mbed TLS v2.16.6
bn_mul.h
Go to the documentation of this file.
1 
6 /*
7  * Copyright (C) 2006-2015, ARM Limited, All Rights Reserved
8  * SPDX-License-Identifier: GPL-2.0
9  *
10  * This program is free software; you can redistribute it and/or modify
11  * it under the terms of the GNU General Public License as published by
12  * the Free Software Foundation; either version 2 of the License, or
13  * (at your option) any later version.
14  *
15  * This program is distributed in the hope that it will be useful,
16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18  * GNU General Public License for more details.
19  *
20  * You should have received a copy of the GNU General Public License along
21  * with this program; if not, write to the Free Software Foundation, Inc.,
22  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
23  *
24  * This file is part of mbed TLS (https://tls.mbed.org)
25  */
26 /*
27  * Multiply source vector [s] with b, add result
28  * to destination vector [d] and set carry c.
29  *
30  * Currently supports:
31  *
32  * . IA-32 (386+) . AMD64 / EM64T
33  * . IA-32 (SSE2) . Motorola 68000
34  * . PowerPC, 32-bit . MicroBlaze
35  * . PowerPC, 64-bit . TriCore
36  * . SPARC v8 . ARM v3+
37  * . Alpha . MIPS32
38  * . C, longlong . C, generic
39  */
40 #ifndef MBEDTLS_BN_MUL_H
41 #define MBEDTLS_BN_MUL_H
42 
43 #if !defined(MBEDTLS_CONFIG_FILE)
44 #include "config.h"
45 #else
46 #include MBEDTLS_CONFIG_FILE
47 #endif
48 
49 #include "bignum.h"
50 
51 #if defined(MBEDTLS_HAVE_ASM)
52 
53 #ifndef asm
54 #define asm __asm
55 #endif
56 
57 /* armcc5 --gnu defines __GNUC__ but doesn't support GNU's extended asm */
58 #if defined(__GNUC__) && \
59  ( !defined(__ARMCC_VERSION) || __ARMCC_VERSION >= 6000000 )
60 
61 /*
62  * Disable use of the i386 assembly code below if option -O0, to disable all
63  * compiler optimisations, is passed, detected with __OPTIMIZE__
64  * This is done as the number of registers used in the assembly code doesn't
65  * work with the -O0 option.
66  */
67 #if defined(__i386__) && defined(__OPTIMIZE__)
68 
69 #define MULADDC_INIT \
70  asm( \
71  "movl %%ebx, %0 \n\t" \
72  "movl %5, %%esi \n\t" \
73  "movl %6, %%edi \n\t" \
74  "movl %7, %%ecx \n\t" \
75  "movl %8, %%ebx \n\t"
76 
77 #define MULADDC_CORE \
78  "lodsl \n\t" \
79  "mull %%ebx \n\t" \
80  "addl %%ecx, %%eax \n\t" \
81  "adcl $0, %%edx \n\t" \
82  "addl (%%edi), %%eax \n\t" \
83  "adcl $0, %%edx \n\t" \
84  "movl %%edx, %%ecx \n\t" \
85  "stosl \n\t"
86 
87 #if defined(MBEDTLS_HAVE_SSE2)
88 
89 #define MULADDC_HUIT \
90  "movd %%ecx, %%mm1 \n\t" \
91  "movd %%ebx, %%mm0 \n\t" \
92  "movd (%%edi), %%mm3 \n\t" \
93  "paddq %%mm3, %%mm1 \n\t" \
94  "movd (%%esi), %%mm2 \n\t" \
95  "pmuludq %%mm0, %%mm2 \n\t" \
96  "movd 4(%%esi), %%mm4 \n\t" \
97  "pmuludq %%mm0, %%mm4 \n\t" \
98  "movd 8(%%esi), %%mm6 \n\t" \
99  "pmuludq %%mm0, %%mm6 \n\t" \
