4 * \brief Multi-precision integer library
7 * Copyright (C) 2006-2015, ARM Limited, All Rights Reserved
8 * SPDX-License-Identifier: Apache-2.0
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
14 * http://www.apache.org/licenses/LICENSE-2.0
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.
22 * This file is part of mbed TLS (https://tls.mbed.org)
25 * Multiply source vector [s] with b, add result
26 * to destination vector [d] and set carry c.
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+
36 * . C, longlong . C, generic
38 #ifndef MBEDTLS_BN_MUL_H
39 #define MBEDTLS_BN_MUL_H
41 #if !defined(MBEDTLS_CONFIG_FILE)
44 #include MBEDTLS_CONFIG_FILE
49 #if defined(MBEDTLS_HAVE_ASM)
55 /* armcc5 --gnu defines __GNUC__ but doesn't support GNU's extended asm */
56 #if defined(__GNUC__) && \
57 ( !defined(__ARMCC_VERSION) || __ARMCC_VERSION >= 6000000 )
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.
65 #if defined(__i386__) && defined(__OPTIMIZE__)
67 #define MULADDC_INIT \
69 "movl %%ebx, %0 \n\t" \
70 "movl %5, %%esi \n\t" \
71 "movl %6, %%edi \n\t" \
72 "movl %7, %%ecx \n\t" \
75 #define MULADDC_CORE \
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" \
85 #if defined(MBEDTLS_HAVE_SSE2)
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"
150 #define MULADDC_STOP \
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" \
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" \
175 #if defined(__amd64__) || defined (__x86_64__)
177 #define MULADDC_INIT \
181 #define MULADDC_CORE \
182 "movq (%%rsi), %%rax\n" \
185 "addq %%rcx, %%rax\n" \
186 "movq %%r8, %%rcx\n" \
189 "addq %%rax, (%%rdi)\n" \
190 "adcq %%rdx, %%rcx\n" \
193 #define MULADDC_STOP \
194 : "+c" (c), "+D" (d), "+S" (s) \
196 : "rax", "rdx", "r8" \
201 #if defined(__mc68020__) || defined(__mcpu32__)
203 #define MULADDC_INIT \
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"
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"
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" \
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"
274 #if defined(__powerpc64__) || defined(__ppc64__)
276 #if defined(__MACH__) && defined(__APPLE__)
278 #define MULADDC_INIT \
284 "addi r3, r3, -8 \n\t" \
285 "addi r4, r4, -8 \n\t" \
286 "addic r5, r5, 0 \n\t"
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"
298 #define MULADDC_STOP \
299 "addze r5, r5 \n\t" \
300 "addi r4, r4, 8 \n\t" \
301 "addi r3, r3, 8 \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" \
311 #else /* __MACH__ && __APPLE__ */
313 #define MULADDC_INIT \
319 "addi %%r3, %%r3, -8 \n\t" \
320 "addi %%r4, %%r4, -8 \n\t" \
321 "addic %%r5, %%r5, 0 \n\t"
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"
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" \
345 #endif /* __MACH__ && __APPLE__ */
347 #elif defined(__powerpc__) || defined(__ppc__) /* end PPC64/begin PPC32 */
349 #if defined(__MACH__) && defined(__APPLE__)
351 #define MULADDC_INIT \
357 "addi r3, r3, -4 \n\t" \
358 "addi r4, r4, -4 \n\t" \
359 "addic r5, r5, 0 \n\t"
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"
371 #define MULADDC_STOP \
372 "addze r5, r5 \n\t" \
373 "addi r4, r4, 4 \n\t" \
374 "addi r3, r3, 4 \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" \
383 #else /* __MACH__ && __APPLE__ */
385 #define MULADDC_INIT \
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"
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"
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" \
417 #endif /* __MACH__ && __APPLE__ */
422 * The Sparc(64) assembly is reported to be broken.
423 * Disable it for now, until we're able to fix it.
425 #if 0 && defined(__sparc__)
426 #if defined(__sparc64__)
428 #define MULADDC_INIT \
430 "ldx %3, %%o0 \n\t" \
431 "ldx %4, %%o1 \n\t" \
435 #define MULADDC_CORE \
436 "ld [%%o0], %%o4 \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" \
448 #define MULADDC_STOP \
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", \
458 #else /* __sparc64__ */
460 #define MULADDC_INIT \
467 #define MULADDC_CORE \
468 "ld [%%o0], %%o4 \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" \
480 #define MULADDC_STOP \
484 : "=m" (c), "=m" (d), "=m" (s) \
485 : "m" (s), "m" (d), "m" (c), "m" (b) \
486 : "g1", "o0", "o1", "o2", "o3", "o4", \
490 #endif /* __sparc64__ */
491 #endif /* __sparc__ */
493 #if defined(__microblaze__) || defined(microblaze)
495 #define MULADDC_INIT \
501 "andi r7, r6, 0xffff \n\t" \
502 "bsrli r6, r6, 16 \n\t"
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"
531 #define MULADDC_STOP \
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" \
541 #endif /* MicroBlaze */
543 #if defined(__tricore__)
545 #define MULADDC_INIT \
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"
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"
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" \
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.
