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 
48 /*
49  * Conversion macros for embedded constants:
50  * build lists of mbedtls_mpi_uint's from lists of unsigned char's grouped by 8, 4 or 2
51  */
52 #if defined(MBEDTLS_HAVE_INT32)
53 
54 #define MBEDTLS_BYTES_TO_T_UINT_4( a, b, c, d )               \
55     ( (mbedtls_mpi_uint) (a) <<  0 ) |                        \
56     ( (mbedtls_mpi_uint) (b) <<  8 ) |                        \
57     ( (mbedtls_mpi_uint) (c) << 16 ) |                        \
58     ( (mbedtls_mpi_uint) (d) << 24 )
59 
60 #define MBEDTLS_BYTES_TO_T_UINT_2( a, b )                   \
61     MBEDTLS_BYTES_TO_T_UINT_4( a, b, 0, 0 )
62 
63 #define MBEDTLS_BYTES_TO_T_UINT_8( a, b, c, d, e, f, g, h ) \
64     MBEDTLS_BYTES_TO_T_UINT_4( a, b, c, d ),                \
65     MBEDTLS_BYTES_TO_T_UINT_4( e, f, g, h )
66 
67 #else /* 64-bits */
68 
69 #define MBEDTLS_BYTES_TO_T_UINT_8( a, b, c, d, e, f, g, h )   \
70     ( (mbedtls_mpi_uint) (a) <<  0 ) |                        \
71     ( (mbedtls_mpi_uint) (b) <<  8 ) |                        \
72     ( (mbedtls_mpi_uint) (c) << 16 ) |                        \
73     ( (mbedtls_mpi_uint) (d) << 24 ) |                        \
74     ( (mbedtls_mpi_uint) (e) << 32 ) |                        \
75     ( (mbedtls_mpi_uint) (f) << 40 ) |                        \
76     ( (mbedtls_mpi_uint) (g) << 48 ) |                        \
77     ( (mbedtls_mpi_uint) (h) << 56 )
78 
79 #define MBEDTLS_BYTES_TO_T_UINT_4( a, b, c, d )             \
80     MBEDTLS_BYTES_TO_T_UINT_8( a, b, c, d, 0, 0, 0, 0 )
81 
82 #define MBEDTLS_BYTES_TO_T_UINT_2( a, b )                   \
83     MBEDTLS_BYTES_TO_T_UINT_8( a, b, 0, 0, 0, 0, 0, 0 )
84 
85 #endif /* bits in mbedtls_mpi_uint */
86 
87 #if defined(MBEDTLS_HAVE_ASM)
88 
89 #ifndef asm
90 #define asm __asm
91 #endif
92 
93 /* armcc5 --gnu defines __GNUC__ but doesn't support GNU's extended asm */
94 #if defined(__GNUC__) && \
95     ( !defined(__ARMCC_VERSION) || __ARMCC_VERSION >= 6000000 )
96 
97 /*
98  * Disable use of the i386 assembly code below if option -O0, to disable all
99  * compiler optimisations, is passed, detected with __OPTIMIZE__
100  * This is done as the number of registers used in the assembly code doesn't
101  * work with the -O0 option.
102  */
103 #if defined(__i386__) && defined(__OPTIMIZE__)
104 
105 #define MULADDC_INIT                        \
106     asm(                                    \
107         "movl   %%ebx, %0           \n\t"   \
108         "movl   %5, %%esi           \n\t"   \
109         "movl   %6, %%edi           \n\t"   \
110         "movl   %7, %%ecx           \n\t"   \
111         "movl   %8, %%ebx           \n\t"
112 
113 #define MULADDC_CORE                        \
114         "lodsl                      \n\t"   \
115         "mull   %%ebx               \n\t"   \
116         "addl   %%ecx,   %%eax      \n\t"   \
117         "adcl   $0,      %%edx      \n\t"   \
118         "addl   (%%edi), %%eax      \n\t"   \
119         "adcl   $0,      %%edx      \n\t"   \
120         "movl   %%edx,   %%ecx      \n\t"   \
121         "stosl                      \n\t"
122 
123 #if defined(MBEDTLS_HAVE_SSE2)
124 
125 #define MULADDC_HUIT                            \
126         "movd     %%ecx,     %%mm1      \n\t"   \
127         "movd     %%ebx,     %%mm0      \n\t"   \
128         "movd     (%%edi),   %%mm3      \n\t"   \
129         "paddq    %%mm3,     %%mm1      \n\t"   \
130         "movd     (%%esi),   %%mm2      \n\t"   \
131         "pmuludq  %%mm0,     %%mm2      \n\t"   \
132         "movd     4(%%esi),  %%mm4      \n\t"   \
133         "pmuludq  %%mm0,     %%mm4      \n\t"   \
134         "movd     8(%%esi),  %%mm6      \n\t"   \
135         "pmuludq  %%mm0,     %%mm6      \n\t"   \
136         "movd     12(%%esi), %%mm7      \n\t"   \
137         "pmuludq  %%mm0,     %%mm7      \n\t"   \
138         "paddq    %%mm2,     %%mm1      \n\t"   \
139         "movd     4(%%edi),  %%mm3      \n\t"   \
140         "paddq    %%mm4,     %%mm3      \n\t"   \
141         "movd     8(%%edi),  %%mm5      \n\t"   \
142         "paddq    %%mm6,     %%mm5      \n\t"   \
143         "movd     12(%%edi), %%mm4      \n\t"   \
144         "paddq    %%mm4,     %%mm7      \n\t"   \
145         "movd     %%mm1,     (%%edi)    \n\t"   \
