1 /**
2  * \file bn_mul.h
3  *
4  * \brief Multi-precision integer library
5  */
6 /*
7  *  Copyright The Mbed TLS Contributors
8  *  SPDX-License-Identifier: Apache-2.0 OR GPL-2.0-or-later
9  *
10  *  This file is provided under the Apache License 2.0, or the
11  *  GNU General Public License v2.0 or later.
12  *
13  *  **********
14  *  Apache License 2.0:
15  *
16  *  Licensed under the Apache License, Version 2.0 (the "License"); you may
17  *  not use this file except in compliance with the License.
18  *  You may obtain a copy of the License at
19  *
20  *  http://www.apache.org/licenses/LICENSE-2.0
21  *
22  *  Unless required by applicable law or agreed to in writing, software
23  *  distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
24  *  WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
25  *  See the License for the specific language governing permissions and
26  *  limitations under the License.
27  *
28  *  **********
29  *
30  *  **********
31  *  GNU General Public License v2.0 or later:
32  *
33  *  This program is free software; you can redistribute it and/or modify
34  *  it under the terms of the GNU General Public License as published by
35  *  the Free Software Foundation; either version 2 of the License, or
36  *  (at your option) any later version.
37  *
38  *  This program is distributed in the hope that it will be useful,
39  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
40  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
41  *  GNU General Public License for more details.
42  *
43  *  You should have received a copy of the GNU General Public License along
44  *  with this program; if not, write to the Free Software Foundation, Inc.,
45  *  51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
46  *
47  *  **********
48  */
49 /*
50  *      Multiply source vector [s] with b, add result
51  *       to destination vector [d] and set carry c.
52  *
53  *      Currently supports:
54  *
55  *         . IA-32 (386+)         . AMD64 / EM64T
56  *         . IA-32 (SSE2)         . Motorola 68000
57  *         . PowerPC, 32-bit      . MicroBlaze
58  *         . PowerPC, 64-bit      . TriCore
59  *         . SPARC v8             . ARM v3+
60  *         . Alpha                . MIPS32
61  *         . C, longlong          . C, generic
62  */
63 #ifndef MBEDTLS_BN_MUL_H
64 #define MBEDTLS_BN_MUL_H
65 
66 #if !defined(MBEDTLS_CONFIG_FILE)
67 #include "config.h"
68 #else
69 #include MBEDTLS_CONFIG_FILE
70 #endif
71 
72 #include "bignum.h"
73 
74 
75 /*
76  * Conversion macros for embedded constants:
77  * build lists of mbedtls_mpi_uint's from lists of unsigned char's grouped by 8, 4 or 2
78  */
79 #if defined(MBEDTLS_HAVE_INT32)
80 
81 #define MBEDTLS_BYTES_TO_T_UINT_4( a, b, c, d )               \
82     ( (mbedtls_mpi_uint) (a) <<  0 ) |                        \
83     ( (mbedtls_mpi_uint) (b) <<  8 ) |                        \
84     ( (mbedtls_mpi_uint) (c) << 16 ) |                        \
85     ( (mbedtls_mpi_uint) (d) << 24 )
86 
87 #define MBEDTLS_BYTES_TO_T_UINT_2( a, b )                   \
88     MBEDTLS_BYTES_TO_T_UINT_4( a, b, 0, 0 )
89 
90 #define MBEDTLS_BYTES_TO_T_UINT_8( a, b, c, d, e, f, g, h ) \
91     MBEDTLS_BYTES_TO_T_UINT_4( a, b, c, d ),                \
92     MBEDTLS_BYTES_TO_T_UINT_4( e, f, g, h )
93 
94 #else /* 64-bits */
95 
96 #define MBEDTLS_BYTES_TO_T_UINT_8( a, b, c, d, e, f, g, h )   \
97     ( (mbedtls_mpi_uint) (a) <<  0 ) |                        \
98     ( (mbedtls_mpi_uint) (b) <<  8 ) |                        \
99     ( (mbedtls_mpi_uint) (c) << 16 ) |                        \
100     ( (mbedtls_mpi_uint) (d) << 24 ) |                        \
101     ( (mbedtls_mpi_uint) (e) << 32 ) |                        \
102     ( (mbedtls_mpi_uint) (f) << 40 ) |                        \
103     ( (mbedtls_mpi_uint) (g) << 48 ) |                        \
104     ( (mbedtls_mpi_uint) (h) << 56 )
105 
106 #define MBEDTLS_BYTES_TO_T_UINT_4( a, b, c, d )             \
107     MBEDTLS_BYTES_TO_T_UINT_8( a, b, c, d, 0, 0, 0, 0 )
108 
109 #define MBEDTLS_BYTES_TO_T_UINT_2( a, b )                   \
110     MBEDTLS_BYTES_TO_T_UINT_8( a, b, 0, 0, 0, 0, 0, 0 )
111 
112 #endif /* bits in mbedtls_mpi_uint */
113 
114 #if defined(MBEDTLS_HAVE_ASM)
115 
116 #ifndef asm
117 #define asm __asm
118 #endif
119 
120 /* armcc5 --gnu defines __GNUC__ but doesn't support GNU's extended asm */
121 #if defined(__GNUC__) && \
122     ( !defined(__ARMCC_VERSION) || __ARMCC_VERSION >= 6000000 )
123 
124 /*
125  * Disable use of the i386 assembly code below if option -O0, to disable all
126  * compiler optimisations, is passed, detected with __OPTIMIZE__
127  * This is done as the number of registers used in the assembly code doesn't
128  * work with the -O0 option.
