___openssl_mangling_h___ /* bird */
|
|---|
| 436 | # undef bn_mul_comba8
|
|---|
| 437 | # undef bn_mul_comba4
|
|---|
| 438 | # undef bn_sqr_comba8
|
|---|
| 439 | # undef bn_sqr_comba4
|
|---|
| 440 | #endif /* !___openssl_mangling_h___*/ /* bird */
|
|---|
| 441 |
|
|---|
| 442 | /* mul_add_c(a,b,c0,c1,c2) -- c+=a*b for three word number c=(c2,c1,c0) */
|
|---|
| 443 | /* mul_add_c2(a,b,c0,c1,c2) -- c+=2*a*b for three word number c=(c2,c1,c0) */
|
|---|
| 444 | /* sqr_add_c(a,i,c0,c1,c2) -- c+=a[i]^2 for three word number c=(c2,c1,c0) */
|
|---|
| 445 | /*
|
|---|
| 446 | * sqr_add_c2(a,i,c0,c1,c2) -- c+=2*a[i]*a[j] for three word number
|
|---|
| 447 | * c=(c2,c1,c0)
|
|---|
| 448 | */
|
|---|
| 449 |
|
|---|
| 450 | # ifdef BN_LLONG
|
|---|
| 451 | /*
|
|---|
| 452 | * Keep in mind that additions to multiplication result can not
|
|---|
| 453 | * overflow, because its high half cannot be all-ones.
|
|---|
| 454 | */
|
|---|
| 455 | # define mul_add_c(a,b,c0,c1,c2) do { \
|
|---|
| 456 | BN_ULONG hi; \
|
|---|
| 457 | BN_ULLONG t = (BN_ULLONG)(a)*(b); \
|
|---|
| 458 | t += c0; /* no carry */ \
|
|---|
| 459 | c0 = (BN_ULONG)Lw(t); \
|
|---|
| 460 | hi = (BN_ULONG)Hw(t); \
|
|---|
| 461 | c1 = (c1+hi)&BN_MASK2; c2 += (c1<hi); \
|
|---|
| 462 | } while(0)
|
|---|
| 463 |
|
|---|
| 464 | # define mul_add_c2(a,b,c0,c1,c2) do { \
|
|---|
| 465 | BN_ULONG hi; \
|
|---|
| 466 | BN_ULLONG t = (BN_ULLONG)(a)*(b); \
|
|---|
| 467 | BN_ULLONG tt = t+c0; /* no carry */ \
|
|---|
| 468 | c0 = (BN_ULONG)Lw(tt); \
|
|---|
| 469 | hi = (BN_ULONG)Hw(tt); \
|
|---|
| 470 | c1 = (c1+hi)&BN_MASK2; c2 += (c1<hi); \
|
|---|
| 471 | t += c0; /* no carry */ \
|
|---|
| 472 | c0 = (BN_ULONG)Lw(t); \
|
|---|
| 473 | hi = (BN_ULONG)Hw(t); \
|
|---|
| 474 | c1 = (c1+hi)&BN_MASK2; c2 += (c1<hi); \
|
|---|
| 475 | } while(0)
|
|---|
| 476 |
|
|---|
| 477 | # define sqr_add_c(a,i,c0,c1,c2) do { \
|
|---|
| 478 | BN_ULONG hi; \
|
|---|
| 479 | BN_ULLONG t = (BN_ULLONG)a[i]*a[i]; \
|
|---|
| 480 | t += c0; /* no carry */ \
|
|---|
| 481 | c0 = (BN_ULONG)Lw(t); \
|
|---|
| 482 | hi = (BN_ULONG)Hw(t); \
|
|---|
| 483 | c1 = (c1+hi)&BN_MASK2; c2 += (c1<hi); \
|
|---|
| 484 | } while(0)
|
|---|
| 485 |
|
|---|
| 486 | # define sqr_add_c2(a,i,j,c0,c1,c2) \
|
|---|
| 487 | mul_add_c2((a)[i],(a)[j],c0,c1,c2)
|
|---|
| 488 |
|
|---|
| 489 | # elif defined(BN_UMULT_LOHI)
|
|---|
| 490 | /*
|
|---|
| 491 | * Keep in mind that additions to hi can not overflow, because
|
|---|
| 492 | * the high word of a multiplication result cannot be all-ones.