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).
581 * So, only use the optimized assembly below for optimized build, which avoids
582 * the build error and is pretty reasonable anyway.
584 #if defined(__GNUC__) && !defined(__OPTIMIZE__)
585 #define MULADDC_CANNOT_USE_R7
588 #if defined(__arm__) && !defined(MULADDC_CANNOT_USE_R7)
590 #if defined(__thumb__) && !defined(__thumb2__)
592 #define MULADDC_INIT \
598 "lsr r7, r3, #16 \n\t" \
600 "lsl r7, r3, #16 \n\t" \
601 "lsr r7, r7, #16 \n\t" \
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" \
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" \
624 "lsl r3, r6, #16 \n\t" \
625 "add r4, r4, r3 \n\t" \
627 "lsl r3, r7, #16 \n\t" \
628 "add r4, r4, r3 \n\t" \
630 "ldr r3, [r1] \n\t" \
631 "add r4, r4, r3 \n\t" \
633 "stmia r1!, {r4} \n\t"
635 #define MULADDC_STOP \
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" \
645 #elif defined (__ARM_FEATURE_DSP) && (__ARM_FEATURE_DSP == 1)
647 #define MULADDC_INIT \
650 #define MULADDC_CORE \
651 "ldr r0, [%0], #4 \n\t" \
652 "ldr r1, [%1] \n\t" \
653 "umaal r1, %2, %3, r0 \n\t" \
654 "str r1, [%1], #4 \n\t"
656 #define MULADDC_STOP \
657 : "=r" (s), "=r" (d), "=r" (c) \
658 : "r" (b), "0" (s), "1" (d), "2" (c) \
659 : "r0", "r1", "memory" \
664 #define MULADDC_INIT \
671 #define MULADDC_CORE \
672 "ldr r4, [r0], #4 \n\t" \
674 "ldr r6, [r1] \n\t" \
675 "umlal r2, r5, r3, r4 \n\t" \
676 "adds r7, r6, r2 \n\t" \
677 "adc r2, r5, #0 \n\t" \
678 "str r7, [r1], #4 \n\t"
680 #define MULADDC_STOP \
684 : "=m" (c), "=m" (d), "=m" (s) \
685 : "m" (s), "m" (d), "m" (c), "m" (b) \
686 : "r0", "r1", "r2", "r3", "r4", "r5", \
694 #if defined(__alpha__)
696 #define MULADDC_INIT \
703 #define MULADDC_CORE \
704 "ldq $6, 0($1) \n\t" \
705 "addq $1, 8, $1 \n\t" \
706 "mulq $6, $4, $7 \n\t" \
707 "umulh $6, $4, $6 \n\t" \
708 "addq $7, $3, $7 \n\t" \
709 "cmpult $7, $3, $3 \n\t" \
710 "ldq $5, 0($2) \n\t" \
711 "addq $7, $5, $7 \n\t" \
712 "cmpult $7, $5, $5 \n\t" \
713 "stq $7, 0($2) \n\t" \
714 "addq $2, 8, $2 \n\t" \
715 "addq $6, $3, $3 \n\t" \
716 "addq $5, $3, $3 \n\t"
718 #define MULADDC_STOP \
722 : "=m" (c), "=m" (d), "=m" (s) \
723 : "m" (s), "m" (d), "m" (c), "m" (b) \
724 : "$1", "$2", "$3", "$4", "$5", "$6", "$7" \
728 #if defined(__mips__) && !defined(__mips64)
730 #define MULADDC_INIT \
737 #define MULADDC_CORE \
738 "lw $14, 0($10) \n\t" \
739 "multu $13, $14 \n\t" \
740 "addi $10, $10, 4 \n\t" \
743 "addu $14, $12, $14 \n\t" \
744 "lw $15, 0($11) \n\t" \
745 "sltu $12, $14, $12 \n\t" \
746 "addu $15, $14, $15 \n\t" \
747 "sltu $14, $15, $14 \n\t" \
748 "addu $12, $12, $9 \n\t" \
749 "sw $15, 0($11) \n\t" \
750 "addu $12, $12, $14 \n\t" \
751 "addi $11, $11, 4 \n\t"
753 #define MULADDC_STOP \
757 : "=m" (c), "=m" (d), "=m" (s) \
758 : "m" (s), "m" (d), "m" (c), "m" (b) \
759 : "$9", "$10", "$11", "$12", "$13", "$14", "$15", "lo", "hi" \
765 #if (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__)
767 #define MULADDC_INIT \
773 #define MULADDC_CORE \
778 __asm add eax, [edi] \
783 #if defined(MBEDTLS_HAVE_SSE2)
785 #define EMIT __asm _emit
787 #define MULADDC_HUIT \
788 EMIT 0x0F EMIT 0x6E EMIT 0xC9 \
789 EMIT 0x0F EMIT 0x6E EMIT 0xC3 \
790 EMIT 0x0F EMIT 0x6E EMIT 0x1F \
791 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
792 EMIT 0x0F EMIT 0x6E EMIT 0x16 \
793 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
794 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x04 \
795 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
796 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x08 \
797 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