100  "movd 12(%%esi), %%mm7 \n\t" \
101  "pmuludq %%mm0, %%mm7 \n\t" \
102  "paddq %%mm2, %%mm1 \n\t" \
103  "movd 4(%%edi), %%mm3 \n\t" \
104  "paddq %%mm4, %%mm3 \n\t" \
105  "movd 8(%%edi), %%mm5 \n\t" \
106  "paddq %%mm6, %%mm5 \n\t" \
107  "movd 12(%%edi), %%mm4 \n\t" \
108  "paddq %%mm4, %%mm7 \n\t" \
109  "movd %%mm1, (%%edi) \n\t" \
110  "movd 16(%%esi), %%mm2 \n\t" \
111  "pmuludq %%mm0, %%mm2 \n\t" \
112  "psrlq $32, %%mm1 \n\t" \
113  "movd 20(%%esi), %%mm4 \n\t" \
114  "pmuludq %%mm0, %%mm4 \n\t" \
115  "paddq %%mm3, %%mm1 \n\t" \
116  "movd 24(%%esi), %%mm6 \n\t" \
117  "pmuludq %%mm0, %%mm6 \n\t" \
118  "movd %%mm1, 4(%%edi) \n\t" \
119  "psrlq $32, %%mm1 \n\t" \
120  "movd 28(%%esi), %%mm3 \n\t" \
121  "pmuludq %%mm0, %%mm3 \n\t" \
122  "paddq %%mm5, %%mm1 \n\t" \
123  "movd 16(%%edi), %%mm5 \n\t" \
124  "paddq %%mm5, %%mm2 \n\t" \
125  "movd %%mm1, 8(%%edi) \n\t" \
126  "psrlq $32, %%mm1 \n\t" \
127  "paddq %%mm7, %%mm1 \n\t" \
128  "movd 20(%%edi), %%mm5 \n\t" \
129  "paddq %%mm5, %%mm4 \n\t" \
130  "movd %%mm1, 12(%%edi) \n\t" \
131  "psrlq $32, %%mm1 \n\t" \
132  "paddq %%mm2, %%mm1 \n\t" \
133  "movd 24(%%edi), %%mm5 \n\t" \
134  "paddq %%mm5, %%mm6 \n\t" \
135  "movd %%mm1, 16(%%edi) \n\t" \
136  "psrlq $32, %%mm1 \n\t" \
137  "paddq %%mm4, %%mm1 \n\t" \
138  "movd 28(%%edi), %%mm5 \n\t" \
139  "paddq %%mm5, %%mm3 \n\t" \
140  "movd %%mm1, 20(%%edi) \n\t" \
141  "psrlq $32, %%mm1 \n\t" \
142  "paddq %%mm6, %%mm1 \n\t" \
143  "movd %%mm1, 24(%%edi) \n\t" \
144  "psrlq $32, %%mm1 \n\t" \
145  "paddq %%mm3, %%mm1 \n\t" \
146  "movd %%mm1, 28(%%edi) \n\t" \
147  "addl $32, %%edi \n\t" \
148  "addl $32, %%esi \n\t" \
149  "psrlq $32, %%mm1 \n\t" \
150  "movd %%mm1, %%ecx \n\t"
151 
152 #define MULADDC_STOP \
153  "emms \n\t" \
154  "movl %4, %%ebx \n\t" \
155  "movl %%ecx, %1 \n\t" \
156  "movl %%edi, %2 \n\t" \
157  "movl %%esi, %3 \n\t" \
158  : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \
159  : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \
160  : "eax", "ebx", "ecx", "edx", "esi", "edi" \
161  );
162 
163 #else
164 
165 #define MULADDC_STOP \
166  "movl %4, %%ebx \n\t" \
167  "movl %%ecx, %1 \n\t" \
168  "movl %%edi, %2 \n\t" \
169  "movl %%esi, %3 \n\t" \
170  : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \
171  : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \
172  : "eax", "ebx", "ecx", "edx", "esi", "edi" \
173  );
174 #endif /* SSE2 */
175 #endif /* i386 */
176 
177 #if defined(__amd64__) || defined (__x86_64__)
178 
179 #define MULADDC_INIT \
180  asm( \
181  "xorq %%r8, %%r8\n"
182 
183 #define MULADDC_CORE \
184  "movq (%%rsi), %%rax\n" \
185  "mulq %%rbx\n" \
186  "addq $8, %%rsi\n" \
187  "addq %%rcx, %%rax\n" \
188  "movq %%r8, %%rcx\n" \
189  "adcq $0, %%rdx\n" \
190  "nop \n" \
191  "addq %%rax, (%%rdi)\n" \
192  "adcq %%rdx, %%rcx\n" \
193  "addq $8, %%rdi\n"