|
|---|
| 493 | */
|
|---|
| 494 | # define mul_add_c(a,b,c0,c1,c2) do { \
|
|---|
| 495 | BN_ULONG ta = (a), tb = (b); \
|
|---|
| 496 | BN_ULONG lo, hi; \
|
|---|
| 497 | BN_UMULT_LOHI(lo,hi,ta,tb); \
|
|---|
| 498 | c0 += lo; hi += (c0<lo); \
|
|---|
| 499 | c1 += hi; c2 += (c1<hi); \
|
|---|
| 500 | } while(0)
|
|---|
| 501 |
|
|---|
| 502 | # define mul_add_c2(a,b,c0,c1,c2) do { \
|
|---|
| 503 | BN_ULONG ta = (a), tb = (b); \
|
|---|
| 504 | BN_ULONG lo, hi, tt; \
|
|---|
| 505 | BN_UMULT_LOHI(lo,hi,ta,tb); \
|
|---|
| 506 | c0 += lo; tt = hi + (c0<lo); \
|
|---|
| 507 | c1 += tt; c2 += (c1<tt); \
|
|---|
| 508 | c0 += lo; hi += (c0<lo); \
|
|---|
| 509 | c1 += hi; c2 += (c1<hi); \
|
|---|
| 510 | } while(0)
|
|---|
| 511 |
|
|---|
| 512 | # define sqr_add_c(a,i,c0,c1,c2) do { \
|
|---|
| 513 | BN_ULONG ta = (a)[i]; \
|
|---|
| 514 | BN_ULONG lo, hi; \
|
|---|
| 515 | BN_UMULT_LOHI(lo,hi,ta,ta); \
|
|---|
| 516 | c0 += lo; hi += (c0<lo); \
|
|---|
| 517 | c1 += hi; c2 += (c1<hi); \
|
|---|
| 518 | } while(0)
|
|---|
| 519 |
|
|---|
| 520 | # define sqr_add_c2(a,i,j,c0,c1,c2) \
|
|---|
| 521 | mul_add_c2((a)[i],(a)[j],c0,c1,c2)
|
|---|
| 522 |
|
|---|
| 523 | # elif defined(BN_UMULT_HIGH)
|
|---|
| 524 | /*
|
|---|
| 525 | * Keep in mind that additions to hi can not overflow, because
|
|---|
| 526 | * the high word of a multiplication result cannot be all-ones.
|
|---|
| 527 | */
|
|---|
| 528 | # define mul_add_c(a,b,c0,c1,c2) do { \
|
|---|
| 529 | BN_ULONG ta = (a), tb = (b); \
|
|---|
| 530 | BN_ULONG lo = ta * tb; \
|
|---|
| 531 | BN_ULONG hi = BN_UMULT_HIGH(ta,tb); \
|
|---|
| 532 | c0 += lo; hi += (c0<lo); \
|
|---|
| 533 | c1 += hi; c2 += (c1<hi); \
|
|---|
| 534 | } while(0)
|
|---|
| 535 |
|
|---|
| 536 | # define mul_add_c2(a,b,c0,c1,c2) do { \
|
|---|
| 537 | BN_ULONG ta = (a), tb = (b), tt; \
|
|---|
| 538 | BN_ULONG lo = ta * tb; \
|
|---|
| 539 | BN_ULONG hi = BN_UMULT_HIGH(ta,tb); \
|
|---|
| 540 | c0 += lo; tt = hi + (c0<lo); \
|
|---|
| 541 | c1 += tt; c2 += (c1<tt); \
|
|---|
| 542 | c0 += lo; hi += (c0<lo); \
|
|---|
| 543 | c1 += hi; c2 += (c1<hi); \
|
|---|
| 544 | } while(0)
|
|---|
| 545 |
|
|---|
| 546 | # define sqr_add_c(a,i,c0,c1,c2) do { \
|
|---|
| 547 | BN_ULONG ta = (a)[i]; \
|
|---|
| 548 | BN_ULONG lo = ta * ta; \
|
|---|
| 549 | BN_ULONG hi = BN_UMULT_HIGH(ta,ta); \
|
|---|
| 550 | c0 += lo; hi += (c0<lo); \
|
|---|
| 551 | c1 += hi; c2 += (c1<hi); \
|
|---|
| 552 | } while(0)
|
|---|
| 553 |
|
|---|
| 554 | # define sqr_add_c2(a,i,j,c0,c1,c2) \
|
|---|
| 555 | mul_add_c2((a)[i],(a)[j],c0,c1,c2)
|
|---|
| 556 |
|
|---|
| 557 | # else /* !BN_LLONG */
|
|---|
| 558 | /*
|
|---|
| 559 | * Keep in mind that additions to hi can not overflow, because
|
|---|
| 560 | * the high word of a multiplication result cannot be all-ones.
