1 /** 2 * \file bn_mul.h 3 * 4 * \brief Multi-precision integer library 5 */ 6 /* 7 * Copyright (C) 2006-2015, ARM Limited, All Rights Reserved 8 * SPDX-License-Identifier: Apache-2.0 9 * 10 * Licensed under the Apache License, Version 2.0 (the "License"); you may 11 * not use this file except in compliance with the License. 12 * You may obtain a copy of the License at 13 * 14 * http://www.apache.org/licenses/LICENSE-2.0 15 * 16 * Unless required by applicable law or agreed to in writing, software 17 * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 18 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 19 * See the License for the specific language governing permissions and 20 * limitations under the License. 21 * 22 * This file is part of mbed TLS (https://tls.mbed.org) 23 */ 24 /* 25 * Multiply source vector [s] with b, add result 26 * to destination vector [d] and set carry c. 27 * 28 * Currently supports: 29 * 30 * . IA-32 (386+) . AMD64 / EM64T 31 * . IA-32 (SSE2) . Motorola 68000 32 * . PowerPC, 32-bit . MicroBlaze 33 * . PowerPC, 64-bit . TriCore 34 * . SPARC v8 . ARM v3+ 35 * . Alpha . MIPS32 36 * . C, longlong . C, generic 37 */ 38 #ifndef MBEDTLS_BN_MUL_H 39 #define MBEDTLS_BN_MUL_H 40 41 #if !defined(MBEDTLS_CONFIG_FILE) 42 #include "config.h" 43 #else 44 #include MBEDTLS_CONFIG_FILE 45 #endif 46 47 #include "bignum.h" 48 49 #if defined(MBEDTLS_HAVE_ASM) 50 51 #ifndef asm 52 #define asm __asm 53 #endif 54 55 /* armcc5 --gnu defines __GNUC__ but doesn't support GNU's extended asm */ 56 #if defined(__GNUC__) && \ 57 ( !defined(__ARMCC_VERSION) || __ARMCC_VERSION >= 6000000 ) 58 59 /* 60 * Disable use of the i386 assembly code below if option -O0, to disable all 61 * compiler optimisations, is passed, detected with __OPTIMIZE__ 62 * This is done as the number of registers used in the assembly code doesn't 63 * work with the -O0 option. 64 */ 65 #if defined(__i386__) && defined(__OPTIMIZE__) 66 67 #define MULADDC_INIT \ 68 asm( \ 69 "movl %%ebx, %0 \n\t" \ 70 "movl %5, %%esi \n\t" \ 71 "movl %6, %%edi \n\t" \ 72 "movl %7, %%ecx \n\t" \ 73 "movl %8, %%ebx \n\t" 74 75 #define MULADDC_CORE \ 76 "lodsl \n\t" \ 77 "mull %%ebx \n\t" \ 78 "addl %%ecx, %%eax \n\t" \ 79 "adcl $0, %%edx \n\t" \ 80 "addl (%%edi), %%eax \n\t" \ 81 "adcl $0, %%edx \n\t" \ 82 "movl %%edx, %%ecx \n\t" \ 83 "stosl \n\t" 84 85 #if defined(MBEDTLS_HAVE_SSE2) 86 87 #define MULADDC_HUIT \ 88 "movd %%ecx, %%mm1 \n\t" \ 89 "movd %%ebx, %%mm0 \n\t" \ 90 "movd (%%edi), %%mm3 \n\t" \ 91 "paddq %%mm3, %%mm1 \n\t" \ 92 "movd (%%esi), %%mm2 \n\t" \ 93 "pmuludq %%mm0, %%mm2 \n\t" \ 94 "movd 4(%%esi), %%mm4 \n\t" \ 95 "pmuludq %%mm0, %%mm4 \n\t" \ 96 "movd 8(%%esi), %%mm6 \n\t" \ 97 "pmuludq %%mm0, %%mm6 \n\t" \ 98 "movd 12(%%esi), %%mm7 \n\t" \ 99 "pmuludq %%mm0, %%mm7 \n\t" \ 100 "paddq %%mm2, %%mm1 \n\t" \ 101 "movd 4(%%edi), %%mm3 \n\t" \ 102 "paddq %%mm4, %%mm3 \n\t" \ 103 "movd 8(%%edi), %%mm5 \n\t" \ 104 "paddq %%mm6, %%mm5 \n\t" \ 105 "movd 12(%%edi), %%mm4 \n\t" \ 106 "paddq %%mm4, %%mm7 \n\t" \ 107 "movd %%mm1, (%%edi) \n\t" \ 108 "movd 16(%%esi), %%mm2 \n\t" \ 109 "pmuludq %%mm0, %%mm2 \n\t" \ 110 "psrlq $32, %%mm1 \n\t" \ 111 "movd 