1 /** 2 * \file bn_mul.h 3 * 4 * \brief Multi-precision integer library 5 */ 6 /* 7 * Copyright The Mbed TLS Contributors 8 * SPDX-License-Identifier: Apache-2.0 OR GPL-2.0-or-later 9 * 10 * This file is provided under the Apache License 2.0, or the 11 * GNU General Public License v2.0 or later. 12 * 13 * ********** 14 * Apache License 2.0: 15 * 16 * Licensed under the Apache License, Version 2.0 (the "License"); you may 17 * not use this file except in compliance with the License. 18 * You may obtain a copy of the License at 19 * 20 * http://www.apache.org/licenses/LICENSE-2.0 21 * 22 * Unless required by applicable law or agreed to in writing, software 23 * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 24 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 25 * See the License for the specific language governing permissions and 26 * limitations under the License. 27 * 28 * ********** 29 * 30 * ********** 31 * GNU General Public License v2.0 or later: 32 * 33 * This program is free software; you can redistribute it and/or modify 34 * it under the terms of the GNU General Public License as published by 35 * the Free Software Foundation; either version 2 of the License, or 36 * (at your option) any later version. 37 * 38 * This program is distributed in the hope that it will be useful, 39 * but WITHOUT ANY WARRANTY; without even the implied warranty of 40 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 41 * GNU General Public License for more details. 42 * 43 * You should have received a copy of the GNU General Public License along 44 * with this program; if not, write to the Free Software Foundation, Inc., 45 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. 46 * 47 * ********** 48 */ 49 /* 50 * Multiply source vector [s] with b, add result 51 * to destination vector [d] and set carry c. 52 * 53 * Currently supports: 54 * 55 * . IA-32 (386+) . AMD64 / EM64T 56 * . IA-32 (SSE2) . Motorola 68000 57 * . PowerPC, 32-bit . MicroBlaze 58 * . PowerPC, 64-bit . TriCore 59 * . SPARC v8 . ARM v3+ 60 * . Alpha . MIPS32 61 * . C, longlong . C, generic 62 */ 63 #ifndef MBEDTLS_BN_MUL_H 64 #define MBEDTLS_BN_MUL_H 65 66 #if !defined(MBEDTLS_CONFIG_FILE) 67 #include "config.h" 68 #else 69 #include MBEDTLS_CONFIG_FILE 70 #endif 71 72 #include "bignum.h" 73 74 #if defined(MBEDTLS_HAVE_ASM) 75 76 #ifndef asm 77 #define asm __asm 78 #endif 79 80 /* armcc5 --gnu defines __GNUC__ but doesn't support GNU's extended asm */ 81 #if defined(__GNUC__) && \ 82 ( !defined(__ARMCC_VERSION) || __ARMCC_VERSION >= 6000000 ) 83 84 /* 85 * Disable use of the i386 assembly code below if option -O0, to disable all 86 * compiler optimisations, is passed, detected with __OPTIMIZE__ 87 * This is done as the number of registers used in the assembly code doesn't 88 * work with the -O0 option. 