|
|---|
| 561 | */
|
|---|
| 562 | # define mul_add_c(a,b,c0,c1,c2) do { \
|
|---|
| 563 | BN_ULONG lo = LBITS(a), hi = HBITS(a); \
|
|---|
| 564 | BN_ULONG bl = LBITS(b), bh = HBITS(b); \
|
|---|
| 565 | mul64(lo,hi,bl,bh); \
|
|---|
| 566 | c0 = (c0+lo)&BN_MASK2; hi += (c0<lo); \
|
|---|
| 567 | c1 = (c1+hi)&BN_MASK2; c2 += (c1<hi); \
|
|---|
| 568 | } while(0)
|
|---|
| 569 |
|
|---|
| 570 | # define mul_add_c2(a,b,c0,c1,c2) do { \
|
|---|
| 571 | BN_ULONG tt; \
|
|---|
| 572 | BN_ULONG lo = LBITS(a), hi = HBITS(a); \
|
|---|
| 573 | BN_ULONG bl = LBITS(b), bh = HBITS(b); \
|
|---|
| 574 | mul64(lo,hi,bl,bh); \
|
|---|
| 575 | tt = hi; \
|
|---|
| 576 | c0 = (c0+lo)&BN_MASK2; tt += (c0<lo); \
|
|---|
| 577 | c1 = (c1+tt)&BN_MASK2; c2 += (c1<tt); \
|
|---|
| 578 | c0 = (c0+lo)&BN_MASK2; hi += (c0<lo); \
|
|---|
| 579 | c1 = (c1+hi)&BN_MASK2; c2 += (c1<hi); \
|
|---|
| 580 | } while(0)
|
|---|
| 581 |
|
|---|
| 582 | # define sqr_add_c(a,i,c0,c1,c2) do { \
|
|---|
| 583 | BN_ULONG lo, hi; \
|
|---|
| 584 | sqr64(lo,hi,(a)[i]); \
|
|---|
| 585 | c0 = (c0+lo)&BN_MASK2; hi += (c0<lo); \
|
|---|
| 586 | c1 = (c1+hi)&BN_MASK2; c2 += (c1<hi); \
|
|---|
| 587 | } while(0)
|
|---|
| 588 |
|
|---|
| 589 | # define sqr_add_c2(a,i,j,c0,c1,c2) \
|
|---|
| 590 | mul_add_c2((a)[i],(a)[j],c0,c1,c2)
|
|---|
| 591 | # endif /* !BN_LLONG */
|
|---|
| 592 |
|
|---|
| 593 | void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
|
|---|
| 594 | {
|
|---|
| 595 | BN_ULONG c1, c2, c3;
|
|---|
| 596 |
|
|---|
| 597 | c1 = 0;
|
|---|
| 598 | c2 = 0;
|
|---|
| 599 | c3 = 0;
|
|---|
| 600 | mul_add_c(a[0], b[0], c1, c2, c3);
|
|---|
| 601 | r[0] = c1;
|
|---|
| 602 | c1 = 0;
|
|---|
| 603 | mul_add_c(a[0], b[1], c2, c3, c1);
|
|---|
| 604 | mul_add_c(a[1], b[0], c2, c3, c1);
|
|---|
| 605 | r[1] = c2;
|
|---|
| 606 | c2 = 0;
|
|---|
| 607 | mul_add_c(a[2], b[0], c3, c1, c2);
|
|---|
| 608 | mul_add_c(a[1], b[1], c3, c1, c2);
|
|---|
| 609 | mul_add_c(a[0], b[2], c3, c1, c2);
|
|---|
| 610 | r[2] = c3;
|
|---|
| 611 | c3 = 0;
|
|---|
| 612 | mul_add_c(a[0], b[3], c1, c2, c3);
|
|---|
| 613 | mul_add_c(a[1], b[2], c1, c2, c3);
|
|---|
| 614 | mul_add_c(a[2], b[1], c1, c2, c3);
|
|---|
| 615 | mul_add_c(a[3], b[0], c1, c2, c3);
|
|---|
| 616 | r[3] = c1;
|
|---|
| 617 | c1 = 0;
|
|---|
| 618 | mul_add_c(a[4], b[0], c2, c3, c1);
|
|---|
| 619 | mul_add_c(a[3], b[1], c2, c3, c1);
|
|---|
| 620 | mul_add_c(a[2], b[2], c2, c3, c1);
|
|---|
| 621 | mul_add_c(a[1], b[3], c2, c3, c1);
|
|---|
| 622 | mul_add_c(a[0], b[4], c2, c3, c1);
|
|---|
| 623 | r[4] = c2;
|
|---|
| 624 | c2 = 0;
|
|---|
| 625 | mul_add_c(a[0], b[5], c3, c1, c2);
|
|---|
| 626 | mul_add_c(a[1], b[4], c3, c1, c2);
|
|---|
| 627 | mul_add_c(a[2], b[3], c3, c1, c2);
|
|---|
| 628 | mul_add_c(a[3], b[2], c3, c1, c2);
|
|---|
| 629 | mul_add_c(a[4], b[1], c3, c1, c2);
|
|---|
| 630 | mul_add_c(a[5], b[0], c3, c1, c2);
|
|---|
| 631 | r[5] = c3;
|
|---|
| 632 | c3 = 0;
|
|---|
| 633 | mul_add_c(a[6], b[0], c1, c2, c3);