129  */
130 #if defined(__i386__) && defined(__OPTIMIZE__)
131 
132 #define MULADDC_INIT                        \
133     asm(                                    \
134         "movl   %%ebx, %0           \n\t"   \
135         "movl   %5, %%esi           \n\t"   \
136         "movl   %6, %%edi           \n\t"   \
137         "movl   %7, %%ecx           \n\t"   \
138         "movl   %8, %%ebx           \n\t"
139 
140 #define MULADDC_CORE                        \
141         "lodsl                      \n\t"   \
142         "mull   %%ebx               \n\t"   \
143         "addl   %%ecx,   %%eax      \n\t"   \
144         "adcl   $0,      %%edx      \n\t"   \
145         "addl   (%%edi), %%eax      \n\t"   \
146         "adcl   $0,      %%edx      \n\t"   \
147         "movl   %%edx,   %%ecx      \n\t"   \
148         "stosl                      \n\t"
149 
150 #if defined(MBEDTLS_HAVE_SSE2)
151 
152 #define MULADDC_HUIT                            \
153         "movd     %%ecx,     %%mm1      \n\t"   \
154         "movd     %%ebx,     %%mm0      \n\t"   \
155         "movd     (%%edi),   %%mm3      \n\t"   \
156         "paddq    %%mm3,     %%mm1      \n\t"   \
157         "movd     (%%esi),   %%mm2      \n\t"   \
158         "pmuludq  %%mm0,     %%mm2      \n\t"   \
159         "movd     4(%%esi),  %%mm4      \n\t"   \
160         "pmuludq  %%mm0,     %%mm4      \n\t"   \
161         "movd     8(%%esi),  %%mm6      \n\t"   \
162         "pmuludq  %%mm0,     %%mm6      \n\t"   \
163         "movd     12(%%esi), %%mm7      \n\t"   \
164         "pmuludq  %%mm0,     %%mm7      \n\t"   \
165         "paddq    %%mm2,     %%mm1      \n\t"   \
166         "movd     4(%%edi),  %%mm3      \n\t"   \
167         "paddq    %%mm4,     %%mm3      \n\t"   \
168         "movd     8(%%edi),  %%mm5      \n\t"   \
169         "paddq    %%mm6,     %%mm5      \n\t"   \
170         "movd     12(%%edi), %%mm4      \n\t"   \
171         "paddq    %%mm4,     %%mm7      \n\t"   \
172         "movd     %%mm1,     (%%edi)    \n\t"   \
173         "movd     16(%%esi), %%mm2      \n\t"   \
174         "pmuludq  %%mm0,     %%mm2      \n\t"   \
175         "psrlq    $32,       %%mm1      \n\t"   \
176         "movd     20(%%esi), %%mm4      \n\t"   \
177         "pmuludq  %%mm0,     %%mm4      \n\t"   \
178         "paddq    %%mm3,     %%mm1      \n\t"   \
179         "movd     24(%%esi), %%mm6      \n\t"   \
180         "pmuludq  %%mm0,     %%mm6      \n\t"   \
181         "movd     %%mm1,     4(%%edi)   \n\t"   \
182         "psrlq    $32,       %%mm1      \n\t"   \
183         "movd     28(%%esi), %%mm3      \n\t"   \
184         "pmuludq  %%mm0,     %%mm3      \n\t"   \
185         "paddq    %%mm5,     %%mm1      \n\t"   \
186         "movd     16(%%edi), %%mm5      \n\t"   \
187         "paddq    %%mm5,     %%mm2      \n\t"   \
188         "movd     %%mm1,     8(%%edi)   \n\t"   \
189         "psrlq    $32,       %%mm1      \n\t"   \
190         "paddq    %%mm7,     %%mm1      \n\t"   \
191         "movd     20(%%edi), %%mm5      \n\t"   \
192         "paddq    %%mm5,     %%mm4      \n\t"   \
193         "movd     %%mm1,     12(%%edi)  \n\t"   \
194         "psrlq    $32,       %%mm1      \n\t"   \
195         "paddq    %%mm2,     %%mm1      \n\t"   \
196         "movd     24(%%edi), %%mm5      \n\t"   \
197         "paddq    %%mm5,     %%mm6      \n\t"   \
198         "movd     %%mm1,     16(%%edi)  \n\t"   \
199         "psrlq    $32,       %%mm1      \n\t"   \
200         "paddq    %%mm4,     %%mm1      \n\t"   \
201         "movd     28(%%edi), %%mm5      \n\t"   \
202         "paddq    %%mm5,     %%mm3      \n\t"   \
203         "movd     %%mm1,     20(%%edi)  \n\t"   \
204         "psrlq    $32,       %%mm1      \n\t"   \
205         "paddq    %%mm6,     %%mm1      \n\t"   \
206         "movd     %%mm1,     24(%%edi)  \n\t"   \
207         "psrlq    $32,       %%mm1      \n\t"   \
208         "paddq    %%mm3,     %%mm1      \n\t"   \
209         "movd     %%mm1,     28(%%edi)  \n\t"   \
210         "addl     $32,       %%edi      \n\t"   \
211         "addl     $32,       %%esi      \n\t"   \
212         "psrlq    $32,       %%mm1      \n\t"   \
213         "movd     %%mm1,     %%ecx      \n\t"
214 
215 #define MULADDC_STOP                    \
216         "emms                   \n\t"   \
217         "movl   %4, %%ebx       \n\t"   \
218         "movl   %%ecx, %1       \n\t"   \
219         "movl   %%edi, %2       \n\t"   \
220         "movl   %%esi, %3       \n\t"   \
221         : "=m" (t), "=m" (c), "=m" (d), "=m" (s)        \
222         : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b)   \
223         : "eax", "ebx", "ecx", "edx", "esi", "edi"      \
224     );
225 
226 #else
227 
228 #define MULADDC_STOP                    \
229         "movl   %4, %%ebx       \n\t"   \
230         "movl   %%ecx, %1       \n\t"   \
231         "movl   %%edi, %2       \n\t"   \
232         "movl   %%esi, %3       \n\t"   \
233         : "=m" (t), "=m" (c), "=m" (d), "=m" (s)        \
234         : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b)   \
235         : "eax", "ebx", "ecx", "edx", "esi", "edi"      \
236     );
237 #endif /* SSE2 */