89 */ 90 #if defined(__i386__) && defined(__OPTIMIZE__) 91 92 #define MULADDC_INIT \ 93 asm( \ 94 "movl %%ebx, %0 \n\t" \ 95 "movl %5, %%esi \n\t" \ 96 "movl %6, %%edi \n\t" \ 97 "movl %7, %%ecx \n\t" \ 98 "movl %8, %%ebx \n\t" 99 100 #define MULADDC_CORE \ 101 "lodsl \n\t" \ 102 "mull %%ebx \n\t" \ 103 "addl %%ecx, %%eax \n\t" \ 104 "adcl $0, %%edx \n\t" \ 105 "addl (%%edi), %%eax \n\t" \ 106 "adcl $0, %%edx \n\t" \ 107 "movl %%edx, %%ecx \n\t" \ 108 "stosl \n\t" 109 110 #if defined(MBEDTLS_HAVE_SSE2) 111 112 #define MULADDC_HUIT \ 113 "movd %%ecx, %%mm1 \n\t" \ 114 "movd %%ebx, %%mm0 \n\t" \ 115 "movd (%%edi), %%mm3 \n\t" \ 116 "paddq %%mm3, %%mm1 \n\t" \ 117 "movd (%%esi), %%mm2 \n\t" \ 118 "pmuludq %%mm0, %%mm2 \n\t" \ 119 "movd 4(%%esi), %%mm4 \n\t" \ 120 "pmuludq %%mm0, %%mm4 \n\t" \ 121 "movd 8(%%esi), %%mm6 \n\t" \ 122 "pmuludq %%mm0, %%mm6 \n\t" \ 123 "movd 12(%%esi), %%mm7 \n\t" \ 124 "pmuludq %%mm0, %%mm7 \n\t" \ 125 "paddq %%mm2, %%mm1 \n\t" \ 126 "movd 4(%%edi), %%mm3 \n\t" \ 127 "paddq %%mm4, %%mm3 \n\t" \ 128 "movd 8(%%edi), %%mm5 \n\t" \ 129 "paddq %%mm6, %%mm5 \n\t" \ 130 "movd 12(%%edi), %%mm4 \n\t" \ 131 "paddq %%mm4, %%mm7 \n\t" \ 132 "movd %%mm1, (%%edi) \n\t" \ 133 "movd 16(%%esi), %%mm2 \n\t" \ 134 "pmuludq %%mm0, %%mm2 \n\t" \ 135 "psrlq $32, %%mm1 \n\t" \ 136 "movd 20(%%esi), %%mm4 \n\t" \ 137 "pmuludq %%mm0, %%mm4 \n\t" \ 138 "paddq %%mm3, %%mm1 \n\t" \ 139 "movd 24(%%esi), %%mm6 \n\t" \ 140 "pmuludq %%mm0, %%mm6 \n\t" \ 141 "movd %%mm1, 4(%%edi) \n\t" \ 142 "psrlq $32, %%mm1 \n\t" \ 143 "movd 28(%%esi), %%mm3 \n\t" \ 144 "pmuludq %%mm0, %%mm3 \n\t" \ 145 "paddq %%mm5, %%mm1 \n\t" \ 146 "movd 16(%%edi), %%mm5 \n\t" \ 147 "paddq %%mm5, %%mm2 \n\t" \ 148 "movd %%mm1, 8(%%edi) \n\t" \ 149 "psrlq $32, %%mm1 \n\t" \ 150 "paddq %%mm7, %%mm1 \n\t" \ 151 "movd 20(%%edi), %%mm5 \n\t" \ 152 "paddq %%mm5, %%mm4 \n\t" \ 153 "movd %%mm1, 12(%%edi) \n\t" \ 154 "psrlq $32, %%mm1 \n\t" \ 155 "paddq %%mm2, %%mm1 \n\t" \ 156 "movd 24(%%edi), %%mm5 \n\t" \ 157 "paddq %%mm5, %%mm6 \n\t" \ 158 "movd %%mm1, 16(%%edi) \n\t" \ 159 "psrlq $32, %%mm1 \n\t" \ 160 "paddq %%mm4, %%mm1 \n\t" \ 161 "movd 28(%%edi), %%mm5 \n\t" \ 162 "paddq %%mm5, %%mm3 \n\t" \ 163 "movd %%mm1, 20(%%edi) \n\t" \ 164 "psrlq $32, %%mm1 \n\t" \ 165 "paddq %%mm6, %%mm1 \n\t" \ 166 "movd %%mm1, 24(%%edi) \n\t" \ 167 "psrlq $32, %%mm1 \n\t" \ 168 "paddq %%mm3, %%mm1 \n\t" \ 169 "movd %%mm1, 28(%%edi) \n\t" \ 170 "addl $32, %%edi \n\t" \ 171 "addl $32, %%esi \n\t" \ 172 "psrlq $32, %%mm1 \n\t" \ 173 "movd %%mm1, %%ecx \n\t" 174 175 #define MULADDC_STOP \ 176 "emms \n\t" \ 177 "movl %4, %%ebx \n\t" \ 178 "movl %%ecx, %1 \n\t" \ 179 "movl %%edi, %2 \n\t" \ 180 "movl %%esi, %3 \n\t" \ 181 : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \ 182 : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \ 183 : "eax", "ebx", "ecx", "edx", "esi", "edi" \ 184 ); 185 186 #else 187 188 #define MULADDC_STOP \ 189 "movl %4, %%ebx \n\t" \ 190 "movl %%ecx, %1 \n\t" \ 191 "movl %%edi, %2 \n\t" \ 192 "movl %%esi, %3 \n\t" \ 193 : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \ 194 : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \ 195 : "eax", "ebx", "ecx", "edx", "esi", "edi" \ 196 ); 197 #endif /* SSE2 */ 198 #endif /* i386 */ 199 200 #if defined(__amd64__) || defined (__x86_64__) 201 202 #define MULADDC_INIT \ 203 asm( \ 204 "xorq %%r8, %%r8\n" 205 206 #define MULADDC_CORE \ 207 "movq (%%rsi), %%rax\n" \ 208 "mulq %%rbx\n" \ 209 "addq $8, %%rsi\n" \ 210 "addq %%rcx, %%rax\n" \ 211 "movq %%r8, %%rcx\n" \ 212 "adcq $0, %%rdx\n" \ 213 "nop \n" \ 214 "addq %%rax, (%%rdi)\n" \ 215 "adcq %%rdx, %%rcx\n" \ 216 "addq $8, %%rdi\n" 217 218 #define MULADDC_STOP \ 219 : "+c" (c), "+D" (d), "+S" (s) \ 220 : "b" (b) \ 221 : "rax", "rdx", "r8" \ 222 ); 223 224 #endif /* AMD64 */ 225 226 #if defined(__mc68020__) || defined(__mcpu32__) 227 228 #define MULADDC_INIT \ 229 asm( \ 230 "movl %3, %%a2 \n\t" \ 231 "movl %4, %%a3 \n\t" \ 232 "movl %5, %%d3 \n\t" \ 233 "movl %6, %%d2 \n\t" \ 234 "moveq #0, %%d0 \n\t" 235 236 #define MULADDC_CORE \ 237 "movel %%a2@+, %%d1 \n\t" \ 238 "mulul %%d2, %%d4:%%d1 \n\t" \ 239 "addl %%d3, %%d1 \n\t" \ 240 "addxl %%d0, %%d4 \n\t" \ 241 "moveq #0, %%d3 \n\t" \ 242 "addl %%d1, %%a3@+ \n\t" \ 243 "addxl %%d4, %%d3 \n\t" 244 245 #define MULADDC_STOP \ 246 "movl %%d3, %0 \n\t" \ 247 "movl %%a3, %1 \n\t" \ 248 "movl %%a2, %2 \n\t" \ 249 : "=m" (c), "=m" (d), "=m" (s) \ 250 : "m" (s), "m" (d), "m" (c), "m" (b) \ 251 : "d0", "d1", "d2", "d3", "d4", "a2", "a3" \ 252 ); 253 254 #define MULADDC_HUIT \ 255 "movel %%a2@+, %%d1 \n\t" \ 256 "mulul %%d2, %%d4:%%d1 \n\t" \ 257 "addxl %%d3, %%d1 \n\t" \ 258 "addxl %%d0, %%d4 \n\t" \ 259 "addl %%d1, %%a3@+ \n\t" \ 260 "movel %%a2@+, %%d1 \n\t" \ 261 "mulul %%d2, %%d3:%%d1 \n\t" \ 262 "addxl %%d4, %%d1 \n\t" \ 263 "addxl %%d0, %%d3 \n\t" \ 264 "addl %%d1, %%a3@+ \n\t" \ 265 "movel %%a2@+, %%d1 \n\t" \ 266 "mulul %%d2, %%d4:%%d1 \n\t" \ 267 "addxl %%d3, %%d1 \n\t" \ 268 "addxl %%d0, %%d4 \n\t" \ 269 "addl %%d1, %%a3@+ \n\t" \ 270 "movel %%a2@+, %%d1 \n\t" \ 271 "mulul %%d2, %%d3:%%d1 \n\t" \ 272 "addxl %%d4, %%d1 \n\t" \ 273 "addxl %%d0, %%d3 \n\t" \ 274 "addl %%d1, %%a3@+ \n\t" \ 275 "movel %%a2@+, %%d1 \n\t" \ 276 "mulul %%d2, %%d4:%%d1 \n\t" \ 277 "addxl %%d3, %%d1 \n\t" \ 278 "addxl %%d0, %%d4 \n\t" \ 279 "addl %%d1, %%a3@+ \n\t" \ 280 "movel %%a2@+, %%d1 \n\t" \ 281 "mulul %%d2, %%d3:%%d1 \n\t" \ 282 "addxl %%d4, %%d1 \n\t" \ 283 "addxl %%d0, %%d3 \n\t" \ 284 "addl %%d1, %%a3@+ \n\t" \ 285 "movel %%a2@+, %%d1 \n\t" \ 286 "mulul %%d2, %%d4:%%d1 \n\t" \ 287 "addxl %%d3, %%d1 \n\t" \ 288 "addxl %%d0, %%d4 \n\t" \ 289 "addl %%d1, %%a3@+ \n\t" \ 290 "movel %%a2@+, %%d1 \n\t" \ 291 "mulul %%d2, %%d3:%%d1 \n\t" \ 292 "addxl %%d4, %%d1 \n\t" \ 293 "addxl %%d0, %%d3 \n\t" \ 294 "addl %%d1, %%a3@+ \n\t" \ 295 "addxl %%d0, %%d3 \n\t" 296 297 #endif /* MC68000 */ 298 299 #if defined(__powerpc64__) || defined(__ppc64__) 300 301 #if defined(__MACH__) && defined(__APPLE__) 302 303 #define MULADDC_INIT \ 304 asm( \ 305 "ld r3, %3 \n\t" \ 306 "ld r4, %4 \n\t" \ 307 "ld r5, %5 \n\t" \ 308 "ld r6, %6 \n\t" \ 309 "addi r3, r3, -8 \n\t" \ 310 "addi r4, r4, -8 \n\t" \ 311 "addic r5, r5, 0 \n\t" 312 313 #define MULADDC_CORE \ 314 "ldu r7, 8(r3) \n\t" \ 315 "mulld r8, r7, r6 \n\t" \ 316 "mulhdu r9, r7, r6 \n\t" \ 317 "adde r8, r8, r5 \n\t" \ 318 "ld r7, 8(r4) \n\t" \ 319 "addze r5, r9 \n\t" \ 320 "addc r8, r8, r7 \n\t" \ 321 "stdu r8, 8(r4) \n\t" 322 323 #define MULADDC_STOP \ 324 "addze r5, r5 \n\t" \ 325 "addi r4, r4, 8 \n\t" \ 326 "addi r3, r3, 8 \n\t" \ 327 "std r5, %0 \n\t" \ 328 "std r4, %1 \n\t" \ 329 "std r3, %2 \n\t" \ 330 : "=m" (c), "=m" (d), "=m" (s) \ 331 : "m" (s), "m" (d), "m" (c), "m" (b) \ 332 : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \ 333 ); 334 335 336 #else /* __MACH__ && __APPLE__ */ 337 338 #define MULADDC_INIT \ 339 asm( \ 340 "ld %%r3, %3 \n\t" \ 341 "ld %%r4, %4 \n\t" \ 342 "ld %%r5, %5 \n\t" \ 343 "ld %%r6, %6 \n\t" \ 344 "addi %%r3, %%r3, -8 \n\t" \ 345 "addi %%r4, %%r4, -8 \n\t" \ 346 "addic %%r5, %%r5, 0 \n\t" 347 348 #define