|
|---|
| 634 | mul_add_c(a[5], b[1], c1, c2, c3);
|
|---|
| 635 | mul_add_c(a[4], b[2], c1, c2, c3);
|
|---|
| 636 | mul_add_c(a[3], b[3], c1, c2, c3);
|
|---|
| 637 | mul_add_c(a[2], b[4], c1, c2, c3);
|
|---|
| 638 | mul_add_c(a[1], b[5], c1, c2, c3);
|
|---|
| 639 | mul_add_c(a[0], b[6], c1, c2, c3);
|
|---|
| 640 | r[6] = c1;
|
|---|
| 641 | c1 = 0;
|
|---|
| 642 | mul_add_c(a[0], b[7], c2, c3, c1);
|
|---|
| 643 | mul_add_c(a[1], b[6], c2, c3, c1);
|
|---|
| 644 | mul_add_c(a[2], b[5], c2, c3, c1);
|
|---|
| 645 | mul_add_c(a[3], b[4], c2, c3, c1);
|
|---|
| 646 | mul_add_c(a[4], b[3], c2, c3, c1);
|
|---|
| 647 | mul_add_c(a[5], b[2], c2, c3, c1);
|
|---|
| 648 | mul_add_c(a[6], b[1], c2, c3, c1);
|
|---|
| 649 | mul_add_c(a[7], b[0], c2, c3, c1);
|
|---|
| 650 | r[7] = c2;
|
|---|
| 651 | c2 = 0;
|
|---|
| 652 | mul_add_c(a[7], b[1], c3, c1, c2);
|
|---|
| 653 | mul_add_c(a[6], b[2], c3, c1, c2);
|
|---|
| 654 | mul_add_c(a[5], b[3], c3, c1, c2);
|
|---|
| 655 | mul_add_c(a[4], b[4], c3, c1, c2);
|
|---|
| 656 | mul_add_c(a[3], b[5], c3, c1, c2);
|
|---|
| 657 | mul_add_c(a[2], b[6], c3, c1, c2);
|
|---|
| 658 | mul_add_c(a[1], b[7], c3, c1, c2);
|
|---|
| 659 | r[8] = c3;
|
|---|
| 660 | c3 = 0;
|
|---|
| 661 | mul_add_c(a[2], b[7], c1, c2, c3);
|
|---|
| 662 | mul_add_c(a[3], b[6], c1, c2, c3);
|
|---|
| 663 | mul_add_c(a[4], b[5], c1, c2, c3);
|
|---|
| 664 | mul_add_c(a[5], b[4], c1, c2, c3);
|
|---|
| 665 | mul_add_c(a[6], b[3], c1, c2, c3);
|
|---|
| 666 | mul_add_c(a[7], b[2], c1, c2, c3);
|
|---|
| 667 | r[9] = c1;
|
|---|
| 668 | c1 = 0;
|
|---|
| 669 | mul_add_c(a[7], b[3], c2, c3, c1);
|
|---|
| 670 | mul_add_c(a[6], b[4], c2, c3, c1);
|
|---|
| 671 | mul_add_c(a[5], b[5], c2, c3, c1);
|
|---|
| 672 | mul_add_c(a[4], b[6], c2, c3, c1);
|
|---|
| 673 | mul_add_c(a[3], b[7], c2, c3, c1);
|
|---|
| 674 | r[10] = c2;
|
|---|
| 675 | c2 = 0;
|
|---|
| 676 | mul_add_c(a[4], b[7], c3, c1, c2);
|
|---|
| 677 | mul_add_c(a[5], b[6], c3, c1, c2);
|
|---|
| 678 | mul_add_c(a[6], b[5], c3, c1, c2);
|
|---|
| 679 | mul_add_c(a[7], b[4], c3, c1, c2);
|
|---|
| 680 | r[11] = c3;
|
|---|
| 681 | c3 = 0;
|
|---|
| 682 | mul_add_c(a[7], b[5], c1, c2, c3);
|
|---|
| 683 | mul_add_c(a[6], b[6], c1, c2, c3);
|
|---|
| 684 | mul_add_c(a[5], b[7], c1, c2, c3);
|
|---|
| 685 | r[12] = c1;
|
|---|
| 686 | c1 = 0;
|
|---|
| 687 | mul_add_c(a[6], b[7], c2, c3, c1);
|
|---|
| 688 | mul_add_c(a[7], b[6], c2, c3, c1);
|
|---|
| 689 | r[13] = c2;
|
|---|
| 690 | c2 = 0;
|
|---|
| 691 | mul_add_c(a[7], b[7], c3, c1, c2);
|
|---|
| 692 | r[14] = c3;
|
|---|
| 693 | r[15] = c1;
|
|---|
| 694 | }
|
|---|
| 695 |
|
|---|
| 696 | void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
|
|---|
| 697 | {
|
|---|
| 698 | BN_ULONG c1, c2, c3;
|
|---|
| 699 |
|
|---|
| 700 | c1 = 0;
|
|---|
| 701 | c2 = 0;
|
|---|
| 702 | c3 = 0;
|
|---|
| 703 | mul_add_c(a[0], b[0], c1, c2, c3);
|
|---|
| 704 | r[0] = c1;
|
|---|
| 705 | c1 = 0;
|
|---|