194 
195 #define MULADDC_STOP \
196  : "+c" (c), "+D" (d), "+S" (s) \
197  : "b" (b) \
198  : "rax", "rdx", "r8" \
199  );
200 
201 #endif /* AMD64 */
202 
203 #if defined(__mc68020__) || defined(__mcpu32__)
204 
205 #define MULADDC_INIT \
206  asm( \
207  "movl %3, %%a2 \n\t" \
208  "movl %4, %%a3 \n\t" \
209  "movl %5, %%d3 \n\t" \
210  "movl %6, %%d2 \n\t" \
211  "moveq #0, %%d0 \n\t"
212 
213 #define MULADDC_CORE \
214  "movel %%a2@+, %%d1 \n\t" \
215  "mulul %%d2, %%d4:%%d1 \n\t" \
216  "addl %%d3, %%d1 \n\t" \
217  "addxl %%d0, %%d4 \n\t" \
218  "moveq #0, %%d3 \n\t" \
219  "addl %%d1, %%a3@+ \n\t" \
220  "addxl %%d4, %%d3 \n\t"
221 
222 #define MULADDC_STOP \
223  "movl %%d3, %0 \n\t" \
224  "movl %%a3, %1 \n\t" \
225  "movl %%a2, %2 \n\t" \
226  : "=m" (c), "=m" (d), "=m" (s) \
227  : "m" (s), "m" (d), "m" (c), "m" (b) \
228  : "d0", "d1", "d2", "d3", "d4", "a2", "a3" \
229  );
230 
231 #define MULADDC_HUIT \
232  "movel %%a2@+, %%d1 \n\t" \
233  "mulul %%d2, %%d4:%%d1 \n\t" \
234  "addxl %%d3, %%d1 \n\t" \
235  "addxl %%d0, %%d4 \n\t" \
236  "addl %%d1, %%a3@+ \n\t" \
237  "movel %%a2@+, %%d1 \n\t" \
238  "mulul %%d2, %%d3:%%d1 \n\t" \
239  "addxl %%d4, %%d1 \n\t" \
240  "addxl %%d0, %%d3 \n\t" \
241  "addl %%d1, %%a3@+ \n\t" \
242  "movel %%a2@+, %%d1 \n\t" \
243  "mulul %%d2, %%d4:%%d1 \n\t" \
244  "addxl %%d3, %%d1 \n\t" \
245  "addxl %%d0, %%d4 \n\t" \
246  "addl %%d1, %%a3@+ \n\t" \
247  "movel %%a2@+, %%d1 \n\t" \
248  "mulul %%d2, %%d3:%%d1 \n\t" \
249  "addxl %%d4, %%d1 \n\t" \
250  "addxl %%d0, %%d3 \n\t" \
251  "addl %%d1, %%a3@+ \n\t" \
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  "addxl %%d0, %%d3 \n\t"
273 
274 #endif /* MC68000 */
275 
276 #if defined(__powerpc64__) || defined(__ppc64__)
277 
278 #if defined(__MACH__) && defined(__APPLE__)
279 
280 #define MULADDC_INIT \
281  asm( \
282  "ld r3, %3 \n\t" \
283  "ld r4, %4 \n\t" \
284  "ld r5, %5 \n\t" \
285  "ld r6, %6 \n\t" \
286  "addi r3, r3, -8 \n\t" \
287  "addi r4, r4, -8 \n\t" \
288  "addic r5, r5, 0 \n\t"
289 
290 #define MULADDC_CORE \
291  "ldu r7, 8(r3) \n\t" \
292  "mulld r8, r7, r6 \n\t" \
293  "mulhdu r9, r7, r6 \n\t" \
294  "adde r8, r8, r5 \n\t" \
295  "ld r7, 8(r4) \n\t" \
296  "addze r5, r9 \n\t" \
297  "addc r8, r8, r7 \n\t" \
298  "stdu r8, 8(r4) \n\t"
299 
300 #define MULADDC_STOP \
301  "addze r5, r5 \n\t" \
302  "addi r4, r4, 8 \n\t" \
303  "addi r3, r3, 8 \n\t" \
304  "std r5, %0 \n\t" \
305  "std r4, %1 \n\t" \
306  "std r3, %2 \n\t" \
307  : "=m" (c), "=m" (d), "=m" (s) \
308  : "m" (s), "m" (d), "m" (c), "m" (b) \
309  : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \
310  );
311 
312 
313 #else /* __MACH__ && __APPLE__ */
314 
315 #define MULADDC_INIT \
316  asm( \
317  "ld %%r3, %3 \n\t" \
318  "ld %%r4, %4 \n\t" \
319  "ld %%r5, %5 \n\t" \
320  "ld %%r6, %6 \n\t" \
321  "addi %%r3, %%r3, -8 \n\t" \
322  "addi %%r4, %%r4, -8 \n\t" \