146         "movd     16(%%esi), %%mm2      \n\t"   \
147         "pmuludq  %%mm0,     %%mm2      \n\t"   \
148         "psrlq    $32,       %%mm1      \n\t"   \
149         "movd     20(%%esi), %%mm4      \n\t"   \
150         "pmuludq  %%mm0,     %%mm4      \n\t"   \
151         "paddq    %%mm3,     %%mm1      \n\t"   \
152         "movd     24(%%esi), %%mm6      \n\t"   \
153         "pmuludq  %%mm0,     %%mm6      \n\t"   \
154         "movd     %%mm1,     4(%%edi)   \n\t"   \
155         "psrlq    $32,       %%mm1      \n\t"   \
156         "movd     28(%%esi), %%mm3      \n\t"   \
157         "pmuludq  %%mm0,     %%mm3      \n\t"   \
158         "paddq    %%mm5,     %%mm1      \n\t"   \
159         "movd     16(%%edi), %%mm5      \n\t"   \
160         "paddq    %%mm5,     %%mm2      \n\t"   \
161         "movd     %%mm1,     8(%%edi)   \n\t"   \
162         "psrlq    $32,       %%mm1      \n\t"   \
163         "paddq    %%mm7,     %%mm1      \n\t"   \
164         "movd     20(%%edi), %%mm5      \n\t"   \
165         "paddq    %%mm5,     %%mm4      \n\t"   \
166         "movd     %%mm1,     12(%%edi)  \n\t"   \
167         "psrlq    $32,       %%mm1      \n\t"   \
168         "paddq    %%mm2,     %%mm1      \n\t"   \
169         "movd     24(%%edi), %%mm5      \n\t"   \
170         "paddq    %%mm5,     %%mm6      \n\t"   \
171         "movd     %%mm1,     16(%%edi)  \n\t"   \
172         "psrlq    $32,       %%mm1      \n\t"   \
173         "paddq    %%mm4,     %%mm1      \n\t"   \
174         "movd     28(%%edi), %%mm5      \n\t"   \
175         "paddq    %%mm5,     %%mm3      \n\t"   \
176         "movd     %%mm1,     20(%%edi)  \n\t"   \
177         "psrlq    $32,       %%mm1      \n\t"   \
178         "paddq    %%mm6,     %%mm1      \n\t"   \
179         "movd     %%mm1,     24(%%edi)  \n\t"   \
180         "psrlq    $32,       %%mm1      \n\t"   \
181         "paddq    %%mm3,     %%mm1      \n\t"   \
182         "movd     %%mm1,     28(%%edi)  \n\t"   \
183         "addl     $32,       %%edi      \n\t"   \
184         "addl     $32,       %%esi      \n\t"   \
185         "psrlq    $32,       %%mm1      \n\t"   \
186         "movd     %%mm1,     %%ecx      \n\t"
187 
188 #define MULADDC_STOP                    \
189         "emms                   \n\t"   \
190         "movl   %4, %%ebx       \n\t"   \
191         "movl   %%ecx, %1       \n\t"   \
192         "movl   %%edi, %2       \n\t"   \
193         "movl   %%esi, %3       \n\t"   \
194         : "=m" (t), "=m" (c), "=m" (d), "=m" (s)        \
195         : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b)   \
196         : "eax", "ebx", "ecx", "edx", "esi", "edi"      \
197     );
198 
199 #else
200 
201 #define MULADDC_STOP                    \
202         "movl   %4, %%ebx       \n\t"   \
203         "movl   %%ecx, %1       \n\t"   \
204         "movl   %%edi, %2       \n\t"   \
205         "movl   %%esi, %3       \n\t"   \
206         : "=m" (t), "=m" (c), "=m" (d), "=m" (s)        \
207         : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b)   \
208         : "eax", "ebx", "ecx", "edx", "esi", "edi"      \
209     );
210 #endif /* SSE2 */
211 #endif /* i386 */
212 
213 #if defined(__amd64__) || defined (__x86_64__)
214 
215 #define MULADDC_INIT                        \
216     asm(                                    \
217         "xorq   %%r8, %%r8\n"
218 
219 #define MULADDC_CORE                        \
220         "movq   (%%rsi), %%rax\n"           \
221         "mulq   %%rbx\n"                    \
222         "addq   $8, %%rsi\n"                \
223         "addq   %%rcx, %%rax\n"             \
224         "movq   %%r8, %%rcx\n"              \
225         "adcq   $0, %%rdx\n"                \
226         "nop    \n"                         \
227         "addq   %%rax, (%%rdi)\n"           \
228         "adcq   %%rdx, %%rcx\n"             \
229         "addq   $8, %%rdi\n"
230 
231 #define MULADDC_STOP                        \
232         : "+c" (c), "+D" (d), "+S" (s)      \
233         : "b" (b)                           \
234         : "rax", "rdx", "r8"                \
235     );
236 
237 #endif /* AMD64 */
238 
239 #if defined(__aarch64__)
240 
241 #define MULADDC_INIT                \
242     asm(
243 
244 #define MULADDC_CORE                \
245         "ldr x4, [%2], #8   \n\t"   \