| 706 | mul_add_c(a[0], b[1], c2, c3, c1);
|
|---|
| 707 | mul_add_c(a[1], b[0], c2, c3, c1);
|
|---|
| 708 | r[1] = c2;
|
|---|
| 709 | c2 = 0;
|
|---|
| 710 | mul_add_c(a[2], b[0], c3, c1, c2);
|
|---|
| 711 | mul_add_c(a[1], b[1], c3, c1, c2);
|
|---|
| 712 | mul_add_c(a[0], b[2], c3, c1, c2);
|
|---|
| 713 | r[2] = c3;
|
|---|
| 714 | c3 = 0;
|
|---|
| 715 | mul_add_c(a[0], b[3], c1, c2, c3);
|
|---|
| 716 | mul_add_c(a[1], b[2], c1, c2, c3);
|
|---|
| 717 | mul_add_c(a[2], b[1], c1, c2, c3);
|
|---|
| 718 | mul_add_c(a[3], b[0], c1, c2, c3);
|
|---|
| 719 | r[3] = c1;
|
|---|
| 720 | c1 = 0;
|
|---|
| 721 | mul_add_c(a[3], b[1], c2, c3, c1);
|
|---|
| 722 | mul_add_c(a[2], b[2], c2, c3, c1);
|
|---|
| 723 | mul_add_c(a[1], b[3], c2, c3, c1);
|
|---|
| 724 | r[4] = c2;
|
|---|
| 725 | c2 = 0;
|
|---|
| 726 | mul_add_c(a[2], b[3], c3, c1, c2);
|
|---|
| 727 | mul_add_c(a[3], b[2], c3, c1, c2);
|
|---|
| 728 | r[5] = c3;
|
|---|
| 729 | c3 = 0;
|
|---|
| 730 | mul_add_c(a[3], b[3], c1, c2, c3);
|
|---|
| 731 | r[6] = c1;
|
|---|
| 732 | r[7] = c2;
|
|---|
| 733 | }
|
|---|
| 734 |
|
|---|
| 735 | void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
|
|---|
| 736 | {
|
|---|
| 737 | BN_ULONG c1, c2, c3;
|
|---|
| 738 |
|
|---|
| 739 | c1 = 0;
|
|---|
| 740 | c2 = 0;
|
|---|
| 741 | c3 = 0;
|
|---|
| 742 | sqr_add_c(a, 0, c1, c2, c3);
|
|---|
| 743 | r[0] = c1;
|
|---|
| 744 | c1 = 0;
|
|---|
| 745 | sqr_add_c2(a, 1, 0, c2, c3, c1);
|
|---|
| 746 | r[1] = c2;
|
|---|
| 747 | c2 = 0;
|
|---|
| 748 | sqr_add_c(a, 1, c3, c1, c2);
|
|---|
| 749 | sqr_add_c2(a, 2, 0, c3, c1, c2);
|
|---|
| 750 | r[2] = c3;
|
|---|
| 751 | c3 = 0;
|
|---|
| 752 | sqr_add_c2(a, 3, 0, c1, c2, c3);
|
|---|
| 753 | sqr_add_c2(a, 2, 1, c1, c2, c3);
|
|---|
| 754 | r[3] = c1;
|
|---|
| 755 | c1 = 0;
|
|---|
| 756 | sqr_add_c(a, 2, c2, c3, c1);
|
|---|
| 757 | sqr_add_c2(a, 3, 1, c2, c3, c1);
|
|---|
| 758 | sqr_add_c2(a, 4, 0, c2, c3, c1);
|
|---|
| 759 | r[4] = c2;
|
|---|
| 760 | c2 = 0;
|
|---|
| 761 | sqr_add_c2(a, 5, 0, c3, c1, c2);
|
|---|
| 762 | sqr_add_c2(a, 4, 1, c3, c1, c2);
|
|---|
| 763 | sqr_add_c2(a, 3, 2, c3, c1, c2);
|
|---|
| 764 | r[5] = c3;
|
|---|
| 765 | c3 = 0;
|
|---|
| 766 | sqr_add_c(a, 3, c1, c2, c3);
|
|---|
| 767 | sqr_add_c2(a, 4, 2, c1, c2, c3);
|
|---|
| 768 | sqr_add_c2(a, 5, 1, c1, c2, c3);
|
|---|
| 769 | sqr_add_c2(a, 6, 0, c1, c2, c3);
|
|---|
| 770 | r[6] = c1;
|
|---|
| 771 | c1 = 0;
|
|---|
| 772 | sqr_add_c2(a, 7, 0, c2, c3, c1);
|
|---|
| 773 | sqr_add_c2(a, 6, 1, c2, c3, c1);
|
|---|
| 774 | sqr_add_c2(a, 5, 2, c2, c3, c1);
|
|---|
| 775 | sqr_add_c2(a, 4, 3, c2, c3, c1);
|
|---|
| 776 | r[7] = c2;
|
|---|
| 777 | c2 = 0;
|
|---|
| 778 | sqr_add_c(a, 4, c3, c1, c2);
|
|---|
| 779 | sqr_add_c2(a, 5, 3, c3, c1, c2);
|
|---|
| 780 | sqr_add_c2(a, 6, 2, c3, c1, c2);
|
|---|
| 781 | sqr_add_c2(a, 7, 1, c3, c1, c2);
|
|---|
| 782 | r[8] = c3;
|
|---|
| 783 | c3 = 0;
|
|---|
| 784 | sqr_add_c2(a, 7, 2, c1, c2, c3);
|
|---|