323  "addic %%r5, %%r5, 0 \n\t"
324 
325 #define MULADDC_CORE \
326  "ldu %%r7, 8(%%r3) \n\t" \
327  "mulld %%r8, %%r7, %%r6 \n\t" \
328  "mulhdu %%r9, %%r7, %%r6 \n\t" \
329  "adde %%r8, %%r8, %%r5 \n\t" \
330  "ld %%r7, 8(%%r4) \n\t" \
331  "addze %%r5, %%r9 \n\t" \
332  "addc %%r8, %%r8, %%r7 \n\t" \
333  "stdu %%r8, 8(%%r4) \n\t"
334 
335 #define MULADDC_STOP \
336  "addze %%r5, %%r5 \n\t" \
337  "addi %%r4, %%r4, 8 \n\t" \
338  "addi %%r3, %%r3, 8 \n\t" \
339  "std %%r5, %0 \n\t" \
340  "std %%r4, %1 \n\t" \
341  "std %%r3, %2 \n\t" \
342  : "=m" (c), "=m" (d), "=m" (s) \
343  : "m" (s), "m" (d), "m" (c), "m" (b) \
344  : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \
345  );
346 
347 #endif /* __MACH__ && __APPLE__ */
348 
349 #elif defined(__powerpc__) || defined(__ppc__) /* end PPC64/begin PPC32 */
350 
351 #if defined(__MACH__) && defined(__APPLE__)
352 
353 #define MULADDC_INIT \
354  asm( \
355  "lwz r3, %3 \n\t" \
356  "lwz r4, %4 \n\t" \
357  "lwz r5, %5 \n\t" \
358  "lwz r6, %6 \n\t" \
359  "addi r3, r3, -4 \n\t" \
360  "addi r4, r4, -4 \n\t" \
361  "addic r5, r5, 0 \n\t"
362 
363 #define MULADDC_CORE \
364  "lwzu r7, 4(r3) \n\t" \
365  "mullw r8, r7, r6 \n\t" \
366  "mulhwu r9, r7, r6 \n\t" \
367  "adde r8, r8, r5 \n\t" \
368  "lwz r7, 4(r4) \n\t" \
369  "addze r5, r9 \n\t" \
370  "addc r8, r8, r7 \n\t" \
371  "stwu r8, 4(r4) \n\t"
372 
373 #define MULADDC_STOP \
374  "addze r5, r5 \n\t" \
375  "addi r4, r4, 4 \n\t" \
376  "addi r3, r3, 4 \n\t" \
377  "stw r5, %0 \n\t" \
378  "stw r4, %1 \n\t" \
379  "stw r3, %2 \n\t" \
380  : "=m" (c), "=m" (d), "=m" (s) \
381  : "m" (s), "m" (d), "m" (c), "m" (b) \
382  : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \
383  );
384 
385 #else /* __MACH__ && __APPLE__ */
386 
387 #define MULADDC_INIT \
388  asm( \
389  "lwz %%r3, %3 \n\t" \
390  "lwz %%r4, %4 \n\t" \
391  "lwz %%r5, %5 \n\t" \
392  "lwz %%r6, %6 \n\t" \
393  "addi %%r3, %%r3, -4 \n\t" \
394  "addi %%r4, %%r4, -4 \n\t" \
395  "addic %%r5, %%r5, 0 \n\t"
396 
397 #define MULADDC_CORE \
398  "lwzu %%r7, 4(%%r3) \n\t" \
399  "mullw %%r8, %%r7, %%r6 \n\t" \
400  "mulhwu %%r9, %%r7, %%r6 \n\t" \
401  "adde %%r8, %%r8, %%r5 \n\t" \
402  "lwz %%r7, 4(%%r4) \n\t" \
403  "addze %%r5, %%r9 \n\t" \
404  "addc %%r8, %%r8, %%r7 \n\t" \
405  "stwu %%r8, 4(%%r4) \n\t"
406 
407 #define MULADDC_STOP \
408  "addze %%r5, %%r5 \n\t" \
409  "addi %%r4, %%r4, 4 \n\t" \
410  "addi %%r3, %%r3, 4 \n\t" \
411  "stw %%r5, %0 \n\t" \
412  "stw %%r4, %1 \n\t" \
413  "stw %%r3, %2 \n\t" \
414  : "=m" (c), "=m" (d), "=m" (s) \
415  : "m" (s), "m" (d), "m" (c), "m" (b) \
416  : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \
417  );
418 
419 #endif /* __MACH__ && __APPLE__ */
420 
421 #endif /* PPC32 */
422 
423 /*
424  * The Sparc(64) assembly is reported to be broken.
425  * Disable it for now, until we're able to fix it.