246         "ldr x5, [%1]       \n\t"   \
247         "mul x6, x4, %3     \n\t"   \
248         "umulh x7, x4, %3   \n\t"   \
249         "adds x5, x5, x6    \n\t"   \
250         "adc x7, x7, xzr    \n\t"   \
251         "adds x5, x5, %0    \n\t"   \
252         "adc %0, x7, xzr    \n\t"   \
253         "str x5, [%1], #8   \n\t"
254 
255 #define MULADDC_STOP                        \
256          : "+r" (c),  "+r" (d), "+r" (s)    \
257          : "r" (b)                          \
258          : "x4", "x5", "x6", "x7", "cc"     \
259     );
260 
261 #endif /* Aarch64 */
262 
263 #if defined(__mc68020__) || defined(__mcpu32__)
264 
265 #define MULADDC_INIT                    \
266     asm(                                \
267         "movl   %3, %%a2        \n\t"   \
268         "movl   %4, %%a3        \n\t"   \
269         "movl   %5, %%d3        \n\t"   \
270         "movl   %6, %%d2        \n\t"   \
271         "moveq  #0, %%d0        \n\t"
272 
273 #define MULADDC_CORE                    \
274         "movel  %%a2@+, %%d1    \n\t"   \
275         "mulul  %%d2, %%d4:%%d1 \n\t"   \
276         "addl   %%d3, %%d1      \n\t"   \
277         "addxl  %%d0, %%d4      \n\t"   \
278         "moveq  #0,   %%d3      \n\t"   \
279         "addl   %%d1, %%a3@+    \n\t"   \
280         "addxl  %%d4, %%d3      \n\t"
281 
282 #define MULADDC_STOP                    \
283         "movl   %%d3, %0        \n\t"   \
284         "movl   %%a3, %1        \n\t"   \
285         "movl   %%a2, %2        \n\t"   \
286         : "=m" (c), "=m" (d), "=m" (s)              \
287         : "m" (s), "m" (d), "m" (c), "m" (b)        \
288         : "d0", "d1", "d2", "d3", "d4", "a2", "a3"  \
289     );
290 
291 #define MULADDC_HUIT                        \
292         "movel  %%a2@+,  %%d1       \n\t"   \
293         "mulul  %%d2,    %%d4:%%d1  \n\t"   \
294         "addxl  %%d3,    %%d1       \n\t"   \
295         "addxl  %%d0,    %%d4       \n\t"   \
296         "addl   %%d1,    %%a3@+     \n\t"   \
297         "movel  %%a2@+,  %%d1       \n\t"   \
298         "mulul  %%d2,    %%d3:%%d1  \n\t"   \
299         "addxl  %%d4,    %%d1       \n\t"   \
300         "addxl  %%d0,    %%d3       \n\t"   \
301         "addl   %%d1,    %%a3@+     \n\t"   \
302         "movel  %%a2@+,  %%d1       \n\t"   \
303         "mulul  %%d2,    %%d4:%%d1  \n\t"   \
304         "addxl  %%d3,    %%d1       \n\t"   \
305         "addxl  %%d0,    %%d4       \n\t"   \
306         "addl   %%d1,    %%a3@+     \n\t"   \
307         "movel  %%a2@+,  %%d1       \n\t"   \
308         "mulul  %%d2,    %%d3:%%d1  \n\t"   \
309         "addxl  %%d4,    %%d1       \n\t"   \
310         "addxl  %%d0,    %%d3       \n\t"   \
311         "addl   %%d1,    %%a3@+     \n\t"   \
312         "movel  %%a2@+,  %%d1       \n\t"   \
313         "mulul  %%d2,    %%d4:%%d1  \n\t"   \
314         "addxl  %%d3,    %%d1       \n\t"   \
315         "addxl  %%d0,    %%d4       \n\t"   \
316         "addl   %%d1,    %%a3@+     \n\t"   \
317         "movel  %%a2@+,  %%d1       \n\t"   \
318         "mulul  %%d2,    %%d3:%%d1  \n\t"   \
319         "addxl  %%d4,    %%d1       \n\t"   \
320         "addxl  %%d0,    %%d3       \n\t"   \
321         "addl   %%d1,    %%a3@+     \n\t"   \
322         "movel  %%a2@+,  %%d1       \n\t"   \
323         "mulul  %%d2,    %%d4:%%d1  \n\t"   \
324         "addxl  %%d3,    %%d1       \n\t"   \
325         "addxl  %%d0,    %%d4       \n\t"   \
326         "addl   %%d1,    %%a3@+     \n\t"   \
327         "movel  %%a2@+,  %%d1       \n\t"   \
328         "mulul  %%d2,    %%d3:%%d1  \n\t"   \
329         "addxl  %%d4,    %%d1       \n\t"   \
330         "addxl  %%d0,    %%d3       \n\t"   \
331         "addl   %%d1,    %%a3@+     \n\t"   \
332         "addxl  %%d0,    %%d3       \n\t"
333 
334 #endif /* MC68000 */
335 
336 #if defined(__powerpc64__) || defined(__ppc64__)
337 
338 #if defined(__MACH__) && defined(__APPLE__)
339 
340 #define MULADDC_INIT                        \
341     asm(                                    \
342         "ld     r3, %3              \n\t"   \
343         "ld     r4, %4              \n\t"   \
344         "ld     r5, %5              \n\t"   \
345         "ld     r6, %6              \n\t"   \
346         "addi   r3, r3, -8          \n\t"   \