| 785 | sqr_add_c2(a, 6, 3, c1, c2, c3);
|
|---|
| 786 | sqr_add_c2(a, 5, 4, c1, c2, c3);
|
|---|
| 787 | r[9] = c1;
|
|---|
| 788 | c1 = 0;
|
|---|
| 789 | sqr_add_c(a, 5, c2, c3, c1);
|
|---|
| 790 | sqr_add_c2(a, 6, 4, c2, c3, c1);
|
|---|
| 791 | sqr_add_c2(a, 7, 3, c2, c3, c1);
|
|---|
| 792 | r[10] = c2;
|
|---|
| 793 | c2 = 0;
|
|---|
| 794 | sqr_add_c2(a, 7, 4, c3, c1, c2);
|
|---|
| 795 | sqr_add_c2(a, 6, 5, c3, c1, c2);
|
|---|
| 796 | r[11] = c3;
|
|---|
| 797 | c3 = 0;
|
|---|
| 798 | sqr_add_c(a, 6, c1, c2, c3);
|
|---|
| 799 | sqr_add_c2(a, 7, 5, c1, c2, c3);
|
|---|
| 800 | r[12] = c1;
|
|---|
| 801 | c1 = 0;
|
|---|
| 802 | sqr_add_c2(a, 7, 6, c2, c3, c1);
|
|---|
| 803 | r[13] = c2;
|
|---|
| 804 | c2 = 0;
|
|---|
| 805 | sqr_add_c(a, 7, c3, c1, c2);
|
|---|
| 806 | r[14] = c3;
|
|---|
| 807 | r[15] = c1;
|
|---|
| 808 | }
|
|---|
| 809 |
|
|---|
| 810 | void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
|
|---|
| 811 | {
|
|---|
| 812 | BN_ULONG c1, c2, c3;
|
|---|
| 813 |
|
|---|
| 814 | c1 = 0;
|
|---|
| 815 | c2 = 0;
|
|---|
| 816 | c3 = 0;
|
|---|
| 817 | sqr_add_c(a, 0, c1, c2, c3);
|
|---|
| 818 | r[0] = c1;
|
|---|
| 819 | c1 = 0;
|
|---|
| 820 | sqr_add_c2(a, 1, 0, c2, c3, c1);
|
|---|
| 821 | r[1] = c2;
|
|---|
| 822 | c2 = 0;
|
|---|
| 823 | sqr_add_c(a, 1, c3, c1, c2);
|
|---|
| 824 | sqr_add_c2(a, 2, 0, c3, c1, c2);
|
|---|
| 825 | r[2] = c3;
|
|---|
| 826 | c3 = 0;
|
|---|
| 827 | sqr_add_c2(a, 3, 0, c1, c2, c3);
|
|---|
| 828 | sqr_add_c2(a, 2, 1, c1, c2, c3);
|
|---|
| 829 | r[3] = c1;
|
|---|
| 830 | c1 = 0;
|
|---|
| 831 | sqr_add_c(a, 2, c2, c3, c1);
|
|---|
| 832 | sqr_add_c2(a, 3, 1, c2, c3, c1);
|
|---|
| 833 | r[4] = c2;
|
|---|
| 834 | c2 = 0;
|
|---|
| 835 | sqr_add_c2(a, 3, 2, c3, c1, c2);
|
|---|
| 836 | r[5] = c3;
|
|---|
| 837 | c3 = 0;
|
|---|
| 838 | sqr_add_c(a, 3, c1, c2, c3);
|
|---|
| 839 | r[6] = c1;
|
|---|
| 840 | r[7] = c2;
|
|---|
| 841 | }
|
|---|
| 842 |
|
|---|
| 843 | # ifdef OPENSSL_NO_ASM
|
|---|
| 844 | # ifdef OPENSSL_BN_ASM_MONT
|
|---|
| 845 | # include <alloca.h>
|
|---|
| 846 | /*
|
|---|
| 847 | * This is essentially reference implementation, which may or may not
|
|---|
| 848 | * result in performance improvement. E.g. on IA-32 this routine was
|
|---|
| 849 | * observed to give 40% faster rsa1024 private key operations and 10%
|
|---|
| 850 | * faster rsa4096 ones, while on AMD64 it improves rsa1024 sign only
|
|---|
| 851 | * by 10% and *worsens* rsa4096 sign by 15%. Once again, it's a
|
|---|
| 852 | * reference implementation, one to be used as starting point for
|
|---|
| 853 | * platform-specific assembler. Mentioned numbers apply to compiler
|
|---|
| 854 | * generated code compiled with and without -DOPENSSL_BN_ASM_MONT and
|
|---|
| 855 | * can vary not only from platform to platform, but even for compiler
|
|---|
| 856 | * versions. Assembler vs. assembler improvement coefficients can
|
|---|
| 857 | * [and are known to] differ and are to be documented elsewhere.