238 #endif /* i386 */
239 
240 #if defined(__amd64__) || defined (__x86_64__)
241 
242 #define MULADDC_INIT                        \
243     asm(                                    \
244         "xorq   %%r8, %%r8\n"
245 
246 #define MULADDC_CORE                        \
247         "movq   (%%rsi), %%rax\n"           \
248         "mulq   %%rbx\n"                    \
249         "addq   $8, %%rsi\n"                \
250         "addq   %%rcx, %%rax\n"             \
251         "movq   %%r8, %%rcx\n"              \
252         "adcq   $0, %%rdx\n"                \
253         "nop    \n"                         \
254         "addq   %%rax, (%%rdi)\n"           \
255         "adcq   %%rdx, %%rcx\n"             \
256         "addq   $8, %%rdi\n"
257 
258 #define MULADDC_STOP                        \
259         : "+c" (c), "+D" (d), "+S" (s)      \
260         : "b" (b)                           \
261         : "rax", "rdx", "r8"                \
262     );
263 
264 #endif /* AMD64 */
265 
266 #if defined(__mc68020__) || defined(__mcpu32__)
267 
268 #define MULADDC_INIT                    \
269     asm(                                \
270         "movl   %3, %%a2        \n\t"   \
271         "movl   %4, %%a3        \n\t"   \
272         "movl   %5, %%d3        \n\t"   \
273         "movl   %6, %%d2        \n\t"   \
274         "moveq  #0, %%d0        \n\t"
275 
276 #define MULADDC_CORE                    \
277         "movel  %%a2@+, %%d1    \n\t"   \
278         "mulul  %%d2, %%d4:%%d1 \n\t"   \
279         "addl   %%d3, %%d1      \n\t"   \
280         "addxl  %%d0, %%d4      \n\t"   \
281         "moveq  #0,   %%d3      \n\t"   \
282         "addl   %%d1, %%a3@+    \n\t"   \
283         "addxl  %%d4, %%d3      \n\t"
284 
285 #define MULADDC_STOP                    \
286         "movl   %%d3, %0        \n\t"   \
287         "movl   %%a3, %1        \n\t"   \
288         "movl   %%a2, %2        \n\t"   \
289         : "=m" (c), "=m" (d), "=m" (s)              \
290         : "m" (s), "m" (d), "m" (c), "m" (b)        \
291         : "d0", "d1", "d2", "d3", "d4", "a2", "a3"  \
292     );
293 
294 #define MULADDC_HUIT                        \
295         "movel  %%a2@+,  %%d1       \n\t"   \
296         "mulul  %%d2,    %%d4:%%d1  \n\t"   \
297         "addxl  %%d3,    %%d1       \n\t"   \
298         "addxl  %%d0,    %%d4       \n\t"   \
299         "addl   %%d1,    %%a3@+     \n\t"   \
300         "movel  %%a2@+,  %%d1       \n\t"   \
301         "mulul  %%d2,    %%d3:%%d1  \n\t"   \
302         "addxl  %%d4,    %%d1       \n\t"   \
303         "addxl  %%d0,    %%d3       \n\t"   \
304         "addl   %%d1,    %%a3@+     \n\t"   \
305         "movel  %%a2@+,  %%d1       \n\t"   \
306         "mulul  %%d2,    %%d4:%%d1  \n\t"   \
307         "addxl  %%d3,    %%d1       \n\t"   \
308         "addxl  %%d0,    %%d4       \n\t"   \
309         "addl   %%d1,    %%a3@+     \n\t"   \
310         "movel  %%a2@+,  %%d1       \n\t"   \
311         "mulul  %%d2,    %%d3:%%d1  \n\t"   \
312         "addxl  %%d4,    %%d1       \n\t"   \
313         "addxl  %%d0,    %%d3       \n\t"   \
314         "addl   %%d1,    %%a3@+     \n\t"   \
315         "movel  %%a2@+,  %%d1       \n\t"   \
316         "mulul  %%d2,    %%d4:%%d1  \n\t"   \
317         "addxl  %%d3,    %%d1       \n\t"   \
318         "addxl  %%d0,    %%d4       \n\t"   \
319         "addl   %%d1,    %%a3@+     \n\t"   \
320         "movel  %%a2@+,  %%d1       \n\t"   \
321         "mulul  %%d2,    %%d3:%%d1  \n\t"   \
322         "addxl  %%d4,    %%d1       \n\t"   \
323         "addxl  %%d0,    %%d3       \n\t"   \
324         "addl   %%d1,    %%a3@+     \n\t"   \
325         "movel  %%a2@+,  %%d1       \n\t"   \
326         "mulul  %%d2,    %%d4:%%d1  \n\t"   \
327         "addxl  %%d3,    %%d1       \n\t"   \
328         "addxl  %%d0,    %%d4       \n\t"   \
329         "addl   %%d1,    %%a3@+     \n\t"   \
330         "movel  %%a2@+,  %%d1       \n\t"   \
331         "mulul  %%d2,    %%d3:%%d1  \n\t"   \
332         "addxl  %%d4,    %%d1       \n\t"   \
333         "addxl  %%d0,    %%d3       \n\t"   \
334         "addl   %%d1,    %%a3@+     \n\t"   \
335         "addxl  %%d0,    %%d3       \n\t"
336 
337 #endif /* MC68000 */
338 
339 #if defined(__powerpc64__) || defined(__ppc64__)
340 
341 #if defined(__MACH__) && defined(__APPLE__)
342 
343 #define MULADDC_INIT                        \
344     asm(                                    \
345         "ld     r3, %3              \n\t"   \
346         "ld     r4, %4              \n\t"   \
347         "ld     r5, %5              \n\t"   \
348         "ld     r6, %6              \n\t"   \
349         "addi   r3, r3, -8          \n\t"   \
350         "addi   r4, r4, -8          \n\t"   \
351         "addic  r5, r5,  0          \n\t"
352 
353 #define MULADDC_CORE                        \
354         "ldu    r7, 8(r3)           \n\t"   \
355         "mulld  r8, r7, r6          \n\t"   \
356         "mulhdu r9, r7, r6          \n\t"   \
357         "adde   r8, r8, r5          \n\t"   \
358         "ld     r7, 8(r4)           \n\t"   \
359         "addze  r5, r9              \n\t"   \
360         "addc   r8, r8, r7          \n\t"   \