347         "addi   r4, r4, -8          \n\t"   \
348         "addic  r5, r5,  0          \n\t"
349 
350 #define MULADDC_CORE                        \
351         "ldu    r7, 8(r3)           \n\t"   \
352         "mulld  r8, r7, r6          \n\t"   \
353         "mulhdu r9, r7, r6          \n\t"   \
354         "adde   r8, r8, r5          \n\t"   \
355         "ld     r7, 8(r4)           \n\t"   \
356         "addze  r5, r9              \n\t"   \
357         "addc   r8, r8, r7          \n\t"   \
358         "stdu   r8, 8(r4)           \n\t"
359 
360 #define MULADDC_STOP                        \
361         "addze  r5, r5              \n\t"   \
362         "addi   r4, r4, 8           \n\t"   \
363         "addi   r3, r3, 8           \n\t"   \
364         "std    r5, %0              \n\t"   \
365         "std    r4, %1              \n\t"   \
366         "std    r3, %2              \n\t"   \
367         : "=m" (c), "=m" (d), "=m" (s)              \
368         : "m" (s), "m" (d), "m" (c), "m" (b)        \
369         : "r3", "r4", "r5", "r6", "r7", "r8", "r9"  \
370     );
371 
372 
373 #else /* __MACH__ && __APPLE__ */
374 
375 #define MULADDC_INIT                        \
376     asm(                                    \
377         "ld     %%r3, %3            \n\t"   \
378         "ld     %%r4, %4            \n\t"   \
379         "ld     %%r5, %5            \n\t"   \
380         "ld     %%r6, %6            \n\t"   \
381         "addi   %%r3, %%r3, -8      \n\t"   \
382         "addi   %%r4, %%r4, -8      \n\t"   \
383         "addic  %%r5, %%r5,  0      \n\t"
384 
385 #define MULADDC_CORE                        \
386         "ldu    %%r7, 8(%%r3)       \n\t"   \
387         "mulld  %%r8, %%r7, %%r6    \n\t"   \
388         "mulhdu %%r9, %%r7, %%r6    \n\t"   \
389         "adde   %%r8, %%r8, %%r5    \n\t"   \
390         "ld     %%r7, 8(%%r4)       \n\t"   \
391         "addze  %%r5, %%r9          \n\t"   \
392         "addc   %%r8, %%r8, %%r7    \n\t"   \
393         "stdu   %%r8, 8(%%r4)       \n\t"
394 
395 #define MULADDC_STOP                        \
396         "addze  %%r5, %%r5          \n\t"   \
397         "addi   %%r4, %%r4, 8       \n\t"   \
398         "addi   %%r3, %%r3, 8       \n\t"   \
399         "std    %%r5, %0            \n\t"   \
400         "std    %%r4, %1            \n\t"   \
401         "std    %%r3, %2            \n\t"   \
402         : "=m" (c), "=m" (d), "=m" (s)              \
403         : "m" (s), "m" (d), "m" (c), "m" (b)        \
404         : "r3", "r4", "r5", "r6", "r7", "r8", "r9"  \
405     );
406 
407 #endif /* __MACH__ && __APPLE__ */
408 
409 #elif defined(__powerpc__) || defined(__ppc__) /* end PPC64/begin PPC32  */
410 
411 #if defined(__MACH__) && defined(__APPLE__)
412 
413 #define MULADDC_INIT                    \
414     asm(                                \
415         "lwz    r3, %3          \n\t"   \
416         "lwz    r4, %4          \n\t"   \
417         "lwz    r5, %5          \n\t"   \
418         "lwz    r6, %6          \n\t"   \
419         "addi   r3, r3, -4      \n\t"   \
420         "addi   r4, r4, -4      \n\t"   \
421         "addic  r5, r5,  0      \n\t"
422 
423 #define MULADDC_CORE                    \
424         "lwzu   r7, 4(r3)       \n\t"   \
425         "mullw  r8, r7, r6      \n\t"   \
426         "mulhwu r9, r7, r6      \n\t"   \
427         "adde   r8, r8, r5      \n\t"   \
428         "lwz    r7, 4(r4)       \n\t"   \
429         "addze  r5, r9          \n\t"   \
430         "addc   r8, r8, r7      \n\t"   \
431         "stwu   r8, 4(r4)       \n\t"
432 
433 #define MULADDC_STOP                    \
434         "addze  r5, r5          \n\t"   \
435         "addi   r4, r4, 4       \n\t"   \
436         "addi   r3, r3, 4       \n\t"   \
437         "stw    r5, %0          \n\t"   \
438         "stw    r4, %1          \n\t"   \
439         "stw    r3, %2          \n\t"   \
440         : "=m" (c), "=m" (d), "=m" (s)              \
441         : "m" (s), "m" (d), "m" (c), "m" (b)        \
442         : "r3", "r4", "r5", "r6", "r7", "r8", "r9"  \
443     );
444 
445 #else /* __MACH__ && __APPLE__ */
446 
447 #define MULADDC_INIT                        \
448     asm(                                    \
449         "lwz    %%r3, %3            \n\t"   \
450         "lwz    %%r4, %4            \n\t"   \