|
|---|
| 858 | */
|
|---|
| 859 | int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
|
|---|
| 860 | const BN_ULONG *np, const BN_ULONG *n0p, int num)
|
|---|
| 861 | {
|
|---|
| 862 | BN_ULONG c0, c1, ml, *tp, n0;
|
|---|
| 863 | # ifdef mul64
|
|---|
| 864 | BN_ULONG mh;
|
|---|
| 865 | # endif
|
|---|
| 866 | volatile BN_ULONG *vp;
|
|---|
| 867 | int i = 0, j;
|
|---|
| 868 |
|
|---|
| 869 | # if 0 /* template for platform-specific
|
|---|
| 870 | * implementation */
|
|---|
| 871 | if (ap == bp)
|
|---|
| 872 | return bn_sqr_mont(rp, ap, np, n0p, num);
|
|---|
| 873 | # endif
|
|---|
| 874 | vp = tp = alloca((num + 2) * sizeof(BN_ULONG));
|
|---|
| 875 |
|
|---|
| 876 | n0 = *n0p;
|
|---|
| 877 |
|
|---|
| 878 | c0 = 0;
|
|---|
| 879 | ml = bp[0];
|
|---|
| 880 | # ifdef mul64
|
|---|
| 881 | mh = HBITS(ml);
|
|---|
| 882 | ml = LBITS(ml);
|
|---|
| 883 | for (j = 0; j < num; ++j)
|
|---|
| 884 | mul(tp[j], ap[j], ml, mh, c0);
|
|---|
| 885 | # else
|
|---|
| 886 | for (j = 0; j < num; ++j)
|
|---|
| 887 | mul(tp[j], ap[j], ml, c0);
|
|---|
| 888 | # endif
|
|---|
| 889 |
|
|---|
| 890 | tp[num] = c0;
|
|---|
| 891 | tp[num + 1] = 0;
|
|---|
| 892 | goto enter;
|
|---|
| 893 |
|
|---|
| 894 | for (i = 0; i < num; i++) {
|
|---|
| 895 | c0 = 0;
|
|---|
| 896 | ml = bp[i];
|
|---|
| 897 | # ifdef mul64
|
|---|
| 898 | mh = HBITS(ml);
|
|---|
| 899 | ml = LBITS(ml);
|
|---|
| 900 | for (j = 0; j < num; ++j)
|
|---|
| 901 | mul_add(tp[j], ap[j], ml, mh, c0);
|
|---|
| 902 | # else
|
|---|
| 903 | for (j = 0; j < num; ++j)
|
|---|
| 904 | mul_add(tp[j], ap[j], ml, c0);
|
|---|
| 905 | # endif
|
|---|
| 906 | c1 = (tp[num] + c0) & BN_MASK2;
|
|---|
| 907 | tp[num] = c1;
|
|---|
| 908 | tp[num + 1] = (c1 < c0 ? 1 : 0);
|
|---|
| 909 | enter:
|
|---|
| 910 | c1 = tp[0];
|
|---|
| 911 | ml = (c1 * n0) & BN_MASK2;
|
|---|
| 912 | c0 = 0;
|
|---|
| 913 | # ifdef mul64
|
|---|
| 914 | mh = HBITS(ml);
|
|---|
| 915 | ml = LBITS(ml);
|
|---|
| 916 | mul_add(c1, np[0], ml, mh, c0);
|
|---|
| 917 | # else
|
|---|
| 918 | mul_add(c1, ml, np[0], c0);
|
|---|
| 919 | # endif
|
|---|
| 920 | for (j = 1; j < num; j++) {
|
|---|
| 921 | c1 = tp[j];
|
|---|
| 922 | # ifdef mul64
|
|---|
| 923 | mul_add(c1, np[j], ml, mh, c0);
|
|---|
| 924 | # else
|
|---|
| 925 | mul_add(c1, ml, np[j], c0);
|
|---|
| 926 | # endif
|
|---|
| 927 | tp[j - 1] = c1 & BN_MASK2;
|
|---|
| 928 | }
|
|---|
| 929 | c1 = (tp[num] + c0) & BN_MASK2;
|
|---|
| 930 | tp[num - 1] = c1;
|
|---|
| 931 | tp[num] = tp[num + 1] + (c1 < c0 ? 1 : 0);
|
|---|
| 932 | }
|
|---|
| 933 |
|
|---|
| 934 | if (tp[num] != 0 || tp[num - 1] >= np[num - 1]) {
|
|---|
| 935 | c0 = bn_sub_words(rp, tp, np, num);
|
|---|
| 936 | if (tp[num] != 0 || c0 == 0) {
|
|---|
| 937 | for (i = 0; i < num + 2; i++)
|
|---|
| 938 | vp[i] = 0;
|
|---|
| 939 | return 1;
|
|---|
| 940 | }
|
|---|
| 941 | }
|
|---|
| 942 | for (i = 0; i < num; i++)
|
|---|
| 943 | rp[i] = tp[i], vp[i] = 0;
|
|---|
| 944 | vp[num] = 0;
|
|---|
| 945 | vp[num + 1] = 0;
|
|---|
| 946 | return 1;
|
|---|
| 947 | }
|
|---|
| 948 | # else
|
|---|
| 949 | /*
|
|---|
| 950 | * Return value of 0 indicates that multiplication/convolution was not
|
|---|
| 951 | * performed to signal the caller to fall down to alternative/original
|
|---|
| 952 | * code-path.