361         "stdu   r8, 8(r4)           \n\t"
362 
363 #define MULADDC_STOP                        \
364         "addze  r5, r5              \n\t"   \
365         "addi   r4, r4, 8           \n\t"   \
366         "addi   r3, r3, 8           \n\t"   \
367         "std    r5, %0              \n\t"   \
368         "std    r4, %1              \n\t"   \
369         "std    r3, %2              \n\t"   \
370         : "=m" (c), "=m" (d), "=m" (s)              \
371         : "m" (s), "m" (d), "m" (c), "m" (b)        \
372         : "r3", "r4", "r5", "r6", "r7", "r8", "r9"  \
373     );
374 
375 
376 #else /* __MACH__ && __APPLE__ */
377 
378 #define MULADDC_INIT                        \
379     asm(                                    \
380         "ld     %%r3, %3            \n\t"   \
381         "ld     %%r4, %4            \n\t"   \
382         "ld     %%r5, %5            \n\t"   \
383         "ld     %%r6, %6            \n\t"   \
384         "addi   %%r3, %%r3, -8      \n\t"   \
385         "addi   %%r4, %%r4, -8      \n\t"   \
386         "addic  %%r5, %%r5,  0      \n\t"
387 
388 #define MULADDC_CORE                        \
389         "ldu    %%r7, 8(%%r3)       \n\t"   \
390         "mulld  %%r8, %%r7, %%r6    \n\t"   \
391         "mulhdu %%r9, %%r7, %%r6    \n\t"   \
392         "adde   %%r8, %%r8, %%r5    \n\t"   \
393         "ld     %%r7, 8(%%r4)       \n\t"   \
394         "addze  %%r5, %%r9          \n\t"   \
395         "addc   %%r8, %%r8, %%r7    \n\t"   \
396         "stdu   %%r8, 8(%%r4)       \n\t"
397 
398 #define MULADDC_STOP                        \
399         "addze  %%r5, %%r5          \n\t"   \
400         "addi   %%r4, %%r4, 8       \n\t"   \
401         "addi   %%r3, %%r3, 8       \n\t"   \
402         "std    %%r5, %0            \n\t"   \
403         "std    %%r4, %1            \n\t"   \
404         "std    %%r3, %2            \n\t"   \
405         : "=m" (c), "=m" (d), "=m" (s)              \
406         : "m" (s), "m" (d), "m" (c), "m" (b)        \
407         : "r3", "r4", "r5", "r6", "r7", "r8", "r9"  \
408     );
409 
410 #endif /* __MACH__ && __APPLE__ */
411 
412 #elif defined(__powerpc__) || defined(__ppc__) /* end PPC64/begin PPC32  */
413 
414 #if defined(__MACH__) && defined(__APPLE__)
415 
416 #define MULADDC_INIT                    \
417     asm(                                \
418         "lwz    r3, %3          \n\t"   \
419         "lwz    r4, %4          \n\t"   \
420         "lwz    r5, %5          \n\t"   \
421         "lwz    r6, %6          \n\t"   \
422         "addi   r3, r3, -4      \n\t"   \
423         "addi   r4, r4, -4      \n\t"   \
424         "addic  r5, r5,  0      \n\t"
425 
426 #define MULADDC_CORE                    \
427         "lwzu   r7, 4(r3)       \n\t"   \
428         "mullw  r8, r7, r6      \n\t"   \
429         "mulhwu r9, r7, r6      \n\t"   \
430         "adde   r8, r8, r5      \n\t"   \
431         "lwz    r7, 4(r4)       \n\t"   \
432         "addze  r5, r9          \n\t"   \
433         "addc   r8, r8, r7      \n\t"   \
434         "stwu   r8, 4(r4)       \n\t"
435 
436 #define MULADDC_STOP                    \
437         "addze  r5, r5          \n\t"   \
438         "addi   r4, r4, 4       \n\t"   \
439         "addi   r3, r3, 4       \n\t"   \
440         "stw    r5, %0          \n\t"   \
441         "stw    r4, %1          \n\t"   \
442         "stw    r3, %2          \n\t"   \
443         : "=m" (c), "=m" (d), "=m" (s)              \
444         : "m" (s), "m" (d), "m" (c), "m" (b)        \
445         : "r3", "r4", "r5", "r6", "r7", "r8", "r9"  \
446     );
447 
448 #else /* __MACH__ && __APPLE__ */
449 
450 #define MULADDC_INIT                        \
451     asm(                                    \
452         "lwz    %%r3, %3            \n\t"   \
453         "lwz    %%r4, %4            \n\t"   \
454         "lwz    %%r5, %5            \n\t"   \
455         "lwz    %%r6, %6            \n\t"   \
456         "addi   %%r3, %%r3, -4      \n\t"   \
457         "addi   %%r4, %%r4, -4      \n\t"   \
458         "addic  %%r5, %%r5,  0      \n\t"
459 
460 #define MULADDC_CORE                        \
461         "lwzu   %%r7, 4(%%r3)       \n\t"   \
462         "mullw  %%r8, %%r7, %%r6    \n\t"   \
463         "mulhwu %%r9, %%r7, %%r6    \n\t"   \
464         "adde   %%r8, %%r8, %%r5    \n\t"   \
465         "lwz    %%r7, 4(%%r4)       \n\t"   \
466         "addze  %%r5, %%r9          \n\t"   \
467         "addc   %%r8, %%r8, %%r7    \n\t"   \
468         "stwu   %%r8, 4(%%r4)       \n\t"
469 
470 #define MULADDC_STOP                        \
471         "addze  %%r5, %%r5          \n\t"   \
472         "addi   %%r4, %%r4, 4       \n\t"   \
473         "addi   %%r3, %%r3, 4       \n\t"   \
474         "stw    %%r5, %0            \n\t"   \
475         "stw    %%r4, %1            \n\t"   \
476         "stw    %%r3, %2            \n\t"   \
477         : "=m" (c), "=m" (d), "=m" (s)              \
478         : "m" (s), "m" (d), "m" (c), "m" (b)        \
479         : "r3", "r4", "r5", "r6", "r7", "r8", "r9"  \
480     );
481 
482 #endif /* __MACH__ && __APPLE__ */
483 
484 #endif /* PPC32 */
485 
486 /*
487  * The Sparc(64) assembly is reported to be broken.