451         "lwz    %%r5, %5            \n\t"   \
452         "lwz    %%r6, %6            \n\t"   \
453         "addi   %%r3, %%r3, -4      \n\t"   \
454         "addi   %%r4, %%r4, -4      \n\t"   \
455         "addic  %%r5, %%r5,  0      \n\t"
456 
457 #define MULADDC_CORE                        \
458         "lwzu   %%r7, 4(%%r3)       \n\t"   \
459         "mullw  %%r8, %%r7, %%r6    \n\t"   \
460         "mulhwu %%r9, %%r7, %%r6    \n\t"   \
461         "adde   %%r8, %%r8, %%r5    \n\t"   \
462         "lwz    %%r7, 4(%%r4)       \n\t"   \
463         "addze  %%r5, %%r9          \n\t"   \
464         "addc   %%r8, %%r8, %%r7    \n\t"   \
465         "stwu   %%r8, 4(%%r4)       \n\t"
466 
467 #define MULADDC_STOP                        \
468         "addze  %%r5, %%r5          \n\t"   \
469         "addi   %%r4, %%r4, 4       \n\t"   \
470         "addi   %%r3, %%r3, 4       \n\t"   \
471         "stw    %%r5, %0            \n\t"   \
472         "stw    %%r4, %1            \n\t"   \
473         "stw    %%r3, %2            \n\t"   \
474         : "=m" (c), "=m" (d), "=m" (s)              \
475         : "m" (s), "m" (d), "m" (c), "m" (b)        \
476         : "r3", "r4", "r5", "r6", "r7", "r8", "r9"  \
477     );
478 
479 #endif /* __MACH__ && __APPLE__ */
480 
481 #endif /* PPC32 */
482 
483 /*
484  * The Sparc(64) assembly is reported to be broken.
485  * Disable it for now, until we're able to fix it.
486  */
487 #if 0 && defined(__sparc__)
488 #if defined(__sparc64__)
489 
490 #define MULADDC_INIT                                    \
491     asm(                                                \
492                 "ldx     %3, %%o0               \n\t"   \
493                 "ldx     %4, %%o1               \n\t"   \
494                 "ld      %5, %%o2               \n\t"   \
495                 "ld      %6, %%o3               \n\t"
496 
497 #define MULADDC_CORE                                    \
498                 "ld      [%%o0], %%o4           \n\t"   \
499                 "inc     4, %%o0                \n\t"   \
500                 "ld      [%%o1], %%o5           \n\t"   \
501                 "umul    %%o3, %%o4, %%o4       \n\t"   \
502                 "addcc   %%o4, %%o2, %%o4       \n\t"   \
503                 "rd      %%y, %%g1              \n\t"   \
504                 "addx    %%g1, 0, %%g1          \n\t"   \
505                 "addcc   %%o4, %%o5, %%o4       \n\t"   \
506                 "st      %%o4, [%%o1]           \n\t"   \
507                 "addx    %%g1, 0, %%o2          \n\t"   \
508                 "inc     4, %%o1                \n\t"
509 
510         #define MULADDC_STOP                            \
511                 "st      %%o2, %0               \n\t"   \
512                 "stx     %%o1, %1               \n\t"   \
513                 "stx     %%o0, %2               \n\t"   \
514         : "=m" (c), "=m" (d), "=m" (s)          \
515         : "m" (s), "m" (d), "m" (c), "m" (b)    \
516         : "g1", "o0", "o1", "o2", "o3", "o4",   \
517           "o5"                                  \
518         );
519 
520 #else /* __sparc64__ */
521 
522 #define MULADDC_INIT                                    \
523     asm(                                                \
524                 "ld      %3, %%o0               \n\t"   \
525                 "ld      %4, %%o1               \n\t"   \
526                 "ld      %5, %%o2               \n\t"   \
527                 "ld      %6, %%o3               \n\t"
528 
529 #define MULADDC_CORE                                    \
530                 "ld      [%%o0], %%o4           \n\t"   \
531                 "inc     4, %%o0                \n\t"   \
532                 "ld      [%%o1], %%o5           \n\t"   \
533                 "umul    %%o3, %%o4, %%o4       \n\t"   \
534                 "addcc   %%o4, %%o2, %%o4       \n\t"   \
535                 "rd      %%y, %%g1              \n\t"   \
536                 "addx    %%g1, 0, %%g1          \n\t"   \
537                 "addcc   %%o4, %%o5, %%o4       \n\t"   \
538                 "st      %%o4, [%%o1]           \n\t"   \
539                 "addx    %%g1, 0, %%o2          \n\t"   \
540                 "inc     4, %%o1                \n\t"
541 
542 #define MULADDC_STOP                                    \