426  */
427 #if 0 && defined(__sparc__)
428 #if defined(__sparc64__)
429 
430 #define MULADDC_INIT \
431  asm( \
432  "ldx %3, %%o0 \n\t" \
433  "ldx %4, %%o1 \n\t" \
434  "ld %5, %%o2 \n\t" \
435  "ld %6, %%o3 \n\t"
436 
437 #define MULADDC_CORE \
438  "ld [%%o0], %%o4 \n\t" \
439  "inc 4, %%o0 \n\t" \
440  "ld [%%o1], %%o5 \n\t" \
441  "umul %%o3, %%o4, %%o4 \n\t" \
442  "addcc %%o4, %%o2, %%o4 \n\t" \
443  "rd %%y, %%g1 \n\t" \
444  "addx %%g1, 0, %%g1 \n\t" \
445  "addcc %%o4, %%o5, %%o4 \n\t" \
446  "st %%o4, [%%o1] \n\t" \
447  "addx %%g1, 0, %%o2 \n\t" \
448  "inc 4, %%o1 \n\t"
449 
450  #define MULADDC_STOP \
451  "st %%o2, %0 \n\t" \
452  "stx %%o1, %1 \n\t" \
453  "stx %%o0, %2 \n\t" \
454  : "=m" (c), "=m" (d), "=m" (s) \
455  : "m" (s), "m" (d), "m" (c), "m" (b) \
456  : "g1", "o0", "o1", "o2", "o3", "o4", \
457  "o5" \
458  );
459 
460 #else /* __sparc64__ */
461 
462 #define MULADDC_INIT \
463  asm( \
464  "ld %3, %%o0 \n\t" \
465  "ld %4, %%o1 \n\t" \
466  "ld %5, %%o2 \n\t" \
467  "ld %6, %%o3 \n\t"
468 
469 #define MULADDC_CORE \
470  "ld [%%o0], %%o4 \n\t" \
471  "inc 4, %%o0 \n\t" \
472  "ld [%%o1], %%o5 \n\t" \
473  "umul %%o3, %%o4, %%o4 \n\t" \
474  "addcc %%o4, %%o2, %%o4 \n\t" \
475  "rd %%y, %%g1 \n\t" \
476  "addx %%g1, 0, %%g1 \n\t" \
477  "addcc %%o4, %%o5, %%o4 \n\t" \
478  "st %%o4, [%%o1] \n\t" \
479  "addx %%g1, 0, %%o2 \n\t" \
480  "inc 4, %%o1 \n\t"
481 
482 #define MULADDC_STOP \
483  "st %%o2, %0 \n\t" \
484  "st %%o1, %1 \n\t" \
485  "st %%o0, %2 \n\t" \
486  : "=m" (c), "=m" (d), "=m" (s) \
487  : "m" (s), "m" (d), "m" (c), "m" (b) \
488  : "g1", "o0", "o1", "o2", "o3", "o4", \
489  "o5" \
490  );
491 
492 #endif /* __sparc64__ */
493 #endif /* __sparc__ */
494 
495 #if defined(__microblaze__) || defined(microblaze)
496 
497 #define MULADDC_INIT \
498  asm( \
499  "lwi r3, %3 \n\t" \
500  "lwi r4, %4 \n\t" \
501  "lwi r5, %5 \n\t" \
502  "lwi r6, %6 \n\t" \
503  "andi r7, r6, 0xffff \n\t" \
504  "bsrli r6, r6, 16 \n\t"
505 
506 #define MULADDC_CORE \
507  "lhui r8, r3, 0 \n\t" \
508  "addi r3, r3, 2 \n\t" \
509  "lhui r9, r3, 0 \n\t" \
510  "addi r3, r3, 2 \n\t" \
511  "mul r10, r9, r6 \n\t" \
512  "mul r11, r8, r7 \n\t" \
513  "mul r12, r9, r7 \n\t" \
514  "mul r13, r8, r6 \n\t" \
515  "bsrli r8, r10, 16 \n\t" \
516  "bsrli r9, r11, 16 \n\t" \
517  "add r13, r13, r8 \n\t" \
518  "add r13, r13, r9 \n\t" \
519  "bslli r10, r10, 16 \n\t" \
520  "bslli r11, r11, 16 \n\t" \
521  "add r12, r12, r10 \n\t" \
522  "addc r13, r13, r0 \n\t" \
523  "add r12, r12, r11 \n\t" \
524  "addc r13, r13, r0 \n\t" \
525  "lwi r10, r4, 0 \n\t" \
526  "add r12, r12, r10 \n\t" \
527  "addc r13, r13, r0 \n\t" \
528  "add r12, r12, r5 \n\t" \
529  "addc r5, r13, r0 \n\t" \
530  "swi r12, r4, 0 \n\t" \
531  "addi r4, r4, 4 \n\t"
532 
533 #define MULADDC_STOP \
534  "swi r5, %0 \n\t" \
535  "swi r4, %1 \n\t" \
536  "swi r3, %2 \n\t" \
537  : "=m" (c), "=m" (d), "=m" (s) \
538  : "m" (s), "m" (d), "m" (c), "m" (b) \
539  : "r3", "r4", "r5", "r6", "r7", "r8", \
540  "r9", "r10", "r11", "r12", "r13" \
541  );
542 
543 #endif /* MicroBlaze */
544 
545 #if defined(__tricore__)
546 
547 #define MULADDC_INIT \
548  asm( \
549  "ld.a %%a2, %3 \n\t" \
550  "ld.a %%a3, %4 \n\t" \
551  "ld.w %%d4, %5 \n\t" \
552  "ld.w %%d1, %6 \n\t" \
553  "xor %%d5, %%d5 \n\t"
554 
555 #define MULADDC_CORE \
556  "ld.w %%d0, [%%a2+] \n\t" \
557  "madd.u %%e2, %%e4, %%d0, %%d1 \n\t" \
558  "ld.w %%d0, [%%a3] \n\t" \
559  "addx %%d2, %%d2, %%d0 \n\t" \
560  "addc %%d3, %%d3, 0 \n\t" \
561  "mov %%d4, %%d3 \n\t" \
562  "st.w [%%a3+], %%d2 \n\t"
563 
564 #define MULADDC_STOP \
565  "st.w %0, %%d4 \n\t" \
566  "st.a %1, %%a3 \n\t" \
567  "st.a %2, %%a2 \n\t" \
568  : "=m" (c), "=m" (d), "=m" (s) \
569  : "m" (s), "m" (d), "m" (c), "m" (b) \
570  : "d0", "d1", "e2", "d4", "a2", "a3" \
571  );
572 
573 #endif /* TriCore */
574 
575 /*
576  * Note, gcc -O0 by default uses r7 for the frame pointer, so it complains about
577  * our use of r7 below, unless -fomit-frame-pointer is passed.