488  * Disable it for now, until we're able to fix it.
489  */
490 #if 0 && defined(__sparc__)
491 #if defined(__sparc64__)
492 
493 #define MULADDC_INIT                                    \
494     asm(                                                \
495                 "ldx     %3, %%o0               \n\t"   \
496                 "ldx     %4, %%o1               \n\t"   \
497                 "ld      %5, %%o2               \n\t"   \
498                 "ld      %6, %%o3               \n\t"
499 
500 #define MULADDC_CORE                                    \
501                 "ld      [%%o0], %%o4           \n\t"   \
502                 "inc     4, %%o0                \n\t"   \
503                 "ld      [%%o1], %%o5           \n\t"   \
504                 "umul    %%o3, %%o4, %%o4       \n\t"   \
505                 "addcc   %%o4, %%o2, %%o4       \n\t"   \
506                 "rd      %%y, %%g1              \n\t"   \
507                 "addx    %%g1, 0, %%g1          \n\t"   \
508                 "addcc   %%o4, %%o5, %%o4       \n\t"   \
509                 "st      %%o4, [%%o1]           \n\t"   \
510                 "addx    %%g1, 0, %%o2          \n\t"   \
511                 "inc     4, %%o1                \n\t"
512 
513         #define MULADDC_STOP                            \
514                 "st      %%o2, %0               \n\t"   \
515                 "stx     %%o1, %1               \n\t"   \
516                 "stx     %%o0, %2               \n\t"   \
517         : "=m" (c), "=m" (d), "=m" (s)          \
518         : "m" (s), "m" (d), "m" (c), "m" (b)    \
519         : "g1", "o0", "o1", "o2", "o3", "o4",   \
520           "o5"                                  \
521         );
522 
523 #else /* __sparc64__ */
524 
525 #define MULADDC_INIT                                    \
526     asm(                                                \
527                 "ld      %3, %%o0               \n\t"   \
528                 "ld      %4, %%o1               \n\t"   \
529                 "ld      %5, %%o2               \n\t"   \
530                 "ld      %6, %%o3               \n\t"
531 
532 #define MULADDC_CORE                                    \
533                 "ld      [%%o0], %%o4           \n\t"   \
534                 "inc     4, %%o0                \n\t"   \
535                 "ld      [%%o1], %%o5           \n\t"   \
536                 "umul    %%o3, %%o4, %%o4       \n\t"   \
537                 "addcc   %%o4, %%o2, %%o4       \n\t"   \
538                 "rd      %%y, %%g1              \n\t"   \
539                 "addx    %%g1, 0, %%g1          \n\t"   \
540                 "addcc   %%o4, %%o5, %%o4       \n\t"   \
541                 "st      %%o4, [%%o1]           \n\t"   \
542                 "addx    %%g1, 0, %%o2          \n\t"   \
543                 "inc     4, %%o1                \n\t"
544 
545 #define MULADDC_STOP                                    \
546                 "st      %%o2, %0               \n\t"   \
547                 "st      %%o1, %1               \n\t"   \
548                 "st      %%o0, %2               \n\t"   \
549         : "=m" (c), "=m" (d), "=m" (s)          \
550         : "m" (s), "m" (d), "m" (c), "m" (b)    \
551         : "g1", "o0", "o1", "o2", "o3", "o4",   \
552           "o5"                                  \
553         );
554 
555 #endif /* __sparc64__ */
556 #endif /* __sparc__ */
557 
558 #if defined(__microblaze__) || defined(microblaze)
559 
560 #define MULADDC_INIT                    \
561     asm(                                \
562         "lwi   r3,   %3         \n\t"   \
563         "lwi   r4,   %4         \n\t"   \
564         "lwi   r5,   %5         \n\t"   \
565         "lwi   r6,   %6         \n\t"   \
566         "andi  r7,   r6, 0xffff \n\t"   \
567         "bsrli r6,   r6, 16     \n\t"
568 
569 #define MULADDC_CORE                    \
570         "lhui  r8,   r3,   0    \n\t"   \
571         "addi  r3,   r3,   2    \n\t"   \
572         "lhui  r9,   r3,   0    \n\t"   \
573         "addi  r3,   r3,   2    \n\t"   \
574         "mul   r10,  r9,  r6    \n\t"   \
575         "mul   r11,  r8,  r7    \n\t"   \
576         "mul   r12,  r9,  r7    \n\t"   \
577         "mul   r13,  r8,  r6    \n\t"   \
578         "bsrli  r8, r10,  16    \n\t"   \
579         "bsrli  r9, r11,  16    \n\t"   \
580         "add   r13, r13,  r8    \n\t"   \
581         "add   r13, r13,  r9    \n\t"   \
582         "bslli r10, r10,  16    \n\t"   \
583         "bslli r11, r11,  16    \n\t"   \
584         "add   r12, r12, r10    \n\t"   \
585         "addc  r13, r13,  r0    \n\t"   \
586         "add   r12, r12, r11    \n\t"   \
587         "addc  r13, r13,  r0    \n\t"   \
588         "lwi   r10,  r4,   0    \n\t"   \
589         "add   r12, r12, r10    \n\t"   \
590         "addc  r13, r13,  r0    \n\t"   \
591         "add   r12, r12,  r5    \n\t"   \
592         "addc   r5, r13,  r0    \n\t"   \
593         "swi   r12,  r4,   0    \n\t"   \
594         "addi   r4,  r4,   4    \n\t"
595 
596 #define MULADDC_STOP                    \
597         "swi   r5,   %0         \n\t"   \
598         "swi   r4,   %1         \n\t"   \
599         "swi   r3,   %2         \n\t"   \
600         : "=m" (c), "=m" (d), "=m" (s)              \
601         : "m" (s), "m" (d), "m" (c), "m" (b)        \
602         : "r3", "r4", "r5", "r6", "r7", "r8",       \
603           "r9", "r10", "r11", "r12", "r13"          \
604     );
605 
606 #endif /* MicroBlaze */
607 
608 #if defined(__tricore__)
609 
610 #define MULADDC_INIT                            \
611     asm(                                        \
612         "ld.a   %%a2, %3                \n\t"   \
613         "ld.a   %%a3, %4                \n\t"   \
614         "ld.w   %%d4, %5                \n\t"   \
615         "ld.w   %%d1, %6                \n\t"   \
616         "xor    %%d5, %%d5              \n\t"
617 
618 #define MULADDC_CORE                            \
619         "ld.w   %%d0,   [%%a2+]         \n\t"   \
620         "madd.u %%e2, %%e4, %%d0, %%d1  \n\t"   \
621         "ld.w   %%d0,   [%%a3]          \n\t"   \
622         "addx   %%d2,    %%d2,  %%d0    \n\t"   \
623         "addc   %%d3,    %%d3,    0     \n\t"   \
624         "mov    %%d4,    %%d3           \n\t"   \
625         "st.w  [%%a3+],  %%d2           \n\t"
626 
627 #define MULADDC_STOP                            \
628         "st.w   %0, %%d4                \n\t"   \
629         "st.a   %1, %%a3                \n\t"   \
630         "st.a   %2, %%a2                \n\t"   \
631         : "=m" (c), "=m" (d), "=m" (s)          \
632         : "m" (s), "m" (d), "m" (c), "m" (b)    \
633         : "d0", "d1", "e2", "d4", "a2", "a3"    \
634     );
635 
636 #endif /* TriCore */
637 
638 /*
639  * Note, gcc -O0 by default uses r7 for the frame pointer, so it complains about
640  * our use of r7 below, unless -fomit-frame-pointer is passed.
