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 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 /* 23 * Multiply source vector [s] with b, add result 24 * to destination vector [d] and set carry c. 25 * 26 * Currently supports: 27 * 28 * . IA-32 (386+) . AMD64 / EM64T 29 * . IA-32 (SSE2) . Motorola 68000 30 * . PowerPC, 32-bit . MicroBlaze 31 * . PowerPC, 64-bit . TriCore 32 * . SPARC v8 . ARM v3+ 33 * . Alpha . MIPS32 34 * . C, longlong . C, generic 35 */ 36 #ifndef MBEDTLS_BN_MUL_H 37 #define MBEDTLS_BN_MUL_H 38 39 #if !defined(MBEDTLS_CONFIG_FILE) 40 #include "mbedtls/config.h" 41 #else 42 #include MBEDTLS_CONFIG_FILE 43 #endif 44 45 #include "mbedtls/bignum.h" 46 47 #if defined(MBEDTLS_HAVE_ASM) 48 49 #ifndef asm 50 #define asm __asm 51 #endif 52 53 /* armcc5 --gnu defines __GNUC__ but doesn't support GNU's extended asm */ 54 #if defined(__GNUC__) && \ 55 ( !defined(__ARMCC_VERSION) || __ARMCC_VERSION >= 6000000 ) 56 57 /* 58 * Disable use of the i386 assembly code below if option -O0, to disable all 59 * compiler optimisations, is passed, detected with __OPTIMIZE__ 60 * This is done as the number of registers used in the assembly code doesn't 61 * work with the -O0 option. 62 */ 63 #if defined(__i386__) && defined(__OPTIMIZE__) 64 65 #define MULADDC_INIT \ 66 asm( \ 67 "movl %%ebx, %0 \n\t" \ 68 "movl %5, %%esi \n\t" \ 69 "movl %6, %%edi \n\t" \ 70 "movl %7, %%ecx \n\t" \ 71 "movl %8, %%ebx \n\t" 72 73 #define MULADDC_CORE \ 74 "lodsl \n\t" \ 75 "mull %%ebx \n\t" \ 76 "addl %%ecx, %%eax \n\t" \ 77 "adcl $0, %%edx \n\t" \ 78 "addl (%%edi), %%eax \n\t" \ 79 "adcl $0, %%edx \n\t" \ 80 "movl %%edx, %%ecx \n\t" \ 81 "stosl \n\t" 82 83 #if defined(MBEDTLS_HAVE_SSE2) 84 85 #define MULADDC_HUIT \ 86 "movd %%ecx, %%mm1 \n\t" \ 87 "movd %%ebx, %%mm0 \n\t" \ 88 "movd (%%edi), %%mm3 \n\t" \ 89 "paddq %%mm3, %%mm1 \n\t" \ 90 "movd (%%esi), %%mm2 \n\t" \ 91 "pmuludq %%mm0, %%mm2 \n\t" \ 92 "movd 4(%%esi), %%mm4 \n\t" \ 93 "pmuludq %%mm0, %%mm4 \n\t" \ 94 "movd 8(%%esi), %%mm6 \n\t" \ 95 "pmuludq %%mm0, %%mm6 \n\t" \ 96 "movd 12(%%esi), %%mm7 \n\t" \ 97 "pmuludq %%mm0, %%mm7 \n\t" \ 98 "paddq %%mm2, %%mm1 \n\t" \ 99 "movd 4(%%edi), %%mm3 \n\t" \ 100 "paddq %%mm4, %%mm3 \n\t" \ 101 "movd 8(%%edi), %%mm5 \n\t" \ 102 "paddq %%mm6, %%mm5 \n\t" \ 103 "movd 12(%%edi), %%mm4 \n\t" \ 104 "paddq %%mm4, %%mm7 \n\t" \ 105 "movd %%mm1, (%%edi) \n\t" \ 106 "movd 16(%%esi), %%mm2 \n\t" \ 107 "pmuludq %%mm0, %%mm2 \n\t" \ 108 "psrlq $32, %%mm1 \n\t" \ 109 "movd 20(%%esi), %%mm4 \n\t" \ 110 "pmuludq %%mm0, %%mm4 \n\t" \ 111 "paddq %%mm3, %%mm1 \n\t" \ 112 "movd 24(%%esi), %%mm6 \n\t" \ 113 "pmuludq %%mm0, %%mm6 \n\t" \ 114 "movd %%mm1, 4(%%edi) \n\t" \ 115 "psrlq $32, %%mm1 \n\t" \ 116 "movd 28(%%esi), %%mm3 \n\t" \ 117 "pmuludq %%mm0, %%mm3 \n\t" \ 118 "paddq %%mm5, %%mm1 \n\t" \ 119 "movd 16(%%edi), %%mm5 \n\t" \ 120 "paddq %%mm5, %%mm2 \n\t" \ 121 "movd %%mm1, 8(%%edi) \n\t" \ 122 "psrlq $32, %%mm1 \n\t" \ 123 "paddq %%mm7, %%mm1 \n\t" \ 124 "movd 20(%%edi), %%mm5 \n\t" \ 125 "paddq %%mm5, %%mm4 \n\t" \ 126 "movd %%mm1, 12(%%edi) \n\t" \ 127 "psrlq $32, %%mm1 \n\t" \ 128 "paddq %%mm2, %%mm1 \n\t" \ 129 "movd 24(%%edi), %%mm5 \n\t" \ 130 "paddq %%mm5, %%mm6 \n\t" \ 131 "movd %%mm1, 16(%%edi) \n\t" \ 132 "psrlq $32, %%mm1 \n\t" \ 133 "paddq %%mm4, %%mm1 \n\t" \ 134 "movd 28(%%edi), %%mm5 \n\t" \ 135 "paddq %%mm5, %%mm3 \n\t" \ 136 "movd %%mm1, 20(%%edi) \n\t" \ 137 "psrlq $32, %%mm1 \n\t" \ 138 "paddq %%mm6, %%mm1 \n\t" \ 139 "movd %%mm1, 24(%%edi) \n\t" \ 140 "psrlq $32, %%mm1 \n\t" \ 141 "paddq %%mm3, %%mm1 \n\t" \ 142 "movd %%mm1, 28(%%edi) \n\t" \ 143 "addl $32, %%edi \n\t" \ 144 "addl $32, %%esi \n\t" \ 145 "psrlq $32, %%mm1 \n\t" \ 146 "movd %%mm1, %%ecx \n\t" 147 148 #define MULADDC_STOP \ 149 "emms \n\t" \ 150 "movl %4, %%ebx \n\t" \ 151 "movl %%ecx, %1 \n\t" \ 152 "movl %%edi, %2 \n\t" \ 153 "movl %%esi, %3 \n\t" \ 154 : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \ 155 : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \ 156 : "eax", "ebx", "ecx", "edx", "esi", "edi" \ 157 ); 158 159 #else 160 161 #define MULADDC_STOP \ 162 "movl %4, %%ebx \n\t" \ 163 "movl %%ecx, %1 \n\t" \ 164 "movl %%edi, %2 \n\t" \ 165 "movl %%esi, %3 \n\t" \ 166 : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \ 167 : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \ 168 : "eax", "ebx", "ecx", "edx", "esi", "edi" \ 169 ); 170 #endif /* SSE2 */ 171 #endif /* i386 */ 172 173 #if defined(__amd64__) || defined (__x86_64__) 174 175 #define MULADDC_INIT \ 176 asm( \ 177 "xorq %%r8, %%r8\n" 178 179 #define MULADDC_CORE \ 180 "movq (%%rsi), %%rax\n" \ 181 "mulq %%rbx\n" \ 182 "addq $8, %%rsi\n" \ 183 "addq %%rcx, %%rax\n" \ 184 "movq %%r8, %%rcx\n" \ 185 "adcq $0, %%rdx\n" \ 186 "nop \n" \ 187 "addq %%rax, (%%rdi)\n" \ 188 "adcq %%rdx, %%rcx\n" \ 189 "addq $8, %%rdi\n" 190 191 #define MULADDC_STOP \ 192 : "+c" (c), "+D" (d), "+S" (s) \ 193 : "b" (b) \ 194 : "rax", "rdx", "r8" \ 195 ); 196 197 #endif /* AMD64 */ 198 199 #if defined(__aarch64__) 200 201 #define MULADDC_INIT \ 202 asm( 203 204 #define MULADDC_CORE \ 205 "ldr x4, [%2], #8 \n\t" \ 206 "ldr x5, [%1] \n\t" \ 207 "mul x6, x4, %3 \n\t" \ 208 "umulh x7, x4, %3 \n\t" \ 209 "adds x5, x5, x6 \n\t" \ 210 "adc x7, x7, xzr \n\t" \ 211 "adds x5, x5, %0 \n\t" \ 212 "adc %0, x7, xzr \n\t" \ 213 "str x5, [%1], #8 \n\t" 214 215 #define MULADDC_STOP \ 216 : "+r" (c), "+r" (d), "+r" (s) \ 217 : "r" (b) \ 218 : "x4", "x5", "x6", "x7", "cc" \ 219 ); 220 221 #endif /* Aarch64 */ 222 223 #if defined(__mc68020__) || defined(__mcpu32__) 224 225 #define MULADDC_INIT \ 226 asm( \ 227 "movl %3, %%a2 \n\t" \ 228 "movl %4, %%a3 \n\t" \ 229 "movl %5, %%d3 \n\t" \ 230 "movl %6, %%d2 \n\t" \ 231 "moveq #0, %%d0 \n\t" 232 233 #define MULADDC_CORE \ 234 "movel %%a2@+, %%d1 \n\t" \ 235 "mulul %%d2, %%d4:%%d1 \n\t" \ 236 "addl %%d3, %%d1 \n\t" \ 237 "addxl %%d0, %%d4 \n\t" \ 238 "moveq #0, %%d3 \n\t" \ 239 "addl %%d1, %%a3@+ \n\t" \ 240 "addxl %%d4, %%d3 \n\t" 241 242 #define MULADDC_STOP \ 243 "movl %%d3, %0 \n\t" \ 244 "movl %%a3, %1 \n\t" \ 245 "movl %%a2, %2 \n\t" \ 246 : "=m" (c), "=m" (d), "=m" (s) \ 247 : "m" (s), "m" (d), "m" (c), "m" (b) \ 248 : "d0", "d1", "d2", "d3", "d4", "a2", "a3" \ 249 ); 250 251 #define MULADDC_HUIT \ 252 "movel %%a2@+, %%d1 \n\t" \ 253 "mulul %%d2, %%d4:%%d1 \n\t" \ 254 "addxl %%d3, %%d1 \n\t" \ 255 "addxl %%d0, %%d4 \n\t" \ 256 "addl %%d1, %%a3@+ \n\t" \ 257 "movel %%a2@+, %%d1 \n\t" \ 258 "mulul %%d2, %%d3:%%d1 \n\t" \ 259 "addxl %%d4, %%d1 \n\t" \ 260 "addxl %%d0, %%d3 \n\t" \ 261 "addl %%d1, %%a3@+ \n\t" \ 262 "movel %%a2@+, %%d1 \n\t" \ 263 "mulul %%d2, %%d4:%%d1 \n\t" \ 264 "addxl %%d3, %%d1 \n\t" \ 265 "addxl %%d0, %%d4 \n\t" \ 266 "addl %%d1, %%a3@+ \n\t" \ 267 "movel %%a2@+, %%d1 \n\t" \ 268 "mulul %%d2, %%d3:%%d1 \n\t" \ 269 "addxl %%d4, %%d1 \n\t" \ 270 "addxl %%d0, %%d3 \n\t" \ 271 "addl %%d1, %%a3@+ \n\t" \ 272 "movel %%a2@+, %%d1 \n\t" \ 273 "mulul %%d2, %%d4:%%d1 \n\t" \ 274 "addxl %%d3, %%d1 \n\t" \ 275 "addxl %%d0, %%d4 \n\t" \ 276 "addl %%d1, %%a3@+ \n\t" \ 277 "movel %%a2@+, %%d1 \n\t" \ 278 "mulul %%d2, %%d3:%%d1 \n\t" \ 279 "addxl %%d4, %%d1 \n\t" \ 280 "addxl %%d0, %%d3 \n\t" \ 281 "addl %%d1, %%a3@+ \n\t" \ 282 "movel %%a2@+, %%d1 \n\t" \ 283 "mulul %%d2, %%d4:%%d1 \n\t" \ 284 "addxl %%d3, %%d1 \n\t" \ 285 "addxl %%d0, %%d4 \n\t" \ 286 "addl %%d1, %%a3@+ \n\t" \ 287 "movel %%a2@+, %%d1 \n\t" \ 288 "mulul %%d2, %%d3:%%d1 \n\t" \ 289 "addxl %%d4, %%d1 \n\t" \ 290 "addxl %%d0, %%d3 \n\t" \ 291 "addl %%d1, %%a3@+ \n\t" \ 292 "addxl %%d0, %%d3 \n\t" 293 294 #endif /* MC68000 */ 295 296 #if defined(__powerpc64__) || defined(__ppc64__) 297 298 #if defined(__MACH__) && defined(__APPLE__) 299 300 #define MULADDC_INIT \ 301 asm( \ 302 "ld r3, %3 \n\t" \ 303 "ld r4, %4 \n\t" \ 304 "ld r5, %5 \n\t" \ 305 "ld r6, %6 \n\t" \ 306 "addi r3, r3, -8 \n\t" \ 307 "addi r4, r4, -8 \n\t" \ 308 "addic r5, r5, 0 \n\t" 309 310 #define MULADDC_CORE \ 311 "ldu r7, 8(r3) \n\t" \ 312 "mulld r8, r7, r6 \n\t" \ 313 "mulhdu r9, r7, r6 \n\t" \ 314 "adde r8, r8, r5 \n\t" \ 315 "ld r7, 8(r4) \n\t" \ 316 "addze r5, r9 \n\t" \ 317 "addc r8, r8, r7 \n\t" \ 318 "stdu r8, 8(r4) \n\t" 319 320 #define MULADDC_STOP \ 321 "addze r5, r5 \n\t" \ 322 "addi r4, r4, 8 \n\t" \ 323 "addi r3, r3, 8 \n\t" \ 324 "std r5, %0 \n\t" \ 325 "std r4, %1 \n\t" \ 326 "std r3, %2 \n\t" \ 327 : "=m" (c), "=m" (d), "=m" (s) \ 328 : "m" (s), "m" (d), "m" (c), "m" (b) \ 329 : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \ 330 ); 331 332 333 #else /* __MACH__ && __APPLE__ */ 334 335 #define MULADDC_INIT \ 336 asm( \ 337 "ld %%r3, %3 \n\t" \ 338 "ld %%r4, %4 \n\t" \ 339 "ld %%r5, %5 \n\t" \ 340 "ld %%r6, %6 \n\t" \ 341 "addi %%r3, %%r3, -8 \n\t" \ 342 "addi %%r4, %%r4, -8 \n\t" \ 343 "addic %%r5, %%r5, 0 \n\t" 344 345 #define MULADDC_CORE \ 346 "ldu %%r7, 8(%%r3) \n\t" \ 347 "mulld %%r8, %%r7, %%r6 \n\t" \ 348 "mulhdu %%r9, %%r7, %%r6 \n\t" \ 349 "adde %%r8, %%r8, %%r5 \n\t" \ 350 "ld %%r7, 8(%%r4) \n\t" \ 351 "addze %%r5, %%r9 \n\t" \ 352 "addc %%r8, %%r8, %%r7 \n\t" \ 353 "stdu %%r8, 8(%%r4) \n\t" 354 355 #define MULADDC_STOP \ 356 "addze %%r5, %%r5 \n\t" \ 357 "addi %%r4, %%r4, 8 \n\t" \ 358 "addi %%r3, %%r3, 8 \n\t" \ 359 "std %%r5, %0 \n\t" \ 360 "std %%r4, %1 \n\t" \ 361 "std %%r3, %2 \n\t" \ 362 : "=m" (c), "=m" (d), "=m" (s) \ 363 : "m" (s), "m" (d), "m" (c), "m" (b) \ 364 : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \ 365 ); 366 367 #endif /* __MACH__ && __APPLE__ */ 368 369 #elif