20(%%esi), %%mm4 \n\t" \ 112 "pmuludq %%mm0, %%mm4 \n\t" \ 113 "paddq %%mm3, %%mm1 \n\t" \ 114 "movd 24(%%esi), %%mm6 \n\t" \ 115 "pmuludq %%mm0, %%mm6 \n\t" \ 116 "movd %%mm1, 4(%%edi) \n\t" \ 117 "psrlq $32, %%mm1 \n\t" \ 118 "movd 28(%%esi), %%mm3 \n\t" \ 119 "pmuludq %%mm0, %%mm3 \n\t" \ 120 "paddq %%mm5, %%mm1 \n\t" \ 121 "movd 16(%%edi), %%mm5 \n\t" \ 122 "paddq %%mm5, %%mm2 \n\t" \ 123 "movd %%mm1, 8(%%edi) \n\t" \ 124 "psrlq $32, %%mm1 \n\t" \ 125 "paddq %%mm7, %%mm1 \n\t" \ 126 "movd 20(%%edi), %%mm5 \n\t" \ 127 "paddq %%mm5, %%mm4 \n\t" \ 128 "movd %%mm1, 12(%%edi) \n\t" \ 129 "psrlq $32, %%mm1 \n\t" \ 130 "paddq %%mm2, %%mm1 \n\t" \ 131 "movd 24(%%edi), %%mm5 \n\t" \ 132 "paddq %%mm5, %%mm6 \n\t" \ 133 "movd %%mm1, 16(%%edi) \n\t" \ 134 "psrlq $32, %%mm1 \n\t" \ 135 "paddq %%mm4, %%mm1 \n\t" \ 136 "movd 28(%%edi), %%mm5 \n\t" \ 137 "paddq %%mm5, %%mm3 \n\t" \ 138 "movd %%mm1, 20(%%edi) \n\t" \ 139 "psrlq $32, %%mm1 \n\t" \ 140 "paddq %%mm6, %%mm1 \n\t" \ 141 "movd %%mm1, 24(%%edi) \n\t" \ 142 "psrlq $32, %%mm1 \n\t" \ 143 "paddq %%mm3, %%mm1 \n\t" \ 144 "movd %%mm1, 28(%%edi) \n\t" \ 145 "addl $32, %%edi \n\t" \ 146 "addl $32, %%esi \n\t" \ 147 "psrlq $32, %%mm1 \n\t" \ 148 "movd %%mm1, %%ecx \n\t" 149 150 #define MULADDC_STOP \ 151 "emms \n\t" \ 152 "movl %4, %%ebx \n\t" \ 153 "movl %%ecx, %1 \n\t" \ 154 "movl %%edi, %2 \n\t" \ 155 "movl %%esi, %3 \n\t" \ 156 : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \ 157 : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \ 158 : "eax", "ebx", "ecx", "edx", "esi", "edi" \ 159 ); 160 161 #else 162 163 #define MULADDC_STOP \ 164 "movl %4, %%ebx \n\t" \ 165 "movl %%ecx, %1 \n\t" \ 166 "movl %%edi, %2 \n\t" \ 167 "movl %%esi, %3 \n\t" \ 168 : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \ 169 : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \ 170 : "eax", "ebx", "ecx", "edx", "esi", "edi" \ 171 ); 172 #endif /* SSE2 */ 173 #endif /* i386 */ 174 175 #if defined(__amd64__) || defined (__x86_64__) 176 177 #define MULADDC_INIT \ 178 asm( \ 179 "xorq %%r8, %%r8\n" 180 181 #define MULADDC_CORE \ 182 "movq (%%rsi), %%rax\n" \ 183 "mulq %%rbx\n" \ 184 "addq $8, %%rsi\n" \ 185 "addq %%rcx, %%rax\n" \ 186 "movq %%r8, %%rcx\n" \ 187 "adcq $0, %%rdx\n" \ 188 "nop \n" \ 189 "addq %%rax, (%%rdi)\n" \ 190 "adcq %%rdx, %%rcx\n" \ 191 "addq $8, %%rdi\n" 192 193 #define MULADDC_STOP \ 194 : "+c" (c), "+D" (d), "+S" (s) \ 195 : "b" (b) \ 196 : "rax", "rdx", "r8" \ 197 ); 198 199 #endif /* AMD64 */ 200 201 #if defined(__mc68020__) || defined(__mcpu32__) 202 203 #define MULADDC_INIT \ 204 asm( \ 205 "movl %3, %%a2 \n\t" \ 206 "movl %4, %%a3 \n\t" \ 207 "movl %5, %%d3 \n\t" \ 208 "movl %6, %%d2 \n\t" \ 209 "moveq #0, %%d0 \n\t" 210 211 #define MULADDC_CORE \ 212 "movel %%a2@+, %%d1 \n\t" \ 213 "mulul %%d2, %%d4:%%d1 \n\t" \ 214 "addl %%d3, %%d1 \n\t" \ 215 "addxl %%d0, %%d4 \n\t" \ 216 "moveq #0, %%d3 \n\t" \ 217 "addl %%d1, %%a3@+ \n\t" \ 218 "addxl %%d4, %%d3 \n\t" 219 220 #define MULADDC_STOP \ 221 "movl %%d3, %0 \n\t" \ 222 "movl %%a3, %1 \n\t" \ 223 "movl %%a2, %2 \n\t" \ 224 : "=m" (c), "=m" (d), "=m" (s) \ 225 : "m" (s), "m" (d), "m" (c), "m" (b) \ 226 : "d0", "d1", "d2", "d3", "d4", "a2", "a3" \ 227 ); 228 229 #define MULADDC_HUIT \ 230 "movel %%a2@+, %%d1 \n\t" \ 231 "mulul %%d2, %%d4:%%d1 \n\t" \ 232 "addxl %%d3, %%d1 \n\t" \ 233 "addxl %%d0, %%d4 \n\t" \ 234 "addl %%d1, %%a3@+ \n\t" \ 235 "movel %%a2@+, %%d1 \n\t" \ 236 "mulul %%d2, %%d3:%%d1 \n\t" \ 237 "addxl %%d4, %%d1 \n\t" \ 238 "addxl %%d0, %%d3 \n\t" \ 239 "addl %%d1, %%a3@+ \n\t" \ 240 "movel %%a2@+, %%d1 \n\t" \ 241 "mulul %%d2, %%d4:%%d1 \n\t" \ 242 "addxl %%d3, %%d1 \n\t" \ 243 "addxl %%d0, %%d4 \n\t" \ 244 "addl %%d1, %%a3@+ \n\t" \ 245 "movel %%a2@+, %%d1 \n\t" \ 246 "mulul %%d2, %%d3:%%d1 \n\t" \ 247 "addxl %%d4, %%d1 \n\t" \ 248 "addxl %%d0, %%d3 \n\t" \ 249 "addl %%d1, %%a3@+ \n\t" \ 250 "movel %%a2@+, %%d1 \n\t" \ 251 "mulul %%d2, %%d4:%%d1 \n\t" \ 252 "addxl %%d3, %%d1 \n\t" \ 253 "addxl %%d0, %%d4 \n\t" \ 254 "addl %%d1, %%a3@+ \n\t" \ 255 "movel %%a2@+, %%d1 \n\t" \ 256 "mulul %%d2, %%d3:%%d1 \n\t" \ 257 "addxl %%d4, %%d1 \n\t" \ 258 "addxl %%d0, %%d3 \n\t" \ 259 "addl %%d1, %%a3@+ \n\t" \ 260 "movel %%a2@+, %%d1 \n\t" \ 261 "mulul %%d2, %%d4:%%d1 \n\t" \ 262 "addxl %%d3, %%d1 \n\t" \ 263 "addxl %%d0, %%d4 \n\t" \ 264 "addl %%d1, %%a3@+ \n\t" \ 265 "movel %%a2@+, %%d1 \n\t" \ 266 "mulul %%d2, %%d3:%%d1 \n\t" \ 267 "addxl %%d4, %%d1 \n\t" \ 268 "addxl %%d0, %%d3 \n\t" \ 269 "addl %%d1, %%a3@+ \n\t" \ 270 "addxl %%d0, %%d3 \n\t" 271 272 #endif /* MC68000 */ 273 274 #if defined(__powerpc64__) || defined(__ppc64__) 275 276 #if defined(__MACH__) && defined(__APPLE__) 277 278 #define MULADDC_INIT \ 279 asm( \ 280 "ld r3, %3 \n\t" \ 281 "ld r4, %4 \n\t" \ 282 "ld r5, %5 \n\t" \ 283 "ld r6, %6 \n\t" \ 284 "addi r3, r3, -8 \n\t" \ 285 "addi r4, r4, -8 \n\t" \ 286 "addic r5, r5, 0 \n\t" 287 288 #define MULADDC_CORE \ 289 "ldu r7, 8(r3) \n\t" \ 290 "mulld r8, r7, r6 \n\t" \ 291 "mulhdu r9, r7, r6 \n\t" \ 292 "adde r8, r8, r5 \n\t" \ 293 "ld r7, 8(r4) \n\t" \ 294 "addze r5, r9 \n\t" \ 295 "addc r8, r8, r7 \n\t" \ 296 "stdu r8, 8(r4) \n\t" 297 298 #define MULADDC_STOP \ 299 "addze r5, r5 \n\t" \ 300 "addi r4, r4, 8 \n\t" \ 301 "addi r3, r3, 8 \n\t" \ 302 "std r5, %0 \n\t" \ 303 "std r4, %1 \n\t" \ 304 "std r3, %2 \n\t" \ 305 : "=m" (c), "=m" (d), "=m" (s) \ 306 : "m" (s), "m" (d), "m" (c), "m" (b) \ 307 : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \ 308 ); 309 310 311 #else /* __MACH__ && __APPLE__ */ 312 313 #define MULADDC_INIT \ 314 asm( \ 315 "ld %%r3, %3 \n\t" \ 316 "ld %%r4, %4 \n\t" \ 317 "ld %%r5, %5 \n\t" \ 318 "ld %%r6, %6 \n\t" \ 319 "addi %%r3, %%r3, -8 \n\t" \ 320 "addi %%r4, %%r4, -8 \n\t" \ 321 "addic %%r5, %%r5, 0 \n\t" 322 323 #define MULADDC_CORE \ 324 "ldu %%r7, 8(%%r3) \n\t" \ 325 "mulld %%r8, %%r7, %%r6 \n\t" \ 326 "mulhdu %%r9, %%r7, %%r6 \n\t" \ 327 "adde %%r8, %%r8, %%r5 \n\t" \ 328 "ld %%r7, 8(%%r4) \n\t" \ 329 "addze %%r5, %%r9 \n\t" \ 330 "addc %%r8, %%r8, %%r7 \n\t" \ 331 "stdu %%r8, 8(%%r4) \n\t" 332 333 #define MULADDC_STOP \ 334 "addze %%r5, %%r5 \n\t" \ 335 "addi %%r4, %%r4, 8 \n\t" \ 336 "addi %%r3, %%r3, 8 \n\t" \ 337 "std %%r5, %0 \n\t" \ 338 "std %%r4, %1 \n\t" \ 339 "std %%r3, %2 \n\t" \ 340 : "=m" (c), "=m" (d), "=m" (s) \ 341 : "m" (s), "m" (d), "m" (c), "m" (b) \ 342 : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \ 343 ); 344 345 #endif /* __MACH__ && __APPLE__ */ 346 347 #elif defined(__powerpc__) || defined(__ppc__) /* end PPC64/begin PPC32 */ 348 349 #if defined(__MACH__) && defined(__APPLE__) 350 351 #define MULADDC_INIT \ 352 asm( \ 353 "lwz r3, %3 \n\t" \ 354 "lwz r4, %4 \n\t" \ 355 "lwz r5, %5 \n\t" \ 356 "lwz r6, %6 \n\t" \ 357 "addi r3, r3, -4 \n\t" \ 358 "addi r4, r4, -4 \n\t" \ 359 "addic r5, r5, 0 \n\t" 360 361 #define MULADDC_CORE \ 362 "lwzu r7, 4(r3) \n\t" \ 363 "mullw r8, r7, r6 \n\t" \ 364 "mulhwu r9, r7, r6 \n\t" \ 365 "adde r8, r8, r5 \n\t" \ 366 "lwz r7, 4(r4) \n\t" \ 367 "addze r5, r9 \n\t" \ 368 "addc r8, r8, r7 \n\t" \ 369 "stwu r8, 4(r4) \n\t" 370 371 #define MULADDC_STOP \ 372 "addze r5, r5 \n\t" \ 373 "addi r4, r4, 4 \n\t" \ 374 "addi r3, r3, 4 \n\t" \ 375 "stw r5, %0 \n\t" \ 376 "stw r4, %1 \n\t" \ 377 "stw r3, %2 \n\t" \ 378 : "=m" (c), "=m" (d), "=m" (s) \ 379 : "m" (s), "m" (d), "m" (c), "m" (b) \ 380 : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \ 381 ); 382 383 #else /* __MACH__ && __APPLE__ */ 384 385 #define MULADDC_INIT \ 386 asm( \ 387 "lwz %%r3, %3 \n\t" \ 388 "lwz %%r4, %4 \n\t" \ 389 "lwz %%r5, %5 \n\t" \ 390 "lwz %%r6, %6 \n\t" \ 391 "addi %%r3, %%r3, -4 \n\t" \ 392 "addi %%r4, %%r4, -4 \n\t" \ 393 "addic %%r5, %%r5, 0 \n\t" 394 395 #define MULADDC_CORE \ 396 "lwzu %%r7, 4(%%r3) \n\t" \ 397 "mullw %%r8, %%r7, %%r6 \n\t" \ 398 "mulhwu %%r9, %%r7, %%r6 \n\t" \ 399 "adde %%r8, %%r8, %%r5 \n\t" \ 400 "lwz %%r7, 4(%%r4) \n\t" \ 401 "addze %%r5, %%r9 \n\t" \ 402 "addc %%r8, %%r8, %%r7 \n\t" \ 403 "stwu %%r8, 4(%%r4) \n\t" 404 405 #define MULADDC_STOP \ 406 "addze %%r5, %%r5 \n\t" \ 407 "addi %%r4, %%r4, 4 \n\t" \ 408 "addi %%r3, %%r3, 4 \n\t" \ 409 "stw %%r5, %0 \n\t" \ 410 "stw %%r4, %1 \n\t" \ 411 "stw %%r3, %2 \n\t" \ 412 : "=m" (c), "=m" (d), "=m" (s) \ 413 : "m" (s), "m" (d), "m" (c), "m" (b) \ 414 : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \ 415 ); 416 417 #endif /* __MACH__ && __APPLE__ */ 418 419 #endif /* PPC32 */ 420 421 /* 422 * The Sparc(64) assembly is reported to be broken. 423 * Disable it for now, until we're able to fix it. 424 */ 425 #if 0 && defined(__sparc__) 426 #if defined(__sparc64__) 427 428 #define MULADDC_INIT \ 429 asm( \ 430 "ldx %3, %%o0 \n\t" \ 431 "ldx %4, %%o1 \n\t" \ 432 "ld %5, %%o2 \n\t" \ 433 "ld %6, %%o3 \n\t" 434 435 #define MULADDC_CORE \ 436 "ld [%%o0], %%o4 \n\t" \ 437 "inc 4, %%o0 \n\t" \ 438 "ld [%%o1], %%o5 \n\t" \ 439 "umul %%o3, %%o4, %%o4 \n\t" \ 440 "addcc %%o4, %%o2, %%o4 \n\t" \ 441 "rd %%y, %%g1 \n\t" \ 442 "addx %%g1, 0, %%g1 \n\t" \ 443 "addcc %%o4, %%o5, %%o4 \n\t" \ 444 "st %%o4, [%%o1] \n\t" \ 445 "addx %%g1, 0, %%o2 \n\t" \ 446 "inc 4, %%o1 \n\t" 447 448 #define MULADDC_STOP \ 449 "st %%o2, %0 \n\t" \ 450 "stx %%o1, %1 \n\t" \ 451 "stx %%o0, %2 \n\t" \ 452 : "=m" (c), "=m" (d), "=m" (s) \ 453 : "m" (s), "m" (d), "m" (c), "m" (b) \ 454 : "g1", "o0", "o1", "o2", "o3", "o4", \ 455 "o5" \ 456 ); 457 458 #else /* __sparc64__ */ 459 460 #define MULADDC_INIT \ 461 asm( \ 462 "ld %3, %%o0 \n\t" \ 463 "ld %4, %%o1 \n\t" \ 464 "ld %5, %%o2 \n\t" \ 465 "ld %6, %%o3 \n\t" 466 467 #define