641  *
642  * On the other hand, -fomit-frame-pointer is implied by any -Ox options with
643  * x !=0, which we can detect using __OPTIMIZE__ (which is also defined by
644  * clang and armcc5 under the same conditions).
645  *
646  * So, only use the optimized assembly below for optimized build, which avoids
647  * the build error and is pretty reasonable anyway.
648  */
649 #if defined(__GNUC__) && !defined(__OPTIMIZE__)
650 #define MULADDC_CANNOT_USE_R7
651 #endif
652 
653 #if defined(__arm__) && !defined(MULADDC_CANNOT_USE_R7)
654 
655 #if defined(__thumb__) && !defined(__thumb2__)
656 
657 #define MULADDC_INIT                                    \
658     asm(                                                \
659             "ldr    r0, %3                      \n\t"   \
660             "ldr    r1, %4                      \n\t"   \
661             "ldr    r2, %5                      \n\t"   \
662             "ldr    r3, %6                      \n\t"   \
663             "lsr    r7, r3, #16                 \n\t"   \
664             "mov    r9, r7                      \n\t"   \
665             "lsl    r7, r3, #16                 \n\t"   \
666             "lsr    r7, r7, #16                 \n\t"   \
667             "mov    r8, r7                      \n\t"
668 
669 #define MULADDC_CORE                                    \
670             "ldmia  r0!, {r6}                   \n\t"   \
671             "lsr    r7, r6, #16                 \n\t"   \
672             "lsl    r6, r6, #16                 \n\t"   \
673             "lsr    r6, r6, #16                 \n\t"   \
674             "mov    r4, r8                      \n\t"   \
675             "mul    r4, r6                      \n\t"   \
676             "mov    r3, r9                      \n\t"   \
677             "mul    r6, r3                      \n\t"   \
678             "mov    r5, r9                      \n\t"   \
679             "mul    r5, r7                      \n\t"   \
680             "mov    r3, r8                      \n\t"   \
681             "mul    r7, r3                      \n\t"   \
682             "lsr    r3, r6, #16                 \n\t"   \
683             "add    r5, r5, r3                  \n\t"   \
684             "lsr    r3, r7, #16                 \n\t"   \
685             "add    r5, r5, r3                  \n\t"   \
686             "add    r4, r4, r2                  \n\t"   \
687             "mov    r2, #0                      \n\t"   \
688             "adc    r5, r2                      \n\t"   \
689             "lsl    r3, r6, #16                 \n\t"   \
690             "add    r4, r4, r3                  \n\t"   \
691             "adc    r5, r2                      \n\t"   \
692             "lsl    r3, r7, #16                 \n\t"   \
693             "add    r4, r4, r3                  \n\t"   \
694             "adc    r5, r2                      \n\t"   \
695             "ldr    r3, [r1]                    \n\t"   \
696             "add    r4, r4, r3                  \n\t"   \
697             "adc    r2, r5                      \n\t"   \
698             "stmia  r1!, {r4}                   \n\t"
699 
700 #define MULADDC_STOP                                    \
701             "str    r2, %0                      \n\t"   \
702             "str    r1, %1                      \n\t"   \
703             "str    r0, %2                      \n\t"   \
704          : "=m" (c),  "=m" (d), "=m" (s)        \
705          : "m" (s), "m" (d), "m" (c), "m" (b)   \
706          : "r0", "r1", "r2", "r3", "r4", "r5",  \
707            "r6", "r7", "r8", "r9", "cc"         \
708          );
709 
710 #elif (__ARM_ARCH >= 6) && \
711     defined (__ARM_FEATURE_DSP) && (__ARM_FEATURE_DSP == 1)
712 
713 #define MULADDC_INIT                            \
714     asm(
715 
716 #define MULADDC_CORE                            \
717             "ldr    r0, [%0], #4        \n\t"   \
718             "ldr    r1, [%1]            \n\t"   \
719             "umaal  r1, %2, %3, r0      \n\t"   \
720             "str    r1, [%1], #4        \n\t"
721 
722 #define MULADDC_STOP                            \
723          : "=r" (s),  "=r" (d), "=r" (c)        \
724          : "r" (b), "0" (s), "1" (d), "2" (c)   \
725          : "r0", "r1", "memory"                 \
726          );
727 
728 #else
729 
730 #define MULADDC_INIT                                    \
731     asm(                                                \
732             "ldr    r0, %3                      \n\t"   \
733             "ldr    r1, %4                      \n\t"   \
734             "ldr    r2, %5                      \n\t"   \
735             "ldr    r3, %6                      \n\t"
736 
737 #define MULADDC_CORE                                    \
738             "ldr    r4, [r0], #4                \n\t"   \
739             "mov    r5, #0                      \n\t"   \
740             "ldr    r6, [r1]                    \n\t"   \
741             "umlal  r2, r5, r3, r4              \n\t"   \
742             "adds   r7, r6, r2                  \n\t"   \