defined(__powerpc__) || defined(__ppc__) /* end PPC64/begin PPC32 */ 370 371 #if defined(__MACH__) && defined(__APPLE__) 372 373 #define MULADDC_INIT \ 374 asm( \ 375 "lwz r3, %3 \n\t" \ 376 "lwz r4, %4 \n\t" \ 377 "lwz r5, %5 \n\t" \ 378 "lwz r6, %6 \n\t" \ 379 "addi r3, r3, -4 \n\t" \ 380 "addi r4, r4, -4 \n\t" \ 381 "addic r5, r5, 0 \n\t" 382 383 #define MULADDC_CORE \ 384 "lwzu r7, 4(r3) \n\t" \ 385 "mullw r8, r7, r6 \n\t" \ 386 "mulhwu r9, r7, r6 \n\t" \ 387 "adde r8, r8, r5 \n\t" \ 388 "lwz r7, 4(r4) \n\t" \ 389 "addze r5, r9 \n\t" \ 390 "addc r8, r8, r7 \n\t" \ 391 "stwu r8, 4(r4) \n\t" 392 393 #define MULADDC_STOP \ 394 "addze r5, r5 \n\t" \ 395 "addi r4, r4, 4 \n\t" \ 396 "addi r3, r3, 4 \n\t" \ 397 "stw r5, %0 \n\t" \ 398 "stw r4, %1 \n\t" \ 399 "stw r3, %2 \n\t" \ 400 : "=m" (c), "=m" (d), "=m" (s) \ 401 : "m" (s), "m" (d), "m" (c), "m" (b) \ 402 : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \ 403 ); 404 405 #else /* __MACH__ && __APPLE__ */ 406 407 #define MULADDC_INIT \ 408 asm( \ 409 "lwz %%r3, %3 \n\t" \ 410 "lwz %%r4, %4 \n\t" \ 411 "lwz %%r5, %5 \n\t" \ 412 "lwz %%r6, %6 \n\t" \ 413 "addi %%r3, %%r3, -4 \n\t" \ 414 "addi %%r4, %%r4, -4 \n\t" \ 415 "addic %%r5, %%r5, 0 \n\t" 416 417 #define MULADDC_CORE \ 418 "lwzu %%r7, 4(%%r3) \n\t" \ 419 "mullw %%r8, %%r7, %%r6 \n\t" \ 420 "mulhwu %%r9, %%r7, %%r6 \n\t" \ 421 "adde %%r8, %%r8, %%r5 \n\t" \ 422 "lwz %%r7, 4(%%r4) \n\t" \ 423 "addze %%r5, %%r9 \n\t" \ 424 "addc %%r8, %%r8, %%r7 \n\t" \ 425 "stwu %%r8, 4(%%r4) \n\t" 426 427 #define MULADDC_STOP \ 428 "addze %%r5, %%r5 \n\t" \ 429 "addi %%r4, %%r4, 4 \n\t" \ 430 "addi %%r3, %%r3, 4 \n\t" \ 431 "stw %%r5, %0 \n\t" \ 432 "stw %%r4, %1 \n\t" \ 433 "stw %%r3, %2 \n\t" \ 434 : "=m" (c), "=m" (d), "=m" (s) \ 435 : "m" (s), "m" (d), "m" (c), "m" (b) \ 436 : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \ 437 ); 438 439 #endif /* __MACH__ && __APPLE__ */ 440 441 #endif /* PPC32 */ 442 443 /* 444 * The Sparc(64) assembly is reported to be broken. 445 * Disable it for now, until we're able to fix it. 446 */ 447 #if 0 && defined(__sparc__) 448 #if defined(__sparc64__) 449 450 #define MULADDC_INIT \ 451 asm( \ 452 "ldx %3, %%o0 \n\t" \ 453 "ldx %4, %%o1 \n\t" \ 454 "ld %5, %%o2 \n\t" \ 455 "ld %6, %%o3 \n\t" 456 457 #define MULADDC_CORE \ 458 "ld [%%o0], %%o4 \n\t" \ 459 "inc 4, %%o0 \n\t" \ 460 "ld [%%o1], %%o5 \n\t" \ 461 "umul %%o3, %%o4, %%o4 \n\t" \ 462 "addcc %%o4, %%o2, %%o4 \n\t" \ 463 "rd %%y, %%g1 \n\t" \ 464 "addx %%g1, 0, %%g1 \n\t" \ 465 "addcc %%o4, %%o5, %%o4 \n\t" \ 466 "st %%o4, [%%o1] \n\t" \ 467 "addx %%g1, 0, %%o2 \n\t" \ 468 "inc 4, %%o1 \n\t" 469 470 #define MULADDC_STOP \ 471 "st %%o2, %0 \n\t" \ 472 "stx %%o1, %1 \n\t" \ 473 "stx %%o0, %2 \n\t" \ 474 : "=m" (c), "=m" (d), "=m" (s) \ 475 : "m" (s), "m" (d), "m" (c), "m" (b) \ 476 : "g1", "o0", "o1", "o2", "o3", "o4", \ 477 "o5" \ 478 ); 479 480 #else /* __sparc64__ */ 481 482 #define MULADDC_INIT \ 483 asm( \ 484 "ld %3, %%o0 \n\t" \ 485 "ld %4, %%o1 \n\t" \ 486 "ld %5, %%o2 \n\t" \ 487 "ld %6, %%o3 \n\t" 488 489 #define MULADDC_CORE \ 490 "ld [%%o0], %%o4 \n\t" \ 491 "inc 4, %%o0 \n\t" \ 492 "ld [%%o1], %%o5 \n\t" \ 493 "umul %%o3, %%o4, %%o4 \n\t" \ 494 "addcc %%o4, %%o2, %%o4 \n\t" \ 495 "rd %%y, %%g1 \n\t" \ 496 "addx %%g1, 0, %%g1 \n\t" \ 497 "addcc %%o4, %%o5, %%o4 \n\t" \ 498 "st %%o4, [%%o1] \n\t" \ 499 "addx %%g1, 0, %%o2 \n\t" \ 500 "inc 4, %%o1 \n\t" 501 502 #define MULADDC_STOP \ 503 "st %%o2, %0 \n\t" \ 504 "st %%o1, %1 \n\t" \ 505 "st %%o0, %2 \n\t" \ 506 : "=m" (c), "=m" (d), "=m" (s) \ 507 : "m" (s), "m" (d), "m" (c), "m" (b) \ 508 : "g1", "o0", "o1", "o2", "o3", "o4", \ 509 "o5" \ 510 ); 511 512 #endif /* __sparc64__ */ 513 #endif /* __sparc__ */ 514 515 #if defined(__microblaze__) || defined(microblaze) 516 517 #define MULADDC_INIT \ 518 asm( \ 519 "lwi r3, %3 \n\t" \ 520 "lwi r4, %4 \n\t" \ 521 "lwi r5, %5 \n\t" \ 522 "lwi r6, %6 \n\t" \ 523 "andi r7, r6, 0xffff \n\t" \ 524 "bsrli r6, r6, 16 \n\t" 525 526 #define MULADDC_CORE \ 527 "lhui r8, r3, 0 \n\t" \ 528 "addi r3, r3, 2 \n\t" \ 529 "lhui r9, r3, 0 \n\t" \ 530 "addi r3, r3, 2 \n\t" \ 531 "mul r10, r9, r6 \n\t" \ 532 "mul r11, r8, r7 \n\t" \ 533 "mul r12, r9, r7 \n\t" \ 534 "mul r13, r8, r6 \n\t" \ 535 "bsrli r8, r10, 16 \n\t" \ 536 "bsrli r9, r11, 16 \n\t" \ 537 "add r13, r13, r8 \n\t" \ 538 "add r13, r13, r9 \n\t" \ 539 "bslli r10, r10, 16 \n\t" \ 540 "bslli r11, r11, 16 \n\t" \ 541 "add r12, r12, r10 \n\t" \ 542 "addc r13, r13, r0 \n\t" \ 543 "add r12, r12, r11 \n\t" \ 544 "addc r13, r13, r0 \n\t" \ 545 "lwi r10, r4, 0 \n\t" \ 546 "add r12, r12, r10 \n\t" \ 547 "addc r13, r13, r0 \n\t" \ 548 "add r12, r12, r5 \n\t" \ 549 "addc r5, r13, r0 \n\t" \ 550 "swi r12, r4, 0 \n\t" \ 551 "addi r4, r4, 4 \n\t" 552 553 #define MULADDC_STOP \ 554 "swi r5, %0 \n\t" \ 555 "swi r4, %1 \n\t" \ 556 "swi r3, %2 \n\t" \ 557 : "=m" (c), "=m" (d), "=m" (s) \ 558 : "m" (s), "m" (d), "m" (c), "m" (b) \ 559 : "r3", "r4", "r5", "r6", "r7", "r8", \ 560 "r9", "r10", "r11", "r12", "r13" \ 561 ); 562 563 #endif /* MicroBlaze */ 564 565 #if defined(__tricore__) 566 567 #define MULADDC_INIT \ 568 asm( \ 569 "ld.a %%a2, %3 \n\t" \ 570 "ld.a %%a3, %4 \n\t" \ 571 "ld.w %%d4, %5 \n\t" \ 572 "ld.w %%d1, %6 \n\t" \ 573 "xor %%d5, %%d5 \n\t" 574 575 #define MULADDC_CORE \ 576 "ld.w %%d0, [%%a2+] \n\t" \ 577 "madd.u %%e2, %%e4, %%d0, %%d1 \n\t" \ 578 "ld.w %%d0, [%%a3] \n\t" \ 579 "addx %%d2, %%d2, %%d0 \n\t" \ 580 "addc %%d3, %%d3, 0 \n\t" \ 581 "mov %%d4, %%d3 \n\t" \ 582 "st.w [%%a3+], %%d2 \n\t" 583 584 #define MULADDC_STOP \ 585 "st.w %0, %%d4 \n\t" \ 586 "st.a %1, %%a3 \n\t" \ 587 "st.a %2, %%a2 \n\t" \ 588 : "=m" (c), "=m" (d), "=m" (s) \ 589 : "m" (s), "m" (d), "m" (c), "m" (b) \ 590 : "d0", "d1", "e2", "d4", "a2", "a3" \ 591 ); 592 593 #endif /* TriCore */ 594 595 /* 596 * Note, gcc -O0 by default uses r7 for the frame pointer, so it complains about 597 * our use of r7 below, unless -fomit-frame-pointer is passed. 598 * 599 * On the other hand, -fomit-frame-pointer is implied by any -Ox options with 600 * x !=0, which we can detect using __OPTIMIZE__ (which is also defined by 601 * clang and armcc5 under the same conditions). 602 * 603 * So, only use the optimized assembly below for optimized build, which avoids 604 * the build error and is pretty reasonable anyway. 