543                 "st      %%o2, %0               \n\t"   \
544                 "st      %%o1, %1               \n\t"   \
545                 "st      %%o0, %2               \n\t"   \
546         : "=m" (c), "=m" (d), "=m" (s)          \
547         : "m" (s), "m" (d), "m" (c), "m" (b)    \
548         : "g1", "o0", "o1", "o2", "o3", "o4",   \
549           "o5"                                  \
550         );
551 
552 #endif /* __sparc64__ */
553 #endif /* __sparc__ */
554 
555 #if defined(__microblaze__) || defined(microblaze)
556 
557 #define MULADDC_INIT                    \
558     asm(                                \
559         "lwi   r3,   %3         \n\t"   \
560         "lwi   r4,   %4         \n\t"   \
561         "lwi   r5,   %5         \n\t"   \
562         "lwi   r6,   %6         \n\t"   \
563         "andi  r7,   r6, 0xffff \n\t"   \
564         "bsrli r6,   r6, 16     \n\t"
565 
566 #define MULADDC_CORE                    \
567         "lhui  r8,   r3,   0    \n\t"   \
568         "addi  r3,   r3,   2    \n\t"   \
569         "lhui  r9,   r3,   0    \n\t"   \
570         "addi  r3,   r3,   2    \n\t"   \
571         "mul   r10,  r9,  r6    \n\t"   \
572         "mul   r11,  r8,  r7    \n\t"   \
573         "mul   r12,  r9,  r7    \n\t"   \
574         "mul   r13,  r8,  r6    \n\t"   \
575         "bsrli  r8, r10,  16    \n\t"   \
576         "bsrli  r9, r11,  16    \n\t"   \
577         "add   r13, r13,  r8    \n\t"   \
578         "add   r13, r13,  r9    \n\t"   \
579         "bslli r10, r10,  16    \n\t"   \
580         "bslli r11, r11,  16    \n\t"   \
581         "add   r12, r12, r10    \n\t"   \
582         "addc  r13, r13,  r0    \n\t"   \
583         "add   r12, r12, r11    \n\t"   \
584         "addc  r13, r13,  r0    \n\t"   \
585         "lwi   r10,  r4,   0    \n\t"   \
586         "add   r12, r12, r10    \n\t"   \
587         "addc  r13, r13,  r0    \n\t"   \
588         "add   r12, r12,  r5    \n\t"   \
589         "addc   r5, r13,  r0    \n\t"   \
590         "swi   r12,  r4,   0    \n\t"   \
591         "addi   r4,  r4,   4    \n\t"
592 
593 #define MULADDC_STOP                    \
594         "swi   r5,   %0         \n\t"   \
595         "swi   r4,   %1         \n\t"   \
596         "swi   r3,   %2         \n\t"   \
597         : "=m" (c), "=m" (d), "=m" (s)              \
598         : "m" (s), "m" (d), "m" (c), "m" (b)        \
599         : "r3", "r4", "r5", "r6", "r7", "r8",       \
600           "r9", "r10", "r11", "r12", "r13"          \
601     );
602 
603 #endif /* MicroBlaze */
604 
605 #if defined(__tricore__)
606 
607 #define MULADDC_INIT                            \
608     asm(                                        \
609         "ld.a   %%a2, %3                \n\t"   \
610         "ld.a   %%a3, %4                \n\t"   \
611         "ld.w   %%d4, %5                \n\t"   \
612         "ld.w   %%d1, %6                \n\t"   \
613         "xor    %%d5, %%d5              \n\t"
614 
615 #define MULADDC_CORE                            \
616         "ld.w   %%d0,   [%%a2+]         \n\t"   \
617         "madd.u %%e2, %%e4, %%d0, %%d1  \n\t"   \
618         "ld.w   %%d0,   [%%a3]          \n\t"   \
619         "addx   %%d2,    %%d2,  %%d0    \n\t"   \
620         "addc   %%d3,    %%d3,    0     \n\t"   \
621         "mov    %%d4,    %%d3           \n\t"   \
622         "st.w  [%%a3+],  %%d2           \n\t"
623 
624 #define MULADDC_STOP                            \
625         "st.w   %0, %%d4                \n\t"   \
626         "st.a   %1, %%a3                \n\t"   \
627         "st.a   %2, %%a2                \n\t"   \
628         : "=m" (c), "=m" (d), "=m" (s)          \
629         : "m" (s), "m" (d), "m" (c), "m" (b)    \
630         : "d0", "d1", "e2", "d4", "a2", "a3"    \
631     );
632 
633 #endif /* TriCore */
634 
635 /*
636  * Note, gcc -O0 by default uses r7 for the frame pointer, so it complains about
637  * our use of r7 below, unless -fomit-frame-pointer is passed.
638  *
639  * On the other hand, -fomit-frame-pointer is implied by any -Ox options with
640  * x !=0, which we can detect using __OPTIMIZE__ (which is also defined by
641  * clang and armcc5 under the same conditions).
642  *
643  * So, only use the optimized assembly below for optimized build, which avoids
644  * the build error and is pretty reasonable anyway.