MULADDC_CORE \ 468 "ld [%%o0], %%o4 \n\t" \ 469 "inc 4, %%o0 \n\t" \ 470 "ld [%%o1], %%o5 \n\t" \ 471 "umul %%o3, %%o4, %%o4 \n\t" \ 472 "addcc %%o4, %%o2, %%o4 \n\t" \ 473 "rd %%y, %%g1 \n\t" \ 474 "addx %%g1, 0, %%g1 \n\t" \ 475 "addcc %%o4, %%o5, %%o4 \n\t" \ 476 "st %%o4, [%%o1] \n\t" \ 477 "addx %%g1, 0, %%o2 \n\t" \ 478 "inc 4, %%o1 \n\t" 479 480 #define MULADDC_STOP \ 481 "st %%o2, %0 \n\t" \ 482 "st %%o1, %1 \n\t" \ 483 "st %%o0, %2 \n\t" \ 484 : "=m" (c), "=m" (d), "=m" (s) \ 485 : "m" (s), "m" (d), "m" (c), "m" (b) \ 486 : "g1", "o0", "o1", "o2", "o3", "o4", \ 487 "o5" \ 488 ); 489 490 #endif /* __sparc64__ */ 491 #endif /* __sparc__ */ 492 493 #if defined(__microblaze__) || defined(microblaze) 494 495 #define MULADDC_INIT \ 496 asm( \ 497 "lwi r3, %3 \n\t" \ 498 "lwi r4, %4 \n\t" \ 499 "lwi r5, %5 \n\t" \ 500 "lwi r6, %6 \n\t" \ 501 "andi r7, r6, 0xffff \n\t" \ 502 "bsrli r6, r6, 16 \n\t" 503 504 #define MULADDC_CORE \ 505 "lhui r8, r3, 0 \n\t" \ 506 "addi r3, r3, 2 \n\t" \ 507 "lhui r9, r3, 0 \n\t" \ 508 "addi r3, r3, 2 \n\t" \ 509 "mul r10, r9, r6 \n\t" \ 510 "mul r11, r8, r7 \n\t" \ 511 "mul r12, r9, r7 \n\t" \ 512 "mul r13, r8, r6 \n\t" \ 513 "bsrli r8, r10, 16 \n\t" \ 514 "bsrli r9, r11, 16 \n\t" \ 515 "add r13, r13, r8 \n\t" \ 516 "add r13, r13, r9 \n\t" \ 517 "bslli r10, r10, 16 \n\t" \ 518 "bslli r11, r11, 16 \n\t" \ 519 "add r12, r12, r10 \n\t" \ 520 "addc r13, r13, r0 \n\t" \ 521 "add r12, r12, r11 \n\t" \ 522 "addc r13, r13, r0 \n\t" \ 523 "lwi r10, r4, 0 \n\t" \ 524 "add r12, r12, r10 \n\t" \ 525 "addc r13, r13, r0 \n\t" \ 526 "add r12, r12, r5 \n\t" \ 527 "addc r5, r13, r0 \n\t" \ 528 "swi r12, r4, 0 \n\t" \ 529 "addi r4, r4, 4 \n\t" 530 531 #define MULADDC_STOP \ 532 "swi r5, %0 \n\t" \ 533 "swi r4, %1 \n\t" \ 534 "swi r3, %2 \n\t" \ 535 : "=m" (c), "=m" (d), "=m" (s) \ 536 : "m" (s), "m" (d), "m" (c), "m" (b) \ 537 : "r3", "r4", "r5", "r6", "r7", "r8", \ 538 "r9", "r10", "r11", "r12", "r13" \ 539 ); 540 541 #endif /* MicroBlaze */ 542 543 #if defined(__tricore__) 544 545 #define MULADDC_INIT \ 546 asm( \ 547 "ld.a %%a2, %3 \n\t" \ 548 "ld.a %%a3, %4 \n\t" \ 549 "ld.w %%d4, %5 \n\t" \ 550 "ld.w %%d1, %6 \n\t" \ 551 "xor %%d5, %%d5 \n\t" 552 553 #define MULADDC_CORE \ 554 "ld.w %%d0, [%%a2+] \n\t" \ 555 "madd.u %%e2, %%e4, %%d0, %%d1 \n\t" \ 556 "ld.w %%d0, [%%a3] \n\t" \ 557 "addx %%d2, %%d2, %%d0 \n\t" \ 558 "addc %%d3, %%d3, 0 \n\t" \ 559 "mov %%d4, %%d3 \n\t" \ 560 "st.w [%%a3+], %%d2 \n\t" 561 562 #define MULADDC_STOP \ 563 "st.w %0, %%d4 \n\t" \ 564 "st.a %1, %%a3 \n\t" \ 565 "st.a %2, %%a2 \n\t" \ 566 : "=m" (c), "=m" (d), "=m" (s) \ 567 : "m" (s), "m" (d), "m" (c), "m" (b) \ 568 : "d0", "d1", "e2", "d4", "a2", "a3" \ 569 ); 570 571 #endif /* TriCore */ 572 573 /* 574 * Note, gcc -O0 by default uses r7 for the frame pointer, so it complains about 575 * our use of r7 below, unless -fomit-frame-pointer is passed. 576 * 577 * On the other hand, -fomit-frame-pointer is implied by any -Ox options with 578 * x !=0, which we can detect using __OPTIMIZE__ (which is also defined by 579 * clang and armcc5 under the same conditions). 