743             "adc    r2, r5, #0                  \n\t"   \
744             "str    r7, [r1], #4                \n\t"
745 
746 #define MULADDC_STOP                                    \
747             "str    r2, %0                      \n\t"   \
748             "str    r1, %1                      \n\t"   \
749             "str    r0, %2                      \n\t"   \
750          : "=m" (c),  "=m" (d), "=m" (s)        \
751          : "m" (s), "m" (d), "m" (c), "m" (b)   \
752          : "r0", "r1", "r2", "r3", "r4", "r5",  \
753            "r6", "r7", "cc"                     \
754          );
755 
756 #endif /* Thumb */
757 
758 #endif /* ARMv3 */
759 
760 #if defined(__alpha__)
761 
762 #define MULADDC_INIT                    \
763     asm(                                \
764         "ldq    $1, %3          \n\t"   \
765         "ldq    $2, %4          \n\t"   \
766         "ldq    $3, %5          \n\t"   \
767         "ldq    $4, %6          \n\t"
768 
769 #define MULADDC_CORE                    \
770         "ldq    $6,  0($1)      \n\t"   \
771         "addq   $1,  8, $1      \n\t"   \
772         "mulq   $6, $4, $7      \n\t"   \
773         "umulh  $6, $4, $6      \n\t"   \
774         "addq   $7, $3, $7      \n\t"   \
775         "cmpult $7, $3, $3      \n\t"   \
776         "ldq    $5,  0($2)      \n\t"   \
777         "addq   $7, $5, $7      \n\t"   \
778         "cmpult $7, $5, $5      \n\t"   \
779         "stq    $7,  0($2)      \n\t"   \
780         "addq   $2,  8, $2      \n\t"   \
781         "addq   $6, $3, $3      \n\t"   \
782         "addq   $5, $3, $3      \n\t"
783 
784 #define MULADDC_STOP                                    \
785         "stq    $3, %0          \n\t"   \
786         "stq    $2, %1          \n\t"   \
787         "stq    $1, %2          \n\t"   \
788         : "=m" (c), "=m" (d), "=m" (s)              \
789         : "m" (s), "m" (d), "m" (c), "m" (b)        \
790         : "$1", "$2", "$3", "$4", "$5", "$6", "$7"  \
791     );
792 #endif /* Alpha */
793 
794 #if defined(__mips__) && !defined(__mips64)
795 
796 #define MULADDC_INIT                    \
797     asm(                                \
798         "lw     $10, %3         \n\t"   \
799         "lw     $11, %4         \n\t"   \
800         "lw     $12, %5         \n\t"   \
801         "lw     $13, %6         \n\t"
802 
803 #define MULADDC_CORE                    \
804         "lw     $14, 0($10)     \n\t"   \
805         "multu  $13, $14        \n\t"   \
806         "addi   $10, $10, 4     \n\t"   \
807         "mflo   $14             \n\t"   \
808         "mfhi   $9              \n\t"   \
809         "addu   $14, $12, $14   \n\t"   \
810         "lw     $15, 0($11)     \n\t"   \
811         "sltu   $12, $14, $12   \n\t"   \
812         "addu   $15, $14, $15   \n\t"   \
813         "sltu   $14, $15, $14   \n\t"   \
814         "addu   $12, $12, $9    \n\t"   \
815         "sw     $15, 0($11)     \n\t"   \
816         "addu   $12, $12, $14   \n\t"   \
817         "addi   $11, $11, 4     \n\t"
818 
819 #define MULADDC_STOP                    \
820         "sw     $12, %0         \n\t"   \
821         "sw     $11, %1         \n\t"   \
822         "sw     $10, %2         \n\t"   \
823         : "=m" (c), "=m" (d), "=m" (s)                      \
824         : "m" (s), "m" (d), "m" (c), "m" (b)                \
825         : "$9", "$10", "$11", "$12", "$13", "$14", "$15", "lo", "hi" \
826     );
827 
828 #endif /* MIPS */
829 #endif /* GNUC */
830 
831 #if (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__)
832 
833 #define MULADDC_INIT                            \
834     __asm   mov     esi, s                      \
835     __asm   mov     edi, d                      \
836     __asm   mov     ecx, c                      \
837     __asm   mov     ebx, b
838 
839 #define MULADDC_CORE                            \
840     __asm   lodsd                               \
841     __asm   mul     ebx                         \
842     __asm   add     eax, ecx                    \
843     __asm   adc     edx, 0                      \
844     __asm   add     eax, [edi]                  \
845     __asm   adc     edx, 0                      \
846     __asm   mov     ecx, edx                    \
847     __asm   stosd
848 
849 #if defined(MBEDTLS_HAVE_SSE2)
850 
851 #define EMIT __asm _emit
852 
853 #define MULADDC_HUIT                            \
854     EMIT 0x0F  EMIT 0x6E  EMIT 0xC9             \
855     EMIT 0x0F  EMIT 0x6E  EMIT 0xC3             \
856     EMIT 0x0F  EMIT 0x6E  EMIT 0x1F             \
857     EMIT 0x0F  EMIT 0xD4  EMIT 0xCB             \
858     EMIT 0x0F  EMIT 0x6E  EMIT 0x16             \
859     EMIT 0x0F  EMIT 0xF4  EMIT 0xD0             \
860     EMIT 0x0F  EMIT 0x6E  EMIT 0x66  EMIT 0x04  \
861     EMIT 0x0F  EMIT 0xF4  EMIT 0xE0             \
862     EMIT 0x0F  EMIT 0x6E  EMIT 0x76  EMIT 0x08  \
863     EMIT 0x0F  EMIT 0xF4  EMIT 0xF0             \
864     EMIT 0x0F  EMIT 0x6E  EMIT 0x7E  EMIT 0x0C  \
865     EMIT 0x0F  EMIT 0xF4  EMIT 0xF8             \