645  */
646 #if defined(__GNUC__) && !defined(__OPTIMIZE__)
647 #define MULADDC_CANNOT_USE_R7
648 #endif
649 
650 #if defined(__arm__) && !defined(MULADDC_CANNOT_USE_R7)
651 
652 #if defined(__thumb__) && !defined(__thumb2__)
653 
654 #define MULADDC_INIT                                    \
655     asm(                                                \
656             "ldr    r0, %3                      \n\t"   \
657             "ldr    r1, %4                      \n\t"   \
658             "ldr    r2, %5                      \n\t"   \
659             "ldr    r3, %6                      \n\t"   \
660             "lsr    r7, r3, #16                 \n\t"   \
661             "mov    r9, r7                      \n\t"   \
662             "lsl    r7, r3, #16                 \n\t"   \
663             "lsr    r7, r7, #16                 \n\t"   \
664             "mov    r8, r7                      \n\t"
665 
666 #define MULADDC_CORE                                    \
667             "ldmia  r0!, {r6}                   \n\t"   \
668             "lsr    r7, r6, #16                 \n\t"   \
669             "lsl    r6, r6, #16                 \n\t"   \
670             "lsr    r6, r6, #16                 \n\t"   \
671             "mov    r4, r8                      \n\t"   \
672             "mul    r4, r6                      \n\t"   \
673             "mov    r3, r9                      \n\t"   \
674             "mul    r6, r3                      \n\t"   \
675             "mov    r5, r9                      \n\t"   \
676             "mul    r5, r7                      \n\t"   \
677             "mov    r3, r8                      \n\t"   \
678             "mul    r7, r3                      \n\t"   \
679             "lsr    r3, r6, #16                 \n\t"   \
680             "add    r5, r5, r3                  \n\t"   \
681             "lsr    r3, r7, #16                 \n\t"   \
682             "add    r5, r5, r3                  \n\t"   \
683             "add    r4, r4, r2                  \n\t"   \
684             "mov    r2, #0                      \n\t"   \
685             "adc    r5, r2                      \n\t"   \
686             "lsl    r3, r6, #16                 \n\t"   \
687             "add    r4, r4, r3                  \n\t"   \
688             "adc    r5, r2                      \n\t"   \
689             "lsl    r3, r7, #16                 \n\t"   \
690             "add    r4, r4, r3                  \n\t"   \
691             "adc    r5, r2                      \n\t"   \
692             "ldr    r3, [r1]                    \n\t"   \
693             "add    r4, r4, r3                  \n\t"   \
694             "adc    r2, r5                      \n\t"   \
695             "stmia  r1!, {r4}                   \n\t"
696 
697 #define MULADDC_STOP                                    \
698             "str    r2, %0                      \n\t"   \
699             "str    r1, %1                      \n\t"   \
700             "str    r0, %2                      \n\t"   \
701          : "=m" (c),  "=m" (d), "=m" (s)        \
702          : "m" (s), "m" (d), "m" (c), "m" (b)   \
703          : "r0", "r1", "r2", "r3", "r4", "r5",  \
704            "r6", "r7", "r8", "r9", "cc"         \
705          );
706 
707 #elif (__ARM_ARCH >= 6) && \
708     defined (__ARM_FEATURE_DSP) && (__ARM_FEATURE_DSP == 1)
709 
710 #define MULADDC_INIT                            \
711     asm(
712 
713 #define MULADDC_CORE                            \
714             "ldr    r0, [%0], #4        \n\t"   \
715             "ldr    r1, [%1]            \n\t"   \
716             "umaal  r1, %2, %3, r0      \n\t"   \
717             "str    r1, [%1], #4        \n\t"
718 
719 #define MULADDC_STOP                            \
720          : "=r" (s),  "=r" (d), "=r" (c)        \
721          : "r" (b), "0" (s), "1" (d), "2" (c)   \
722          : "r0", "r1", "memory"                 \
723          );
724 
725 #else
726 
727 #define MULADDC_INIT                                    \
728     asm(                                                \
729             "ldr    r0, %3                      \n\t"   \
730             "ldr    r1, %4                      \n\t"   \
731             "ldr    r2, %5                      \n\t"   \
732             "ldr    r3, %6                      \n\t"
733 
734 #define MULADDC_CORE                                    \
735             "ldr    r4, [r0], #4                \n\t"   \
736             "mov    r5, #0                      \n\t"   \
737             "ldr    r6, [r1]                    \n\t"   \
738             "umlal  r2, r5, r3, r4              \n\t"   \
739             "adds   r7, r6, r2                  \n\t"   \
740             "adc    r2, r5, #0                  \n\t"   \
741             "str    r7, [r1], #4                \n\t"
742 
743 #define MULADDC_STOP                                    \