798 EMIT 0x0F EMIT 0x6E EMIT 0x7E EMIT 0x0C \
799 EMIT 0x0F EMIT 0xF4 EMIT 0xF8 \
800 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
801 EMIT 0x0F EMIT 0x6E EMIT 0x5F EMIT 0x04 \
802 EMIT 0x0F EMIT 0xD4 EMIT 0xDC \
803 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x08 \
804 EMIT 0x0F EMIT 0xD4 EMIT 0xEE \
805 EMIT 0x0F EMIT 0x6E EMIT 0x67 EMIT 0x0C \
806 EMIT 0x0F EMIT 0xD4 EMIT 0xFC \
807 EMIT 0x0F EMIT 0x7E EMIT 0x0F \
808 EMIT 0x0F EMIT 0x6E EMIT 0x56 EMIT 0x10 \
809 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
810 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
811 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x14 \
812 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
813 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
814 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x18 \
815 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
816 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x04 \
817 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
818 EMIT 0x0F EMIT 0x6E EMIT 0x5E EMIT 0x1C \
819 EMIT 0x0F EMIT 0xF4 EMIT 0xD8 \
820 EMIT 0x0F EMIT 0xD4 EMIT 0xCD \
821 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x10 \
822 EMIT 0x0F EMIT 0xD4 EMIT 0xD5 \
823 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x08 \
824 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
825 EMIT 0x0F EMIT 0xD4 EMIT 0xCF \
826 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x14 \
827 EMIT 0x0F EMIT 0xD4 EMIT 0xE5 \
828 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x0C \
829 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
830 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
831 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x18 \
832 EMIT 0x0F EMIT 0xD4 EMIT 0xF5 \
833 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x10 \
834 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
835 EMIT 0x0F EMIT 0xD4 EMIT 0xCC \
836 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x1C \
837 EMIT 0x0F EMIT 0xD4 EMIT 0xDD \
838 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x14 \
839 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
840 EMIT 0x0F EMIT 0xD4 EMIT 0xCE \
841 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x18 \
842 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
843 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
844 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x1C \
845 EMIT 0x83 EMIT 0xC7 EMIT 0x20 \
846 EMIT 0x83 EMIT 0xC6 EMIT 0x20 \
847 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
848 EMIT 0x0F EMIT 0x7E EMIT 0xC9
850 #define MULADDC_STOP \
851 EMIT 0x0F EMIT 0x77 \
858 #define MULADDC_STOP \
866 #endif /* MBEDTLS_HAVE_ASM */
868 #if !defined(MULADDC_CORE)
869 #if defined(MBEDTLS_HAVE_UDBL)
871 #define MULADDC_INIT \
874 mbedtls_mpi_uint r0, r1;
876 #define MULADDC_CORE \
877 r = *(s++) * (mbedtls_t_udbl) b; \
878 r0 = (mbedtls_mpi_uint) r; \
879 r1 = (mbedtls_mpi_uint)( r >> biL ); \
880 r0 += c; r1 += (r0 < c); \
881 r0 += *d; r1 += (r0 < *d); \
884 #define MULADDC_STOP \
888 #define MULADDC_INIT \
890 mbedtls_mpi_uint s0, s1, b0, b1; \
891 mbedtls_mpi_uint r0, r1, rx, ry; \
892 b0 = ( b << biH ) >> biH; \
895 #define MULADDC_CORE \
896 s0 = ( *s << biH ) >> biH; \
897 s1 = ( *s >> biH ); s++; \
898 rx = s0 * b1; r0 = s0 * b0; \
899 ry = s1 * b0; r1 = s1 * b1; \
900 r1 += ( rx >> biH ); \
901 r1 += ( ry >> biH ); \
902 rx <<= biH; ry <<= biH; \
903 r0 += rx; r1 += (r0 < rx); \
904 r0 += ry; r1 += (r0 < ry); \
905 r0 += c; r1 += (r0 < c); \
906 r0 += *d; r1 += (r0 < *d); \
909 #define MULADDC_STOP \
912 #endif /* C (generic) */
913 #endif /* C (longlong) */
915 #endif /* bn_mul.h */