578  *
579  * On the other hand, -fomit-frame-pointer is implied by any -Ox options with
580  * x !=0, which we can detect using __OPTIMIZE__ (which is also defined by
581  * clang and armcc5 under the same conditions).
582  *
583  * So, only use the optimized assembly below for optimized build, which avoids
584  * the build error and is pretty reasonable anyway.
585  */
586 #if defined(__GNUC__) && !defined(__OPTIMIZE__)
587 #define MULADDC_CANNOT_USE_R7
588 #endif
589 
590 #if defined(__arm__) && !defined(MULADDC_CANNOT_USE_R7)
591 
592 #if defined(__thumb__) && !defined(__thumb2__)
593 
594 #define MULADDC_INIT \
595  asm( \
596  "ldr r0, %3 \n\t" \
597  "ldr r1, %4 \n\t" \
598  "ldr r2, %5 \n\t" \
599  "ldr r3, %6 \n\t" \
600  "lsr r7, r3, #16 \n\t" \
601  "mov r9, r7 \n\t" \
602  "lsl r7, r3, #16 \n\t" \
603  "lsr r7, r7, #16 \n\t" \
604  "mov r8, r7 \n\t"
605 
606 #define MULADDC_CORE \
607  "ldmia r0!, {r6} \n\t" \
608  "lsr r7, r6, #16 \n\t" \
609  "lsl r6, r6, #16 \n\t" \
610  "lsr r6, r6, #16 \n\t" \
611  "mov r4, r8 \n\t" \
612  "mul r4, r6 \n\t" \
613  "mov r3, r9 \n\t" \
614  "mul r6, r3 \n\t" \
615  "mov r5, r9 \n\t" \
616  "mul r5, r7 \n\t" \
617  "mov r3, r8 \n\t" \
618  "mul r7, r3 \n\t" \
619  "lsr r3, r6, #16 \n\t" \
620  "add r5, r5, r3 \n\t" \
621  "lsr r3, r7, #16 \n\t" \
622  "add r5, r5, r3 \n\t" \
623  "add r4, r4, r2 \n\t" \
624  "mov r2, #0 \n\t" \
625  "adc r5, r2 \n\t" \
626  "lsl r3, r6, #16 \n\t" \
627  "add r4, r4, r3 \n\t" \
628  "adc r5, r2 \n\t" \
629  "lsl r3, r7, #16 \n\t" \
630  "add r4, r4, r3 \n\t" \
631  "adc r5, r2 \n\t" \
632  "ldr r3, [r1] \n\t" \
633  "add r4, r4, r3 \n\t" \
634  "adc r2, r5 \n\t" \
635  "stmia r1!, {r4} \n\t"
636 
637 #define MULADDC_STOP \
638  "str r2, %0 \n\t" \
639  "str r1, %1 \n\t" \
640  "str r0, %2 \n\t" \
641  : "=m" (c), "=m" (d), "=m" (s) \
642  : "m" (s), "m" (d), "m" (c), "m" (b) \
643  : "r0", "r1", "r2", "r3", "r4", "r5", \
644  "r6", "r7", "r8", "r9", "cc" \
645  );
646 
647 #elif (__ARM_ARCH >= 6) && \
648  defined (__ARM_FEATURE_DSP) && (__ARM_FEATURE_DSP == 1)
649 
650 #define MULADDC_INIT \
651  asm(
652 
653 #define MULADDC_CORE \
654  "ldr r0, [%0], #4 \n\t" \
655  "ldr r1, [%1] \n\t" \
656  "umaal r1, %2, %3, r0 \n\t" \
657  "str r1, [%1], #4 \n\t"
658 
659 #define MULADDC_STOP \
660  : "=r" (s), "=r" (d), "=r" (c) \
661  : "r" (b), "0" (s), "1" (d), "2" (c) \
662  : "r0", "r1", "memory" \
663  );
664 
665 #else
666 
667 #define MULADDC_INIT \
668  asm( \
669  "ldr r0, %3 \n\t" \
670  "ldr r1, %4 \n\t" \
671  "ldr r2, %5 \n\t" \
672  "ldr r3, %6 \n\t"
673 
674 #define MULADDC_CORE \
675  "ldr r4, [r0], #4 \n\t" \
676  "mov r5, #0 \n\t" \
677  "ldr r6, [r1] \n\t" \
678  "umlal r2, r5, r3, r4 \n\t" \
679  "adds r7, r6, r2 \n\t" \
680  "adc r2, r5, #0 \n\t" \
681  "str r7, [r1], #4 \n\t"
682 
683 #define MULADDC_STOP \
684  "str r2, %0 \n\t" \
685  "str r1, %1 \n\t" \
686  "str r0, %2 \n\t" \
687  : "=m" (c), "=m" (d), "=m" (s) \
688  : "m" (s), "m" (d), "m" (c), "m" (b) \
689  : "r0", "r1", "r2", "r3", "r4", "r5", \
690  "r6", "r7", "cc" \
691  );
692 
693 #endif /* Thumb */
694 
695 #endif /* ARMv3 */
696 
697 #if defined(__alpha__)
698 
699 #define MULADDC_INIT \
700  asm( \
701  "ldq $1, %3 \n\t" \
702  "ldq $2, %4 \n\t" \
703  "ldq $3, %5 \n\t" \
704  "ldq $4, %6 \n\t"
705 
706 #define MULADDC_CORE \
707  "ldq $6, 0($1) \n\t" \
708  "addq $1, 8, $1 \n\t" \