MULADDC_CORE \ 349 "ldu %%r7, 8(%%r3) \n\t" \ 350 "mulld %%r8, %%r7, %%r6 \n\t" \ 351 "mulhdu %%r9, %%r7, %%r6 \n\t" \ 352 "adde %%r8, %%r8, %%r5 \n\t" \ 353 "ld %%r7, 8(%%r4) \n\t" \ 354 "addze %%r5, %%r9 \n\t" \ 355 "addc %%r8, %%r8, %%r7 \n\t" \ 356 "stdu %%r8, 8(%%r4) \n\t" 357 358 #define MULADDC_STOP \ 359 "addze %%r5, %%r5 \n\t" \ 360 "addi %%r4, %%r4, 8 \n\t" \ 361 "addi %%r3, %%r3, 8 \n\t" \ 362 "std %%r5, %0 \n\t" \ 363 "std %%r4, %1 \n\t" \ 364 "std %%r3, %2 \n\t" \ 365 : "=m" (c), "=m" (d), "=m" (s) \ 366 : "m" (s), "m" (d), "m" (c), "m" (b) \ 367 : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \ 368 ); 369 370 #endif /* __MACH__ && __APPLE__ */ 371 372 #elif defined(__powerpc__) || defined(__ppc__) /* end PPC64/begin PPC32 */ 373 374 #if defined(__MACH__) && defined(__APPLE__) 375 376 #define MULADDC_INIT \ 377 asm( \ 378 "lwz r3, %3 \n\t" \ 379 "lwz r4, %4 \n\t" \ 380 "lwz r5, %5 \n\t" \ 381 "lwz r6, %6 \n\t" \ 382 "addi r3, r3, -4 \n\t" \ 383 "addi r4, r4, -4 \n\t" \ 384 "addic r5, r5, 0 \n\t" 385 386 #define MULADDC_CORE \ 387 "lwzu r7, 4(r3) \n\t" \ 388 "mullw r8, r7, r6 \n\t" \ 389 "mulhwu r9, r7, r6 \n\t" \ 390 "adde r8, r8, r5 \n\t" \ 391 "lwz r7, 4(r4) \n\t" \ 392 "addze r5, r9 \n\t" \ 393 "addc r8, r8, r7 \n\t" \ 394 "stwu r8, 4(r4) \n\t" 395 396 #define MULADDC_STOP \ 397 "addze r5, r5 \n\t" \ 398 "addi r4, r4, 4 \n\t" \ 399 "addi r3, r3, 4 \n\t" \ 400 "stw r5, %0 \n\t" \ 401 "stw r4, %1 \n\t" \ 402 "stw r3, %2 \n\t" \ 403 : "=m" (c), "=m" (d), "=m" (s) \ 404 : "m" (s), "m" (d), "m" (c), "m" (b) \ 405 : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \ 406 ); 407 408 #else /* __MACH__ && __APPLE__ */ 409 410 #define MULADDC_INIT \ 411 asm( \ 412 "lwz %%r3, %3 \n\t" \ 413 "lwz %%r4, %4 \n\t" \ 414 "lwz %%r5, %5 \n\t" \ 415 "lwz %%r6, %6 \n\t" \ 416 "addi %%r3, %%r3, -4 \n\t" \ 417 "addi %%r4, %%r4, -4 \n\t" \ 418 "addic %%r5, %%r5, 0 \n\t" 419 420 #define MULADDC_CORE \ 421 "lwzu %%r7, 4(%%r3) \n\t" \ 422 "mullw %%r8, %%r7, %%r6 \n\t" \ 423 "mulhwu %%r9, %%r7, %%r6 \n\t" \ 424 "adde %%r8, %%r8, %%r5 \n\t" \ 425 "lwz %%r7, 4(%%r4) \n\t" \ 426 "addze %%r5, %%r9 \n\t" \ 427 "addc %%r8, %%r8, %%r7 \n\t" \ 428 "stwu %%r8, 4(%%r4) \n\t" 429 430 #define MULADDC_STOP \ 431 "addze %%r5, %%r5 \n\t" \ 432 "addi %%r4, %%r4, 4 \n\t" \ 433 "addi %%r3, %%r3, 4 \n\t" \ 434 "stw %%r5, %0 \n\t" \ 435 "stw %%r4, %1 \n\t" \ 436 "stw %%r3, %2 \n\t" \ 437 : "=m" (c), "=m" (d), "=m" (s) \ 438 : "m" (s), "m" (d), "m" (c), "m" (b) \ 439 : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \ 440 ); 441 442 #endif /* __MACH__ && __APPLE__ */ 443 444 #endif /* PPC32 */ 445 446 /* 447 * The Sparc(64) assembly is reported to be broken. 448 * Disable it for now, until we're able to fix it. 449 */ 450 #if 0 && defined(__sparc__) 451 #if defined(__sparc64__) 452 453 #define MULADDC_INIT \ 454 asm( \ 455 "ldx %3, %%o0 \n\t" \ 456 "ldx %4, %%o1 \n\t" \ 457 "ld %5, %%o2 \n\t" \ 458 "ld %6, %%o3 \n\t" 459 460 #define MULADDC_CORE \ 461 "ld [%%o0], %%o4 \n\t" \ 462 "inc 4, %%o0 \n\t" \ 463 "ld [%%o1], %%o5 \n\t" \ 464 "umul %%o3, %%o4, %%o4 \n\t" \ 465 "addcc %%o4, %%o2, %%o4 \n\t" \ 466 "rd %%y, %%g1 \n\t" \ 467 "addx %%g1, 0, %%g1 \n\t" \ 468 "addcc %%o4, %%o5, %%o4 \n\t" \ 469 "st %%o4, [%%o1] \n\t" \ 470 "addx %%g1, 0, %%o2 \n\t" \ 471 "inc 4, %%o1 \n\t" 472 473 #define MULADDC_STOP \ 474 "st %%o2, %0 \n\t" \ 475 "stx %%o1, %1 \n\t" \ 476 "stx %%o0, %2 \n\t" \ 477 : "=m" (c), "=m" (d), "=m" (s) \ 478 : "m" (s), "m" (d), "m" (c), "m" (b) \ 479 : "g1", "o0", "o1", "o2", "o3", "o4", \ 480 "o5" \ 481 ); 482 483 #else /* __sparc64__ */ 484 485 #define MULADDC_INIT \ 486 asm( \ 487 "ld %3, %%o0 \n\t" \ 488 "ld %4, %%o1 \n\t" \ 489 "ld %5, %%o2 \n\t" \ 490 "ld %6, %%o3 \n\t" 491 492 #define MULADDC_CORE \ 493 "ld [%%o0], %%o4 \n\t" \ 494 "inc 4, %%o0 \n\t" \ 495 "ld [%%o1], %%o5 \n\t" \ 496 "umul %%o3, %%o4, %%o4 \n\t" \ 497 "addcc %%o4, %%o2, %%o4 \n\t" \ 498 "rd %%y, %%g1 \n\t" \ 499 "addx %%g1, 0, %%g1 \n\t" \ 500 "addcc %%o4, %%o5, %%o4 \n\t" \ 501 "st %%o4, [%%o1] \n\t" \ 502 "addx %%g1, 0, %%o2 \n\t" \ 503 "inc 4, %%o1 \n\t" 504 505 #define MULADDC_STOP \ 506 "st %%o2, %0 \n\t" \ 507 "st %%o1, %1 \n\t" \ 508 "st %%o0, %2 \n\t" \ 509 : "=m" (c), "=m" (d), "=m" (s) \ 510 : "m" (s), "m" (d), "m" (c), "m" (b) \ 511 : "g1", "o0", "o1", "o2", "o3", "o4", \ 512 "o5" \ 513 ); 514 515 #endif /* __sparc64__ */ 516 #endif /* __sparc__ */ 517 518 #if defined(__microblaze__) || defined(microblaze) 519 520 #define MULADDC_INIT \ 521 asm( \ 522 "lwi r3, %3 \n\t" \ 523 "lwi r4, %4 \n\t" \ 524 "lwi r5, %5 \n\t" \ 525 "lwi r6, %6 \n\t" \ 526 "andi r7, r6, 0xffff \n\t" \ 527 "bsrli r6, r6, 16 \n\t" 528 529 #define MULADDC_CORE \ 530 "lhui r8, r3, 0 \n\t" \ 531 "addi r3, r3, 2 \n\t" \ 532 "lhui r9, r3, 0 \n\t" \ 533 "addi r3, r3, 2 \n\t" \ 534 "mul r10, r9, r6 \n\t" \ 535 "mul r11, r8, r7 \n\t" \ 536 "mul r12, r9, r7 \n\t" \ 537 "mul r13, r8, r6 \n\t" \ 538 "bsrli r8, r10, 16 \n\t" \ 539 "bsrli r9, r11, 16 \n\t" \ 540 "add r13, r13, r8 \n\t" \ 541 "add r13, r13, r9 \n\t" \ 542 "bslli r10, r10, 16 \n\t" \ 543 "bslli r11, r11, 16 \n\t" \ 544 "add r12, r12, r10 \n\t" \ 545 "addc r13, r13, r0 \n\t" \ 546 "add r12, r12, r11 \n\t" \ 547 "addc r13, r13, r0 \n\t" \ 548 "lwi r10, r4, 0 \n\t" \ 549 "add r12, r12, r10 \n\t" \ 550 "addc r13, r13, r0 \n\t" \ 551 "add r12, r12, r5 \n\t" \ 552 "addc r5, r13, r0 \n\t" \ 553 "swi r12, r4, 0 \n\t" \ 554 "addi r4, r4, 4 \n\t" 555 556 #define MULADDC_STOP \ 557 "swi r5, %0 \n\t" \ 558 "swi r4, %1 \n\t" \ 559 "swi r3, %2 \n\t" \ 560 : "=m" (c), "=m" (d), "=m" (s) \ 561 : "m" (s), "m" (d), "m" (c), "m" (b) \ 562 : "r3", "r4", "r5", "r6", "r7", "r8", \ 563 "r9", "r10", "r11", "r12", "r13" \ 564 ); 565 566 #endif /* MicroBlaze */ 567 568 #if defined(__tricore__) 569 570 #define MULADDC_INIT \ 571 asm( \ 572 "ld.a %%a2, %3 \n\t" \ 573 "ld.a %%a3, %4 \n\t" \ 574 "ld.w %%d4, %5 \n\t" \ 575 "ld.w %%d1, %6 \n\t" \ 576 "xor %%d5, %%d5 \n\t" 577 578 #define MULADDC_CORE \ 579 "ld.w %%d0, [%%a2+] \n\t" \ 580 "madd.u %%e2, %%e4, %%d0, %%d1 \n\t" \ 581 "ld.w %%d0, [%%a3] \n\t" \ 582 "addx %%d2, %%d2, %%d0 \n\t" \ 583 "addc %%d3, %%d3, 0 \n\t" \ 584 "mov %%d4, %%d3 \n\t" \ 585 "st.w [%%a3+], %%d2 \n\t" 586 587 #define MULADDC_STOP \ 588 "st.w %0, %%d4 \n\t" \ 589 "st.a %1, %%a3 \n\t" \ 590 "st.a %2, %%a2 \n\t" \ 591 : "=m" (c), "=m" (d), "=m" (s) \ 592 : "m" (s), "m" (d), "m" (c), "m" (b) \ 593 : "d0", "d1", "e2", "d4", "a2", "a3" \ 594 ); 595 596 #endif /* TriCore */ 597 598 /* 599 * Note, gcc -O0 by default uses r7 for the frame pointer, so it complains about 600 * our use of r7 below, unless -fomit-frame-pointer is passed. 