|
|---|
| 953 | */
|
|---|
| 954 | int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
|
|---|
| 955 | const BN_ULONG *np, const BN_ULONG *n0, int num)
|
|---|
| 956 | {
|
|---|
| 957 | return 0;
|
|---|
| 958 | }
|
|---|
| 959 | # endif /* OPENSSL_BN_ASM_MONT */
|
|---|
| 960 | # endif
|
|---|
| 961 |
|
|---|
| 962 | #else /* !BN_MUL_COMBA */
|
|---|
| 963 |
|
|---|
| 964 | /* hmm... is it faster just to do a multiply? */
|
|---|
| 965 | #ifndef ___openssl_mangling_h___ /* bird */
|
|---|
| 966 | # undef bn_sqr_comba4
|
|---|
| 967 | # undef bn_sqr_comba8
|
|---|
| 968 | #endif
|
|---|
| 969 | void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
|
|---|
| 970 | {
|
|---|
| 971 | BN_ULONG t[8];
|
|---|
| 972 | bn_sqr_normal(r, a, 4, t);
|
|---|
| 973 | }
|
|---|
| 974 |
|
|---|
| 975 | void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
|
|---|
| 976 | {
|
|---|
| 977 | BN_ULONG t[16];
|
|---|
| 978 | bn_sqr_normal(r, a, 8, t);
|
|---|
| 979 | }
|
|---|
| 980 |
|
|---|
| 981 | void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
|
|---|
| 982 | {
|
|---|
| 983 | r[4] = bn_mul_words(&(r[0]), a, 4, b[0]);
|
|---|
| 984 | r[5] = bn_mul_add_words(&(r[1]), a, 4, b[1]);
|
|---|
| 985 | r[6] = bn_mul_add_words(&(r[2]), a, 4, b[2]);
|
|---|
| 986 | r[7] = bn_mul_add_words(&(r[3]), a, 4, b[3]);
|
|---|
| 987 | }
|
|---|
| 988 |
|
|---|
| 989 | void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
|
|---|
| 990 | {
|
|---|
| 991 | r[8] = bn_mul_words(&(r[0]), a, 8, b[0]);
|
|---|
| 992 | r[9] = bn_mul_add_words(&(r[1]), a, 8, b[1]);
|
|---|
| 993 | r[10] = bn_mul_add_words(&(r[2]), a, 8, b[2]);
|
|---|
| 994 | r[11] = bn_mul_add_words(&(r[3]), a, 8, b[3]);
|
|---|
| 995 | r[12] = bn_mul_add_words(&(r[4]), a, 8, b[4]);
|
|---|
| 996 | r[13] = bn_mul_add_words(&(r[5]), a, 8, b[5]);
|
|---|
| 997 | r[14] = bn_mul_add_words(&(r[6]), a, 8, b[6]);
|
|---|
| 998 | r[15] = bn_mul_add_words(&(r[7]), a, 8, b[7]);
|
|---|
| 999 | }
|
|---|
| 1000 |
|
|---|
| 1001 | # ifdef OPENSSL_NO_ASM
|
|---|
| 1002 | # ifdef OPENSSL_BN_ASM_MONT
|
|---|
| 1003 | # include <alloca.h>
|
|---|
| 1004 | int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
|
|---|
| 1005 | const BN_ULONG *np, const BN_ULONG *n0p, int num)
|
|---|
| 1006 | {
|
|---|
| 1007 | BN_ULONG c0, c1, *tp, n0 = *n0p;
|
|---|
| 1008 | volatile BN_ULONG *vp;
|
|---|
| 1009 | int i = 0, j;
|
|---|
| 1010 |
|
|---|
| 1011 | vp = tp = alloca((num + 2) * sizeof(BN_ULONG));
|
|---|
| 1012 |
|
|---|
| 1013 | for (i = 0; i <= num; i++)
|
|---|
| 1014 | tp[i] = 0;
|
|---|
| 1015 |
|
|---|
| 1016 | for (i = 0; i < num; i++) {
|
|---|
| 1017 | c0 = bn_mul_add_words(tp, ap, num, bp[i]);
|
|---|
| 1018 | c1 = (tp[num] + c0) & BN_MASK2;
|
|---|
| 1019 | tp[num] = c1;
|
|---|
| 1020 | tp[num + 1] = (c1 < c0 ? 1 : 0);
|
|---|
| 1021 |
|
|---|
| 1022 | c0 = bn_mul_add_words(tp, np, num, tp[0] * n0);
|
|---|
| 1023 | c1 = (tp[num] + c0) & BN_MASK2;
|
|---|
| 1024 | tp[num] = c1;
|
|---|
| 1025 | tp[num + 1] += (c1 < c0 ? 1 : 0);
|
|---|
| 1026 | for (j = 0; j <= num; j++)
|
|---|
| 1027 | tp[j] = tp[j + 1];
|
|---|
| 1028 | }
|
|---|
| 1029 |
|
|---|
| 1030 | if (tp[num] != 0 || tp[num - 1] >= np[num - 1]) {
|
|---|
| 1031 | c0 = bn_sub_words(rp, tp, np, num);
|
|---|
| 1032 | if (tp[num] != 0 || c0 == 0) {
|
|---|
| 1033 | for (i = 0; i < num + 2; i++)
|
|---|
| 1034 | vp[i] = 0;
|
|---|
| 1035 | return 1;
|
|---|
| 1036 | }
|
|---|
| 1037 | }
|
|---|
| 1038 | for (i = 0; i < num; i++)
|
|---|
| 1039 | rp[i] = tp[i], vp[i] = 0;
|
|---|
| 1040 | vp[num] = 0;
|
|---|
| 1041 | vp[num + 1] = 0;
|
|---|
| 1042 | return 1;
|
|---|
| 1043 | }
|
|---|
| 1044 | # else
|
|---|
| 1045 | int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
|
|---|
| 1046 | const BN_ULONG *np, const BN_ULONG *n0, int num)
|
|---|
| 1047 | {
|
|---|
| 1048 | return 0;
|
|---|
| 1049 | }
|
|---|
| 1050 | # endif /* OPENSSL_BN_ASM_MONT */
|
|---|
| 1051 | # endif
|
|---|
| 1052 |
|
|---|
| 1053 | #endif /* !BN_MUL_COMBA */
|
|---|