605 */ 606 #if defined(__GNUC__) && !defined(__OPTIMIZE__) 607 #define MULADDC_CANNOT_USE_R7 608 #endif 609 610 #if defined(__arm__) && !defined(MULADDC_CANNOT_USE_R7) 611 612 #if defined(__thumb__) && !defined(__thumb2__) 613 614 #define MULADDC_INIT \ 615 asm( \ 616 "ldr r0, %3 \n\t" \ 617 "ldr r1, %4 \n\t" \ 618 "ldr r2, %5 \n\t" \ 619 "ldr r3, %6 \n\t" \ 620 "lsr r7, r3, #16 \n\t" \ 621 "mov r9, r7 \n\t" \ 622 "lsl r7, r3, #16 \n\t" \ 623 "lsr r7, r7, #16 \n\t" \ 624 "mov r8, r7 \n\t" 625 626 #define MULADDC_CORE \ 627 "ldmia r0!, {r6} \n\t" \ 628 "lsr r7, r6, #16 \n\t" \ 629 "lsl r6, r6, #16 \n\t" \ 630 "lsr r6, r6, #16 \n\t" \ 631 "mov r4, r8 \n\t" \ 632 "mul r4, r6 \n\t" \ 633 "mov r3, r9 \n\t" \ 634 "mul r6, r3 \n\t" \ 635 "mov r5, r9 \n\t" \ 636 "mul r5, r7 \n\t" \ 637 "mov r3, r8 \n\t" \ 638 "mul r7, r3 \n\t" \ 639 "lsr r3, r6, #16 \n\t" \ 640 "add r5, r5, r3 \n\t" \ 641 "lsr r3, r7, #16 \n\t" \ 642 "add r5, r5, r3 \n\t" \ 643 "add r4, r4, r2 \n\t" \ 644 "mov r2, #0 \n\t" \ 645 "adc r5, r2 \n\t" \ 646 "lsl r3, r6, #16 \n\t" \ 647 "add r4, r4, r3 \n\t" \ 648 "adc r5, r2 \n\t" \ 649 "lsl r3, r7, #16 \n\t" \ 650 "add r4, r4, r3 \n\t" \ 651 "adc r5, r2 \n\t" \ 652 "ldr r3, [r1] \n\t" \ 653 "add r4, r4, r3 \n\t" \ 654 "adc r2, r5 \n\t" \ 655 "stmia r1!, {r4} \n\t" 656 657 #define MULADDC_STOP \ 658 "str r2, %0 \n\t" \ 659 "str r1, %1 \n\t" \ 660 "str r0, %2 \n\t" \ 661 : "=m" (c), "=m" (d), "=m" (s) \ 662 : "m" (s), "m" (d), "m" (c), "m" (b) \ 663 : "r0", "r1", "r2", "r3", "r4", "r5", \ 664 "r6", "r7", "r8", "r9", "cc" \ 665 ); 666 667 #elif (__ARM_ARCH >= 6) && \ 668 defined (__ARM_FEATURE_DSP) && (__ARM_FEATURE_DSP == 1) 669 670 #define MULADDC_INIT \ 671 asm( 672 673 #define MULADDC_CORE \ 674 "ldr r0, [%0], #4 \n\t" \ 675 "ldr r1, [%1] \n\t" \ 676 "umaal r1, %2, %3, r0 \n\t" \ 677 "str r1, [%1], #4 \n\t" 678 679 #define MULADDC_STOP \ 680 : "=r" (s), "=r" (d), "=r" (c) \ 681 : "r" (b), "0" (s), "1" (d), "2" (c) \ 682 : "r0", "r1", "memory" \ 683 ); 684 685 #else 686 687 #define MULADDC_INIT \ 688 asm( \ 689 "ldr r0, %3 \n\t" \ 690 "ldr r1, %4 \n\t" \ 691 "ldr r2, %5 \n\t" \ 692 "ldr r3, %6 \n\t" 693 694 #define MULADDC_CORE \ 695 "ldr r4, [r0], #4 \n\t" \ 696 "mov r5, #0 \n\t" \ 697 "ldr r6, [r1] \n\t" \ 698 "umlal r2, r5, r3, r4 \n\t" \ 699 "adds r7, r6, r2 \n\t" \ 700 "adc r2, r5, #0 \n\t" \ 701 "str r7, [r1], #4 \n\t" 702 703 #define MULADDC_STOP \ 704 "str r2, %0 \n\t" \ 705 "str r1, %1 \n\t" \ 706 "str r0, %2 \n\t" \ 707 : "=m" (c), "=m" (d), "=m" (s) \ 708 : "m" (s), "m" (d), "m" (c), "m" (b) \ 709 : "r0", "r1", "r2", "r3", "r4", "r5", \ 710 "r6", "r7", "cc" \ 711 ); 712 713 #endif /* Thumb */ 714 715 #endif /* ARMv3 */ 716 717 #if defined(__alpha__) 718 719 #define MULADDC_INIT \ 720 asm( \ 721 "ldq $1, %3 \n\t" \ 722 "ldq $2, %4 \n\t" \ 723 "ldq $3, %5 \n\t" \ 724 "ldq $4, %6 \n\t" 725 726 #define MULADDC_CORE \ 727 "ldq $6, 0($1) \n\t" \ 728 "addq $1, 8, $1 \n\t" \ 729 "mulq $6, $4, $7 \n\t" \ 730 "umulh $6, $4, $6 \n\t" \ 731 "addq $7, $3, $7 \n\t" \ 732 "cmpult $7, $3, $3 \n\t" \ 733 "ldq $5, 0($2) \n\t" \ 734 "addq $7, $5, $7 \n\t" \ 735 "cmpult $7, $5, $5 \n\t" \ 736 "stq $7, 0($2) \n\t" \ 737 "addq $2, 8, $2 \n\t" \ 