601 * 602 * On the other hand, -fomit-frame-pointer is implied by any -Ox options with 603 * x !=0, which we can detect using __OPTIMIZE__ (which is also defined by 604 * clang and armcc5 under the same conditions). 605 * 606 * So, only use the optimized assembly below for optimized build, which avoids 607 * the build error and is pretty reasonable anyway. 608 */ 609 #if defined(__GNUC__) && !defined(__OPTIMIZE__) 610 #define MULADDC_CANNOT_USE_R7 611 #endif 612 613 #if defined(__arm__) && !defined(MULADDC_CANNOT_USE_R7) 614 615 #if defined(__thumb__) && !defined(__thumb2__) 616 617 #define MULADDC_INIT \ 618 asm( \ 619 "ldr r0, %3 \n\t" \ 620 "ldr r1, %4 \n\t" \ 621 "ldr r2, %5 \n\t" \ 622 "ldr r3, %6 \n\t" \ 623 "lsr r7, r3, #16 \n\t" \ 624 "mov r9, r7 \n\t" \ 625 "lsl r7, r3, #16 \n\t" \ 626 "lsr r7, r7, #16 \n\t" \ 627 "mov r8, r7 \n\t" 628 629 #define MULADDC_CORE \ 630 "ldmia r0!, {r6} \n\t" \ 631 "lsr r7, r6, #16 \n\t" \ 632 "lsl r6, r6, #16 \n\t" \ 633 "lsr r6, r6, #16 \n\t" \ 634 "mov r4, r8 \n\t" \ 635 "mul r4, r6 \n\t" \ 636 "mov r3, r9 \n\t" \ 637 "mul r6, r3 \n\t" \ 638 "mov r5, r9 \n\t" \ 639 "mul r5, r7 \n\t" \ 640 "mov r3, r8 \n\t" \ 641 "mul r7, r3 \n\t" \ 642 "lsr r3, r6, #16 \n\t" \ 643 "add r5, r5, r3 \n\t" \ 644 "lsr r3, r7, #16 \n\t" \ 645 "add r5, r5, r3 \n\t" \ 646 "add r4, r4, r2 \n\t" \ 647 "mov r2, #0 \n\t" \ 648 "adc r5, r2 \n\t" \ 649 "lsl r3, r6, #16 \n\t" \ 650 "add r4, r4, r3 \n\t" \ 651 "adc r5, r2 \n\t" \ 652 "lsl r3, r7, #16 \n\t" \ 653 "add r4, r4, r3 \n\t" \ 654 "adc r5, r2 \n\t" \ 655 "ldr r3, [r1] \n\t" \ 656 "add r4, r4, r3 \n\t" \ 657 "adc r2, r5 \n\t" \ 658 "stmia r1!, {r4} \n\t" 659 660 #define MULADDC_STOP \ 661 "str r2, %0 \n\t" \ 662 "str r1, %1 \n\t" \ 663 "str r0, %2 \n\t" \ 664 : "=m" (c), "=m" (d), "=m" (s) \ 665 : "m" (s), "m" (d), "m" (c), "m" (b) \ 666 : "r0", "r1", "r2", "r3", "r4", "r5", \ 667 "r6", "r7", "r8", "r9", "cc" \ 668 ); 669 670 #elif (__ARM_ARCH >= 6) && \ 671 defined (__ARM_FEATURE_DSP) && (__ARM_FEATURE_DSP == 1) 672 673 #define MULADDC_INIT \ 674 asm( 675 676 #define MULADDC_CORE \ 677 "ldr r0, [%0], #4 \n\t" \ 678 "ldr r1, [%1] \n\t" \ 679 "umaal r1, %2, %3, r0 \n\t" \ 680 "str r1, [%1], #4 \n\t" 681 682 #define MULADDC_STOP \ 683 : "=r" (s), "=r" (d), "=r" (c) \ 684 : "r" (b), "0" (s), "1" (d), "2" (c) \ 685 : "r0", "r1", "memory" \ 686 ); 687 688 #else 689 690 #define MULADDC_INIT \ 691 asm( \ 692 "ldr r0, %3 \n\t" \ 693 "ldr r1, %4 \n\t" \ 694 "ldr r2, %5 \n\t" \ 695 "ldr r3, %6 \n\t" 696 697 #define MULADDC_CORE \ 698 "ldr r4, [r0], #4 \n\t" \ 699 "mov r5, #0 \n\t" \ 700 "ldr r6, [r1] \n\t" \ 701 "umlal r2, r5, r3, r4 \n\t" \ 702 "adds r7, r6, r2 \n\t" \ 703 "adc r2, r5, #0 \n\t" \ 704 "str r7, [r1], #4 \n\t" 705 706 #define MULADDC_STOP \ 707 "str r2, %0 \n\t" \ 708 "str r1, %1 \n\t" \ 709 "str r0, %2 \n\t" \ 710 : "=m" (c), "=m" (d), "=m" (s) \ 711 : "m" (s), "m" (d), "m" (c), "m" (b) \ 712 : "r0", "r1", "r2", "r3", "r4", "r5", \ 713 "r6", "r7", "cc" \ 714 ); 715 716 #endif /* Thumb */ 717 718 #endif /* ARMv3 */ 719 720 #if defined(__alpha__) 721 722 #define MULADDC_INIT \ 723 asm( \ 724 "ldq $1, %3 \n\t" \ 725 "ldq $2, %4 \n\t" \ 726 "ldq $3, %5 \n\t" \ 727 "ldq $4, %6 \n\t" 728 729 #define MULADDC_CORE \ 730 "ldq $6, 0($1) \n\t" \ 731 "addq $1, 8, $1 \n\t" \ 732 "mulq $6, $4, $7 \n\t" \ 733 "umulh $6, $4, $6 \n\t" \ 734 "addq $7, $3, $7 \n\t" \ 735 "cmpult $7, $3, $3 \n\t" \ 736 "ldq $5, 0($2) \n\t" \ 737 "addq $7, $5, $7 \n\t" \ 738 "cmpult $7, $5, $5 \n\t" \ 739 "stq $7, 0($2) \n\t" \ 740 "addq $2, 8, $2 \n\t" \ 741 "addq $6, $3, $3 \n\t" \ 742 "addq $5, $3, $3 \n\t" 743 744 #define MULADDC_STOP \ 745 "stq $3, %0 \n\t" \ 746 "stq $2, %1 \n\t" \ 747 "stq $1, %2 \n\t" \ 748 : "=m" (c), "=m" (d), "=m" (s) \ 749 : "m" (s), "m" (d), "m" (c), "m" (b) \ 750 : "$1", "$2", "$3", "$4", "$5", "$6", "$7" \ 751 ); 752 #endif /* Alpha */ 753 754 #if defined(__mips__) && !defined(__mips64) 755 756 #define MULADDC_INIT \ 757 asm( \ 758 "lw $10, %3 \n\t" \ 759 "lw $11, %4 \n\t" \ 760 "lw $12, %5 \n\t" \ 761 "lw $13, %6 \n\t" 762 763 #define MULADDC_CORE \ 764 "lw $14, 0($10) \n\t" \ 765 "multu $13, $14 \n\t" \ 766 "addi $10, $10, 4 \n\t" \ 767 "mflo $14 \n\t" \ 768 "mfhi $9 \n\t" \ 769 "addu $14, $12, $14 \n\t" \ 770 "lw $15, 0($11) \n\t" \ 771 "sltu $12, $14, $12 \n\t" \ 772 "addu $15, $14, $15 \n\t" \ 773 "sltu $14, $15, $14 \n\t" \ 774 "addu $12, $12, $9 \n\t" \ 775 "sw $15, 0($11) \n\t" \ 776 "addu $12, $12, $14 \n\t" \ 777 "addi $11, $11, 4 \n\t" 778 779 #define MULADDC_STOP \ 780 "sw $12, %0 \n\t" \ 781 "sw $11, %1 \n\t" \ 782 "sw $10, %2 \n\t" \ 783 : "=m" (c), "=m" (d), "=m" (s) \ 784 : "m" (s), "m" (d), "m" (c), "m" (b) \ 785 : "$9", "$10", "$11", "$12", "$13", "$14", "$15", "lo", "hi" \ 786 ); 787 788 #endif /* MIPS */ 789 #endif /* GNUC */ 790 791 #if (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__) 792 793 #define MULADDC_INIT \ 794 __asm mov esi, s \ 795 __asm mov edi, d \ 796 __asm mov ecx, c \ 797 __asm mov ebx, b 798 799 #define MULADDC_CORE \ 800 __asm lodsd \ 801 __asm mul ebx \ 802 __asm add eax, ecx \ 803 __asm adc edx, 0 \ 804 __asm add eax, [edi] \ 805 __asm adc edx, 0 \ 806 __asm mov ecx, edx \ 807 __asm stosd 808 809 #if defined(MBEDTLS_HAVE_SSE2) 810 811 #define EMIT __asm _emit 812 813 #define MULADDC_HUIT \ 814 EMIT 0x0F EMIT 0x6E EMIT 0xC9 \ 815 EMIT 0x0F EMIT 0x6E EMIT 0xC3 \ 816 EMIT 0x0F EMIT 0x6E EMIT 0x1F \ 817 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \ 818 EMIT 0x0F EMIT 0x6E EMIT 0x16 \ 819 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \ 820 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x04 \ 821 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \ 822 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x08 \ 823 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \ 824 EMIT 0x0F EMIT 0x6E EMIT 0x7E EMIT 0x0C \ 825 EMIT 0x0F EMIT 0xF4 EMIT 0xF8 \ 826 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \ 827 EMIT 0x0F EMIT 0x6E EMIT 0x5F EMIT 0x04 \ 828 EMIT 0x0F EMIT 0xD4 EMIT 0xDC \ 829 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x08 \ 830 EMIT 0x0F EMIT 0xD4 EMIT 0xEE \ 831 EMIT 0x0F EMIT 0x6E EMIT 0x67 EMIT 0x0C \ 832 EMIT 0x0F EMIT 0xD4 EMIT 0xFC \ 833 EMIT 0x0F EMIT 0x7E EMIT 0x0F \ 834 EMIT 0x0F EMIT 0x6E EMIT 0x56 EMIT 0x10 \ 835 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \ 836 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 837 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x14 \ 838 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \ 839 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \ 840 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x18 \ 841 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \ 842 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x04 \ 843 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 844 EMIT 0x0F EMIT 0x6E EMIT 0x5E EMIT 0x1C \ 845 EMIT 0x0F EMIT 0xF4 EMIT 0xD8 \ 846 EMIT 0x0F EMIT 0xD4 EMIT 0xCD \ 847 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x10 \ 848 EMIT 0x0F EMIT 0xD4 EMIT 0xD5 \ 849 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x08 \ 850 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 851 EMIT 0x0F EMIT 0xD4 EMIT 0xCF \ 852 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x14 \ 853 EMIT 0x0F EMIT 0xD4 EMIT 0xE5 \ 854 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x0C \ 855 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 856 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \ 857 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x18 \ 858 EMIT 0x0F EMIT 0xD4 EMIT 0xF5 \ 859 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x10 \ 860 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 861 EMIT 0x0F EMIT 0xD4 EMIT 0xCC \ 862 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x1C \ 863 EMIT 0x0F EMIT 0xD4 EMIT 0xDD \ 864 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x14 \ 865 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 866 EMIT 0x0F EMIT 0xD4 EMIT 0xCE \ 867 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x18 \ 868 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 869 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \ 870 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x1C \ 871 EMIT 0x83 EMIT 0xC7 EMIT 0x20 \ 872 EMIT 0x83 EMIT 0xC6 EMIT 0x20 \ 873 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 874 EMIT 0x0F EMIT 0x7E EMIT 0xC9 875 876 #define MULADDC_STOP \ 877 EMIT 0x0F EMIT 0x77 \ 878 __asm mov c, ecx \ 879 __asm mov d, edi \ 880 __asm mov s, esi \ 881 882 #else 883 884 #define MULADDC_STOP \ 885 __asm mov c, ecx \ 886 __asm mov d, edi \ 887 __asm mov s, esi \ 888 889 #endif /* SSE2 */ 890 #endif /* MSVC */ 891 892 #endif /* MBEDTLS_HAVE_ASM */ 893 894 #if !defined(MULADDC_CORE) 895 #if defined(MBEDTLS_HAVE_UDBL) 896 897 #define MULADDC_INIT \ 898 { \ 899 mbedtls_t_udbl r; \ 900 mbedtls_mpi_uint r0, r1; 901 902 #define MULADDC_CORE \ 903 r = *(s++) * (mbedtls_t_udbl) b; \ 904 r0 = (mbedtls_mpi_uint) r; \ 905 r1 = (mbedtls_mpi_uint)( r >> biL ); \ 906 r0 += c; r1 += (r0 < c); \ 907 r0 += *d; r1 += (r0 < *d); \ 908 c = r1; *(d++) = r0; 909 910 #define MULADDC_STOP \ 911 } 912 913 #else 914 #define MULADDC_INIT \ 915 { \ 916 mbedtls_mpi_uint s0, s1, b0, b1; \ 917 mbedtls_mpi_uint r0, r1, rx, ry; \ 918 b0 = ( b << biH ) >> biH; \ 919 b1 = ( b >> biH ); 920 921 #define MULADDC_CORE \ 922 s0 = ( *s << biH ) >> biH; \ 923 s1 = ( *s >> biH ); s++; \ 924 rx = s0 * b1; r0 = s0 * b0; \ 925 ry = s1 * b0; r1 = s1 * b1; \ 926 r1 += ( rx >> biH ); \ 927 r1 += ( ry >> biH ); \ 928 rx <<= biH; ry <<= biH; \ 929 r0 += rx; r1 += (r0 < rx); \ 930 r0 += ry; r1 += (r0 < ry); \ 931 r0 += c; r1 += (r0 < c); \ 932 r0 += *d; r1 += (r0 < *d); \ 933 c = r1; *(d++) = r0; 934 935 #define MULADDC_STOP \ 936 } 937 938 #endif /* C (generic) */ 939 #endif /* C (longlong) */ 940 941 #endif /* bn_mul.h */ 942