580 * 581 * So, only use the optimized assembly below for optimized build, which avoids 582 * the build error and is pretty reasonable anyway. 583 */ 584 #if defined(__GNUC__) && !defined(__OPTIMIZE__) 585 #define MULADDC_CANNOT_USE_R7 586 #endif 587 588 #if defined(__arm__) && !defined(MULADDC_CANNOT_USE_R7) 589 590 #if defined(__thumb__) && !defined(__thumb2__) 591 592 #define MULADDC_INIT \ 593 asm( \ 594 "ldr r0, %3 \n\t" \ 595 "ldr r1, %4 \n\t" \ 596 "ldr r2, %5 \n\t" \ 597 "ldr r3, %6 \n\t" \ 598 "lsr r7, r3, #16 \n\t" \ 599 "mov r9, r7 \n\t" \ 600 "lsl r7, r3, #16 \n\t" \ 601 "lsr r7, r7, #16 \n\t" \ 602 "mov r8, r7 \n\t" 603 604 #define MULADDC_CORE \ 605 "ldmia r0!, {r6} \n\t" \ 606 "lsr r7, r6, #16 \n\t" \ 607 "lsl r6, r6, #16 \n\t" \ 608 "lsr r6, r6, #16 \n\t" \ 609 "mov r4, r8 \n\t" \ 610 "mul r4, r6 \n\t" \ 611 "mov r3, r9 \n\t" \ 612 "mul r6, r3 \n\t" \ 613 "mov r5, r9 \n\t" \ 614 "mul r5, r7 \n\t" \ 615 "mov r3, r8 \n\t" \ 616 "mul r7, r3 \n\t" \ 617 "lsr r3, r6, #16 \n\t" \ 618 "add r5, r5, r3 \n\t" \ 619 "lsr r3, r7, #16 \n\t" \ 620 "add r5, r5, r3 \n\t" \ 621 "add r4, r4, r2 \n\t" \ 622 "mov r2, #0 \n\t" \ 623 "adc r5, r2 \n\t" \ 624 "lsl r3, r6, #16 \n\t" \ 625 "add r4, r4, r3 \n\t" \ 626 "adc r5, r2 \n\t" \ 627 "lsl r3, r7, #16 \n\t" \ 628 "add r4, r4, r3 \n\t" \ 629 "adc r5, r2 \n\t" \ 630 "ldr r3, [r1] \n\t" \ 631 "add r4, r4, r3 \n\t" \ 632 "adc r2, r5 \n\t" \ 633 "stmia r1!, {r4} \n\t" 634 635 #define MULADDC_STOP \ 636 "str r2, %0 \n\t" \ 637 "str r1, %1 \n\t" \ 638 "str r0, %2 \n\t" \ 639 : "=m" (c), "=m" (d), "=m" (s) \ 640 : "m" (s), "m" (d), "m" (c), "m" (b) \ 641 : "r0", "r1", "r2", "r3", "r4", "r5", \ 642 "r6", "r7", "r8", "r9", "cc" \ 643 ); 644 645 #elif (__ARM_ARCH >= 6) && \ 646 defined (__ARM_FEATURE_DSP) && (__ARM_FEATURE_DSP == 1) 647 648 #define MULADDC_INIT \ 649 asm( 650 651 #define MULADDC_CORE \ 652 "ldr r0, [%0], #4 \n\t" \ 653 "ldr r1, [%1] \n\t" \ 654 "umaal r1, %2, %3, r0 \n\t" \ 655 "str r1, [%1], #4 \n\t" 656 657 #define MULADDC_STOP \ 658 : "=r" (s), "=r" (d), "=r" (c) \ 659 : "r" (b), "0" (s), "1" (d), "2" (c) \ 660 : "r0", "r1", "memory" \ 661 ); 662 663 #else 664 665 #define MULADDC_INIT \ 666 asm( \ 667 "ldr r0, %3 \n\t" \ 668 "ldr r1, %4 \n\t" \ 669 "ldr r2, %5 \n\t" \ 670 "ldr r3, %6 \n\t" 671 672 #define MULADDC_CORE \ 673 "ldr r4, [r0], #4 \n\t" \ 674 "mov r5, #0 \n\t" \ 675 "ldr r6, [r1] \n\t" \ 676 "umlal r2, r5, r3, r4 \n\t" \ 677 "adds r7, r6, r2 \n\t" \ 678 "adc r2, r5, #0 \n\t" \ 679 "str r7, [r1], #4 \n\t" 680 681 #define MULADDC_STOP \ 682 "str r2, %0 \n\t" \ 683 "str r1, %1 \n\t" \ 684 "str r0, %2 \n\t" \ 685 : "=m" (c), "=m" (d), "=m" (s) \ 686 : "m" (s), "m" (d), "m" (c), "m" (b) \ 687 : "r0", "r1", "r2", "r3", "r4", "r5", \ 688 "r6", "r7", "cc" \ 689 ); 690 691 #endif /* Thumb */ 692 693 #endif /* ARMv3 */ 694 695 #if defined(__alpha__) 696 697 #define MULADDC_INIT \ 698 asm( \ 699 "ldq $1, %3 \n\t" \ 700 "ldq $2, %4 \n\t" \ 701 "ldq $3, %5 \n\t" \ 702 "ldq $4, %6 \n\t" 703 704 #define MULADDC_CORE \ 705 "ldq $6, 0($1) \n\t" \ 706 "addq $1, 8, $1 \n\t" \ 707 "mulq $6, $4, $7 \n\t" \ 708 "umulh $6, $4, $6 \n\t" \ 709 "addq $7, $3, $7 \n\t" \ 710 "cmpult $7, $3, $3 \n\t" \ 711 "ldq $5, 0($2) \n\t" \ 712 "addq $7, $5, $7 \n\t" \ 713 "cmpult $7, $5, $5 \n\t" \ 714 "stq $7, 0($2) \n\t" \ 715 "addq $2, 8, $2 \n\t" \ 716 "addq $6, $3, $3 \n\t" \ 717 "addq $5, $3, $3 \n\t" 718 719 #define MULADDC_STOP \ 720 "stq $3, %0 \n\t" \ 721 "stq $2, %1 \n\t" \ 722 "stq $1, %2 \n\t" \ 723 : "=m" (c), "=m" (d), "=m" (s) \ 724 : "m" (s), "m" (d), "m" (c), "m" (b) \ 725 : "$1", "$2", "$3", "$4", "$5", "$6", "$7" \ 726 ); 727 #endif /* Alpha */ 728 729 #if defined(__mips__) && !defined(__mips64) 730 731 #define MULADDC_INIT \ 732 asm( \ 733 "lw $10, %3 \n\t" \ 734 "lw $11, %4 \n\t" \ 735 "lw $12, %5 \n\t" \ 736 "lw $13, %6 \n\t" 737 738 #define MULADDC_CORE \ 739 "lw $14, 0($10) \n\t" \ 740 "multu $13, $14 \n\t" \ 741 "addi $10, $10, 4 \n\t" \ 742 "mflo $14 \n\t" \ 743 "mfhi $9 \n\t" \ 744 "addu $14, $12, $14 \n\t" \ 745 "lw $15, 0($11) \n\t" \ 746 "sltu $12, $14, $12 \n\t" \ 747 "addu $15, $14, $15 \n\t" \ 748 "sltu $14, $15, $14 \n\t" \ 749 "addu $12, $12, $9 \n\t" \ 750 "sw $15, 0($11) \n\t" \ 751 "addu $12, $12, $14 \n\t" \ 752 "addi $11, $11, 4 \n\t" 753 754 #define MULADDC_STOP \ 755 "sw $12, %0 \n\t" \ 756 "sw $11, %1 \n\t" \ 757 "sw $10, %2 \n\t" \ 758 : "=m" (c), "=m" (d), "=m" (s) \ 759 : "m" (s), "m" (d), "m" (c), "m" (b) \ 760 : "$9", "$10", "$11", "$12", "$13", "$14", "$15", "lo", "hi" \ 761 ); 762 763 #endif /* MIPS */ 764 #endif /* GNUC */ 765 766 #if (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__) 767 768 #define MULADDC_INIT \ 769 __asm mov esi, s \ 770 __asm mov edi, d \ 771 __asm mov ecx, c \ 772 __asm mov ebx, b 773 774 #define MULADDC_CORE \ 775 __asm lodsd \ 776 __asm mul ebx \ 777 __asm add eax, ecx \ 778 __asm adc edx, 0 \ 779 __asm add eax, [edi] \ 780 __asm adc edx, 0 \ 781 __asm mov ecx, edx \ 782 __asm stosd 783 784 #if defined(MBEDTLS_HAVE_SSE2) 785 786 #define EMIT __asm _emit 787 788 #define MULADDC_HUIT \ 789 EMIT 0x0F EMIT 0x6E EMIT 0xC9 \ 790 EMIT 0x0F EMIT 0x6E EMIT 0xC3 \ 791 EMIT 0x0F EMIT 0x6E EMIT 0x1F \ 792 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \ 793 EMIT 0x0F EMIT 0x6E EMIT 0x16 \ 794 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \ 795 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x04 \ 796 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \ 797 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x08 \ 798 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \ 799 EMIT 0x0F EMIT 0x6E EMIT 0x7E EMIT 0x0C \ 800 EMIT 0x0F EMIT 0xF4 EMIT 0xF8 \ 801 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \ 802 EMIT 0x0F EMIT 0x6E EMIT 0x5F EMIT 0x04 \ 803 EMIT 0x0F EMIT 0xD4 EMIT 0xDC \ 804 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x08 \ 805 EMIT 0x0F EMIT 0xD4 EMIT 0xEE \ 806 EMIT 0x0F EMIT 0x6E EMIT 0x67 EMIT 0x0C \ 807 EMIT 0x0F EMIT 0xD4 EMIT 0xFC \ 808 EMIT 0x0F EMIT 0x7E EMIT 0x0F \ 809 EMIT 0x0F EMIT 0x6E EMIT 0x56 EMIT 0x10 \ 810 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \ 811 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 812 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x14 \ 813 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \ 814 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \ 815 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x18 \ 816 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \ 817 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x04 \ 818 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 819 EMIT 0x0F EMIT 0x6E EMIT 0x5E EMIT 0x1C \ 820 EMIT 0x0F EMIT 0xF4 EMIT 0xD8 \ 821 EMIT 0x0F EMIT 0xD4 EMIT 0xCD \ 822 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x10 \ 823 EMIT 0x0F EMIT 0xD4 EMIT 0xD5 \ 824 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x08 \ 825 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 826 EMIT 0x0F EMIT 0xD4 EMIT 0xCF \ 827 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x14 \ 828 EMIT 0x0F EMIT 0xD4 EMIT 0xE5 \ 829 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x0C \ 830 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 831 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \ 832 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x18 \ 833 EMIT 0x0F EMIT 0xD4 EMIT 0xF5 \ 834 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x10 \ 835 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 836 EMIT 0x0F EMIT 0xD4 EMIT 0xCC \ 837 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x1C \ 838 EMIT 0x0F EMIT 0xD4 EMIT 0xDD \ 839 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x14 \ 840 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 841 EMIT 0x0F EMIT 0xD4 EMIT 0xCE \ 842 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x18 \ 843 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 844 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \ 845 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x1C \ 846 EMIT 0x83 EMIT 0xC7 EMIT 0x20 \ 847 EMIT 0x83 EMIT 0xC6 EMIT 0x20 \ 848 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 849 EMIT 0x0F EMIT 0x7E EMIT 0xC9 850 851 #define MULADDC_STOP \ 852 EMIT 0x0F EMIT 0x77 \ 853 __asm mov c, ecx \ 854 __asm mov d, edi \ 855 __asm mov s, esi \ 856 857 #else 858 859 #define MULADDC_STOP \ 860 __asm mov c, ecx \ 861 __asm mov d, edi \ 862 __asm mov s, esi \ 863 864 #endif /* SSE2 */ 865 #endif /* MSVC */ 866 867 #endif /* MBEDTLS_HAVE_ASM */ 868 869 #if !defined(MULADDC_CORE) 870 #if defined(MBEDTLS_HAVE_UDBL) 871 872 #define MULADDC_INIT \ 873 { \ 874 mbedtls_t_udbl r; \ 875 mbedtls_mpi_uint r0, r1; 876 877 #define MULADDC_CORE \ 878 r = *(s++) * (mbedtls_t_udbl) b; \ 879 r0 = (mbedtls_mpi_uint) r; \ 880 r1 = (mbedtls_mpi_uint)( r >> biL ); \ 881 r0 += c; r1 += (r0 < c); \ 882 r0 += *d; r1 += (r0 < *d); \ 883 c = r1; *(d++) = r0; 884 885 #define MULADDC_STOP \ 886 } 887 888 #else 889 #define MULADDC_INIT \ 890 { \ 891 mbedtls_mpi_uint s0, s1, b0, b1; \ 892 mbedtls_mpi_uint r0, r1, rx, ry; \ 893 b0 = ( b << biH ) >> biH; \ 894 b1 = ( b >> biH ); 895 896 #define MULADDC_CORE \ 897 s0 = ( *s << biH ) >> biH; \ 898 s1 = ( *s >> biH ); s++; \ 899 rx = s0 * b1; r0 = s0 * b0; \ 900 ry = s1 * b0; r1 = s1 * b1; \ 901 r1 += ( rx >> biH ); \ 902 r1 += ( ry >> biH ); \ 903 rx <<= biH; ry <<= biH; \ 904 r0 += rx; r1 += (r0 < rx); \ 905 r0 += ry; r1 += (r0 < ry); \ 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 #endif /* C (generic) */ 914 #endif /* C (longlong) */ 915 916 #endif /* bn_mul.h */ 917