738 "addq $6, $3, $3 \n\t" \ 739 "addq $5, $3, $3 \n\t" 740 741 #define MULADDC_STOP \ 742 "stq $3, %0 \n\t" \ 743 "stq $2, %1 \n\t" \ 744 "stq $1, %2 \n\t" \ 745 : "=m" (c), "=m" (d), "=m" (s) \ 746 : "m" (s), "m" (d), "m" (c), "m" (b) \ 747 : "$1", "$2", "$3", "$4", "$5", "$6", "$7" \ 748 ); 749 #endif /* Alpha */ 750 751 #if defined(__mips__) && !defined(__mips64) 752 753 #define MULADDC_INIT \ 754 asm( \ 755 "lw $10, %3 \n\t" \ 756 "lw $11, %4 \n\t" \ 757 "lw $12, %5 \n\t" \ 758 "lw $13, %6 \n\t" 759 760 #define MULADDC_CORE \ 761 "lw $14, 0($10) \n\t" \ 762 "multu $13, $14 \n\t" \ 763 "addi $10, $10, 4 \n\t" \ 764 "mflo $14 \n\t" \ 765 "mfhi $9 \n\t" \ 766 "addu $14, $12, $14 \n\t" \ 767 "lw $15, 0($11) \n\t" \ 768 "sltu $12, $14, $12 \n\t" \ 769 "addu $15, $14, $15 \n\t" \ 770 "sltu $14, $15, $14 \n\t" \ 771 "addu $12, $12, $9 \n\t" \ 772 "sw $15, 0($11) \n\t" \ 773 "addu $12, $12, $14 \n\t" \ 774 "addi $11, $11, 4 \n\t" 775 776 #define MULADDC_STOP \ 777 "sw $12, %0 \n\t" \ 778 "sw $11, %1 \n\t" \ 779 "sw $10, %2 \n\t" \ 780 : "=m" (c), "=m" (d), "=m" (s) \ 781 : "m" (s), "m" (d), "m" (c), "m" (b) \ 782 : "$9", "$10", "$11", "$12", "$13", "$14", "$15", "lo", "hi" \ 783 ); 784 785 #endif /* MIPS */ 786 #endif /* GNUC */ 787 788 #if (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__) 789 790 #define MULADDC_INIT \ 791 __asm mov esi, s \ 792 __asm mov edi, d \ 793 __asm mov ecx, c \ 794 __asm mov ebx, b 795 796 #define MULADDC_CORE \ 797 __asm lodsd \ 798 __asm mul ebx \ 799 __asm add eax, ecx \ 800 __asm adc edx, 0 \ 801 __asm add eax, [edi] \ 802 __asm adc edx, 0 \ 803 __asm mov ecx, edx \ 804 __asm stosd 805 806 #if defined(MBEDTLS_HAVE_SSE2) 807 808 #define EMIT __asm _emit 809 810 #define MULADDC_HUIT \ 811 EMIT 0x0F EMIT 0x6E EMIT 0xC9 \ 812 EMIT 0x0F EMIT 0x6E EMIT 0xC3 \ 813 EMIT 0x0F EMIT 0x6E EMIT 0x1F \ 814 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \ 815 EMIT 0x0F EMIT 0x6E EMIT 0x16 \ 816 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \ 817 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x04 \ 818 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \ 819 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x08 \ 820 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \ 821 EMIT 0x0F EMIT 0x6E EMIT 0x7E EMIT 0x0C \ 822 EMIT 0x0F EMIT 0xF4 EMIT 0xF8 \ 823 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \ 824 EMIT 0x0F EMIT 0x6E EMIT 0x5F EMIT 0x04 \ 825 EMIT 0x0F EMIT 0xD4 EMIT 0xDC \ 826 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x08 \ 827 EMIT 0x0F EMIT 0xD4 EMIT 0xEE \ 828 EMIT 0x0F EMIT 0x6E EMIT 0x67 EMIT 0x0C \ 829 EMIT 0x0F EMIT 0xD4 EMIT 0xFC \ 830 EMIT 0x0F EMIT 0x7E EMIT 0x0F \ 831 EMIT 0x0F EMIT 0x6E EMIT 0x56 EMIT 0x10 \ 832 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \ 833 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 834 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x14 \ 835 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \ 836 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \ 837 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x18 \ 838 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \ 839 