709  "mulq $6, $4, $7 \n\t" \
710  "umulh $6, $4, $6 \n\t" \
711  "addq $7, $3, $7 \n\t" \
712  "cmpult $7, $3, $3 \n\t" \
713  "ldq $5, 0($2) \n\t" \
714  "addq $7, $5, $7 \n\t" \
715  "cmpult $7, $5, $5 \n\t" \
716  "stq $7, 0($2) \n\t" \
717  "addq $2, 8, $2 \n\t" \
718  "addq $6, $3, $3 \n\t" \
719  "addq $5, $3, $3 \n\t"
720 
721 #define MULADDC_STOP \
722  "stq $3, %0 \n\t" \
723  "stq $2, %1 \n\t" \
724  "stq $1, %2 \n\t" \
725  : "=m" (c), "=m" (d), "=m" (s) \
726  : "m" (s), "m" (d), "m" (c), "m" (b) \
727  : "$1", "$2", "$3", "$4", "$5", "$6", "$7" \
728  );
729 #endif /* Alpha */
730 
731 #if defined(__mips__) && !defined(__mips64)
732 
733 #define MULADDC_INIT \
734  asm( \
735  "lw $10, %3 \n\t" \
736  "lw $11, %4 \n\t" \
737  "lw $12, %5 \n\t" \
738  "lw $13, %6 \n\t"
739 
740 #define MULADDC_CORE \
741  "lw $14, 0($10) \n\t" \
742  "multu $13, $14 \n\t" \
743  "addi $10, $10, 4 \n\t" \
744  "mflo $14 \n\t" \
745  "mfhi $9 \n\t" \
746  "addu $14, $12, $14 \n\t" \
747  "lw $15, 0($11) \n\t" \
748  "sltu $12, $14, $12 \n\t" \
749  "addu $15, $14, $15 \n\t" \
750  "sltu $14, $15, $14 \n\t" \
751  "addu $12, $12, $9 \n\t" \
752  "sw $15, 0($11) \n\t" \
753  "addu $12, $12, $14 \n\t" \
754  "addi $11, $11, 4 \n\t"
755 
756 #define MULADDC_STOP \
757  "sw $12, %0 \n\t" \
758  "sw $11, %1 \n\t" \
759  "sw $10, %2 \n\t" \
760  : "=m" (c), "=m" (d), "=m" (s) \
761  : "m" (s), "m" (d), "m" (c), "m" (b) \
762  : "$9", "$10", "$11", "$12", "$13", "$14", "$15", "lo", "hi" \
763  );
764 
765 #endif /* MIPS */
766 #endif /* GNUC */
767 
768 #if (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__)
769 
770 #define MULADDC_INIT \
771  __asm mov esi, s \
772  __asm mov edi, d \
773  __asm mov ecx, c \
774  __asm mov ebx, b
775 
776 #define MULADDC_CORE \
777  __asm lodsd \
778  __asm mul ebx \
779  __asm add eax, ecx \
780  __asm adc edx, 0 \
781  __asm add eax, [edi] \
782  __asm adc edx, 0 \
783  __asm mov ecx, edx \
784  __asm stosd
785 
786 #if defined(MBEDTLS_HAVE_SSE2)
787 
788 #define EMIT __asm _emit
789 
790 #define MULADDC_HUIT \
791  EMIT 0x0F EMIT 0x6E EMIT 0xC9 \
792  EMIT 0x0F EMIT 0x6E EMIT 0xC3 \
793  EMIT 0x0F EMIT 0x6E EMIT 0x1F \
794  EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
795  EMIT 0x0F EMIT 0x6E EMIT 0x16 \
796  EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
797  EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x04 \
798  EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
799  EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x08 \
800  EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
801  EMIT 0x0F EMIT 0x6E EMIT 0x7E EMIT 0x0C \
802  EMIT 0x0F EMIT 0xF4 EMIT 0xF8 \
803  EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
804  EMIT 0x0F EMIT 0x6E EMIT 0x5F EMIT 0x04 \
805  EMIT 0x0F EMIT 0xD4 EMIT 0xDC \
806  EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x08 \
807  EMIT 0x0F EMIT 0xD4 EMIT 0xEE \
808  EMIT 0x0F EMIT 0x6E EMIT 0x67 EMIT 0x0C \
809  EMIT 0x0F EMIT 0xD4 EMIT 0xFC \
810  EMIT 0x0F EMIT 0x7E EMIT 0x0F \
811  EMIT 0x0F EMIT 0x6E EMIT 0x56 EMIT 0x10 \
812  EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
813  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
814  EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x14 \
815  EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
816  EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