866     EMIT 0x0F  EMIT 0xD4  EMIT 0xCA             \
867     EMIT 0x0F  EMIT 0x6E  EMIT 0x5F  EMIT 0x04  \
868     EMIT 0x0F  EMIT 0xD4  EMIT 0xDC             \
869     EMIT 0x0F  EMIT 0x6E  EMIT 0x6F  EMIT 0x08  \
870     EMIT 0x0F  EMIT 0xD4  EMIT 0xEE             \
871     EMIT 0x0F  EMIT 0x6E  EMIT 0x67  EMIT 0x0C  \
872     EMIT 0x0F  EMIT 0xD4  EMIT 0xFC             \
873     EMIT 0x0F  EMIT 0x7E  EMIT 0x0F             \
874     EMIT 0x0F  EMIT 0x6E  EMIT 0x56  EMIT 0x10  \
875     EMIT 0x0F  EMIT 0xF4  EMIT 0xD0             \
876     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
877     EMIT 0x0F  EMIT 0x6E  EMIT 0x66  EMIT 0x14  \
878     EMIT 0x0F  EMIT 0xF4  EMIT 0xE0             \
879     EMIT 0x0F  EMIT 0xD4  EMIT 0xCB             \
880     EMIT 0x0F  EMIT 0x6E  EMIT 0x76  EMIT 0x18  \
881     EMIT 0x0F  EMIT 0xF4  EMIT 0xF0             \
882     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x04  \
883     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
884     EMIT 0x0F  EMIT 0x6E  EMIT 0x5E  EMIT 0x1C  \
885     EMIT 0x0F  EMIT 0xF4  EMIT 0xD8             \
886     EMIT 0x0F  EMIT 0xD4  EMIT 0xCD             \
887     EMIT 0x0F  EMIT 0x6E  EMIT 0x6F  EMIT 0x10  \
888     EMIT 0x0F  EMIT 0xD4  EMIT 0xD5             \
889     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x08  \
890     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
891     EMIT 0x0F  EMIT 0xD4  EMIT 0xCF             \
892     EMIT 0x0F  EMIT 0x6E  EMIT 0x6F  EMIT 0x14  \
893     EMIT 0x0F  EMIT 0xD4  EMIT 0xE5             \
894     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x0C  \
895     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
896     EMIT 0x0F  EMIT 0xD4  EMIT 0xCA             \
897     EMIT 0x0F  EMIT 0x6E  EMIT 0x6F  EMIT 0x18  \
898     EMIT 0x0F  EMIT 0xD4  EMIT 0xF5             \
899     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x10  \
900     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
901     EMIT 0x0F  EMIT 0xD4  EMIT 0xCC             \
902     EMIT 0x0F  EMIT 0x6E  EMIT 0x6F  EMIT 0x1C  \
903     EMIT 0x0F  EMIT 0xD4  EMIT 0xDD             \
904     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x14  \
905     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
906     EMIT 0x0F  EMIT 0xD4  EMIT 0xCE             \
907     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x18  \
908     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
909     EMIT 0x0F  EMIT 0xD4  EMIT 0xCB             \
910     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x1C  \
911     EMIT 0x83  EMIT 0xC7  EMIT 0x20             \
912     EMIT 0x83  EMIT 0xC6  EMIT 0x20             \
913     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
914     EMIT 0x0F  EMIT 0x7E  EMIT 0xC9
915 
916 #define MULADDC_STOP                            \
917     EMIT 0x0F  EMIT 0x77                        \
918     __asm   mov     c, ecx                      \
919     __asm   mov     d, edi                      \
920     __asm   mov     s, esi                      \
921 
922 #else
923 
924 #define MULADDC_STOP                            \
925     __asm   mov     c, ecx                      \
926     __asm   mov     d, edi                      \
927     __asm   mov     s, esi                      \
928 
929 #endif /* SSE2 */
930 #endif /* MSVC */
931 
932 #endif /* MBEDTLS_HAVE_ASM */
933 
934 #if !defined(MULADDC_CORE)
935 #if defined(MBEDTLS_HAVE_UDBL)
936 
937 #define MULADDC_INIT                    \
938 {                                       \
939     mbedtls_t_udbl r;                           \
940     mbedtls_mpi_uint r0, r1;
941 
942 #define MULADDC_CORE                    \
943     r   = *(s++) * (mbedtls_t_udbl) b;          \
944     r0  = (mbedtls_mpi_uint) r;                   \
945     r1  = (mbedtls_mpi_uint)( r >> biL );         \
946     r0 += c;  r1 += (r0 <  c);          \
947     r0 += *d; r1 += (r0 < *d);          \
948     c = r1; *(d++) = r0;
949 
950 #define MULADDC_STOP                    \
951 }
952 
953 #else
954 #define MULADDC_INIT                    \
955 {                                       \
956     mbedtls_mpi_uint s0, s1, b0, b1;              \
957     mbedtls_mpi_uint r0, r1, rx, ry;              \
958     b0 = ( b << biH ) >> biH;           \
959     b1 = ( b >> biH );
960 
961 #define MULADDC_CORE                    \
962     s0 = ( *s << biH ) >> biH;          \
963     s1 = ( *s >> biH ); s++;            \
964     rx = s0 * b1; r0 = s0 * b0;         \
965     ry = s1 * b0; r1 = s1 * b1;         \
966     r1 += ( rx >> biH );                \
967     r1 += ( ry >> biH );                \
968     rx <<= biH; ry <<= biH;             \
969     r0 += rx; r1 += (r0 < rx);          \
970     r0 += ry; r1 += (r0 < ry);          \
971     r0 +=  c; r1 += (r0 <  c);          \
972     r0 += *d; r1 += (r0 < *d);          \
973     c = r1; *(d++) = r0;
974 
975 #define MULADDC_STOP                    \
976 }
977 
978 #endif /* C (generic)  */
979 #endif /* C (longlong) */
980 
981 #endif /* bn_mul.h */
982