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x04 \ 840 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 841 EMIT 0x0F EMIT 0x6E EMIT 0x5E EMIT 0x1C \ 842 EMIT 0x0F EMIT 0xF4 EMIT 0xD8 \ 843 EMIT 0x0F EMIT 0xD4 EMIT 0xCD \ 844 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x10 \ 845 EMIT 0x0F EMIT 0xD4 EMIT 0xD5 \ 846 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x08 \ 847 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 848 EMIT 0x0F EMIT 0xD4 EMIT 0xCF \ 849 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x14 \ 850 EMIT 0x0F EMIT 0xD4 EMIT 0xE5 \ 851 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x0C \ 852 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 853 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \ 854 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x18 \ 855 EMIT 0x0F EMIT 0xD4 EMIT 0xF5 \ 856 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x10 \ 857 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 858 EMIT 0x0F EMIT 0xD4 EMIT 0xCC \ 859 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x1C \ 860 EMIT 0x0F EMIT 0xD4 EMIT 0xDD \ 861 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x14 \ 862 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 863 EMIT 0x0F EMIT 0xD4 EMIT 0xCE \ 864 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x18 \ 865 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 866 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \ 867 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x1C \ 868 EMIT 0x83 EMIT 0xC7 EMIT 0x20 \ 869 EMIT 0x83 EMIT 0xC6 EMIT 0x20 \ 870 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \ 871 EMIT 0x0F EMIT 0x7E EMIT 0xC9 872 873 #define MULADDC_STOP \ 874 EMIT 0x0F EMIT 0x77 \ 875 __asm mov c, ecx \ 876 __asm mov d, edi \ 877 __asm mov s, esi \ 878 879 #else 880 881 #define MULADDC_STOP \ 882 __asm mov c, ecx \ 883 __asm mov d, edi \ 884 __asm mov s, esi \ 885 886 #endif /* SSE2 */ 887 #endif /* MSVC */ 888 889 #endif /* MBEDTLS_HAVE_ASM */ 890 891 #if !defined(MULADDC_CORE) 892 #if defined(MBEDTLS_HAVE_UDBL) 893 894 #define MULADDC_INIT \ 895 { \ 896 mbedtls_t_udbl r; \ 897 mbedtls_mpi_uint r0, r1; 898 899 #define MULADDC_CORE \ 900 r = *(s++) * (mbedtls_t_udbl) b; \ 901 r0 = (mbedtls_mpi_uint) r; \ 902 r1 = (mbedtls_mpi_uint)( r >> biL ); \ 903 r0 += c; r1 += (r0 < c); \ 904 r0 += *d; r1 += (r0 < *d); \ 905 c = r1; *(d++) = r0; 906 907 #define MULADDC_STOP \ 908 } 909 910 #else 911 #define MULADDC_INIT \ 912 { \ 913 mbedtls_mpi_uint s0, s1, b0, b1; \ 914 mbedtls_mpi_uint r0, r1, rx, ry; \ 915 b0 = ( b << biH ) >> biH; \ 916 b1 = ( b >> biH ); 917 918 #define MULADDC_CORE \ 919 s0 = ( *s << biH ) >> biH; \ 920 s1 = ( *s >> biH ); s++; \ 921 rx = s0 * b1; r0 = s0 * b0; \ 922 ry = s1 * b0; r1 = s1 * b1; \ 923 r1 += ( rx >> biH ); \ 924 r1 += ( ry >> biH ); \ 925 rx <<= biH; ry <<= biH; \ 926 r0 += rx; r1 += (r0 < rx); \ 927 r0 += ry; r1 += (r0 < ry); \ 928 r0 += c; r1 += (r0 < c); \ 929 r0 += *d; r1 += (r0 < *d); \ 930 c = r1; *(d++) = r0; 931 932 #define MULADDC_STOP \ 933 } 934 935 #endif /* C (generic) */ 936 #endif /* C (longlong) */ 937 938 #endif /* bn_mul.h */ 939