817  EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x18 \
818  EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
819  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x04 \
820  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
821  EMIT 0x0F EMIT 0x6E EMIT 0x5E EMIT 0x1C \
822  EMIT 0x0F EMIT 0xF4 EMIT 0xD8 \
823  EMIT 0x0F EMIT 0xD4 EMIT 0xCD \
824  EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x10 \
825  EMIT 0x0F EMIT 0xD4 EMIT 0xD5 \
826  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x08 \
827  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
828  EMIT 0x0F EMIT 0xD4 EMIT 0xCF \
829  EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x14 \
830  EMIT 0x0F EMIT 0xD4 EMIT 0xE5 \
831  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x0C \
832  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
833  EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
834  EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x18 \
835  EMIT 0x0F EMIT 0xD4 EMIT 0xF5 \
836  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x10 \
837  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
838  EMIT 0x0F EMIT 0xD4 EMIT 0xCC \
839  EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x1C \
840  EMIT 0x0F EMIT 0xD4 EMIT 0xDD \
841  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x14 \
842  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
843  EMIT 0x0F EMIT 0xD4 EMIT 0xCE \
844  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x18 \
845  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
846  EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
847  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x1C \
848  EMIT 0x83 EMIT 0xC7 EMIT 0x20 \
849  EMIT 0x83 EMIT 0xC6 EMIT 0x20 \
850  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
851  EMIT 0x0F EMIT 0x7E EMIT 0xC9
852 
853 #define MULADDC_STOP \
854  EMIT 0x0F EMIT 0x77 \
855  __asm mov c, ecx \
856  __asm mov d, edi \
857  __asm mov s, esi \
858 
859 #else
860 
861 #define MULADDC_STOP \
862  __asm mov c, ecx \
863  __asm mov d, edi \
864  __asm mov s, esi \
865 
866 #endif /* SSE2 */
867 #endif /* MSVC */
868 
869 #endif /* MBEDTLS_HAVE_ASM */
870 
871 #if !defined(MULADDC_CORE)
872 #if defined(MBEDTLS_HAVE_UDBL)
873 
874 #define MULADDC_INIT \
875 { \
876  mbedtls_t_udbl r; \
877  mbedtls_mpi_uint r0, r1;
878 
879 #define MULADDC_CORE \
880  r = *(s++) * (mbedtls_t_udbl) b; \
881  r0 = (mbedtls_mpi_uint) r; \
882  r1 = (mbedtls_mpi_uint)( r >> biL ); \
883  r0 += c; r1 += (r0 < c); \
884  r0 += *d; r1 += (r0 < *d); \
885  c = r1; *(d++) = r0;
886 
887 #define MULADDC_STOP \
888 }
889 
890 #else
891 #define MULADDC_INIT \
892 { \
893  mbedtls_mpi_uint s0, s1, b0, b1; \
894  mbedtls_mpi_uint r0, r1, rx, ry; \
895  b0 = ( b << biH ) >> biH; \
896  b1 = ( b >> biH );
897 
898 #define MULADDC_CORE \
899  s0 = ( *s << biH ) >> biH; \
900  s1 = ( *s >> biH ); s++; \
901  rx = s0 * b1; r0 = s0 * b0; \
902  ry = s1 * b0; r1 = s1 * b1; \
903  r1 += ( rx >> biH ); \
904  r1 += ( ry >> biH ); \
905  rx <<= biH; ry <<= biH; \
906  r0 += rx; r1 += (r0 < rx); \
907  r0 += ry; r1 += (r0 < ry); \
908  r0 += c; r1 += (r0 < c); \
909  r0 += *d; r1 += (r0 < *d); \
910  c = r1; *(d++) = r0;
911 
912 #define MULADDC_STOP \
913 }
914 
915 #endif /* C (generic) */
916 #endif /* C (longlong) */
917 
918 #endif /* bn_mul.h */
Configuration options (set of defines)
Multi-precision integer library.