744             "str    r2, %0                      \n\t"   \
745             "str    r1, %1                      \n\t"   \
746             "str    r0, %2                      \n\t"   \
747          : "=m" (c),  "=m" (d), "=m" (s)        \
748          : "m" (s), "m" (d), "m" (c), "m" (b)   \
749          : "r0", "r1", "r2", "r3", "r4", "r5",  \
750            "r6", "r7", "cc"                     \
751          );
752 
753 #endif /* Thumb */
754 
755 #endif /* ARMv3 */
756 
757 #if defined(__alpha__)
758 
759 #define MULADDC_INIT                    \
760     asm(                                \
761         "ldq    $1, %3          \n\t"   \
762         "ldq    $2, %4          \n\t"   \
763         "ldq    $3, %5          \n\t"   \
764         "ldq    $4, %6          \n\t"
765 
766 #define MULADDC_CORE                    \
767         "ldq    $6,  0($1)      \n\t"   \
768         "addq   $1,  8, $1      \n\t"   \
769         "mulq   $6, $4, $7      \n\t"   \
770         "umulh  $6, $4, $6      \n\t"   \
771         "addq   $7, $3, $7      \n\t"   \
772         "cmpult $7, $3, $3      \n\t"   \
773         "ldq    $5,  0($2)      \n\t"   \
774         "addq   $7, $5, $7      \n\t"   \
775         "cmpult $7, $5, $5      \n\t"   \
776         "stq    $7,  0($2)      \n\t"   \
777         "addq   $2,  8, $2      \n\t"   \
778         "addq   $6, $3, $3      \n\t"   \
779         "addq   $5, $3, $3      \n\t"
780 
781 #define MULADDC_STOP                                    \
782         "stq    $3, %0          \n\t"   \
783         "stq    $2, %1          \n\t"   \
784         "stq    $1, %2          \n\t"   \
785         : "=m" (c), "=m" (d), "=m" (s)              \
786         : "m" (s), "m" (d), "m" (c), "m" (b)        \
787         : "$1", "$2", "$3", "$4", "$5", "$6", "$7"  \
788     );
789 #endif /* Alpha */
790 
791 #if defined(__mips__) && !defined(__mips64)
792 
793 #define MULADDC_INIT                    \
794     asm(                                \
795         "lw     $10, %3         \n\t"   \
796         "lw     $11, %4         \n\t"   \
797         "lw     $12, %5         \n\t"   \
798         "lw     $13, %6         \n\t"
799 
800 #define MULADDC_CORE                    \
801         "lw     $14, 0($10)     \n\t"   \
802         "multu  $13, $14        \n\t"   \
803         "addi   $10, $10, 4     \n\t"   \
804         "mflo   $14             \n\t"   \
805         "mfhi   $9              \n\t"   \
806         "addu   $14, $12, $14   \n\t"   \
807         "lw     $15, 0($11)     \n\t"   \
808         "sltu   $12, $14, $12   \n\t"   \
809         "addu   $15, $14, $15   \n\t"   \
810         "sltu   $14, $15, $14   \n\t"   \
811         "addu   $12, $12, $9    \n\t"   \
812         "sw     $15, 0($11)     \n\t"   \
813         "addu   $12, $12, $14   \n\t"   \
814         "addi   $11, $11, 4     \n\t"
815 
816 #define MULADDC_STOP                    \
817         "sw     $12, %0         \n\t"   \
818         "sw     $11, %1         \n\t"   \
819         "sw     $10, %2         \n\t"   \
820         : "=m" (c), "=m" (d), "=m" (s)                      \
821         : "m" (s), "m" (d), "m" (c), "m" (b)                \
822         : "$9", "$10", "$11", "$12", "$13", "$14", "$15", "lo", "hi" \
823     );
824 
825 #endif /* MIPS */
826 #endif /* GNUC */
827 
828 #if (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__)
829 
830 #define MULADDC_INIT                            \
831     __asm   mov     esi, s                      \
832     __asm   mov     edi, d                      \
833     __asm   mov     ecx, c                      \
834     __asm   mov     ebx, b
835 
836 #define MULADDC_CORE                            \
837     __asm   lodsd                               \
838     __asm   mul     ebx                         \
839     __asm   add     eax, ecx                    \
840     __asm   adc     edx, 0                      \
841     __asm   add     eax, [edi]                  \
842     __asm   adc     edx, 0                      \
843     __asm   mov     ecx, edx                    \
844     __asm   stosd
845 
846 #if defined(MBEDTLS_HAVE_SSE2)
847 
848 #define EMIT __asm _emit
849 
850 #define MULADDC_HUIT                            \
851     EMIT 0x0F  EMIT 0x6E  EMIT 0xC9             \
852     EMIT 0x0F  EMIT 0x6E  EMIT 0xC3             \
853     EMIT 0x0F  EMIT 0x6E  EMIT 0x1F             \
854     EMIT 0x0F  EMIT 0xD4  EMIT 0xCB             \
855     EMIT 0x0F  EMIT 0x6E  EMIT 0x16             \
856     EMIT 0x0F  EMIT 0xF4  EMIT 0xD0             \
857     EMIT 0x0F  EMIT 0x6E  EMIT 0x66  EMIT 0x04  \
858     EMIT 0x0F  EMIT 0xF4  EMIT 0xE0             \
859     EMIT 0x0F  EMIT 0x6E  EMIT 0x76  EMIT 0x08  \
860     EMIT 0x0F  EMIT 0xF4  EMIT 0xF0             \
861     EMIT 0x0F  EMIT 0x6E  EMIT 0x7E  EMIT 0x0C  \
862     EMIT 0x0F  EMIT 0xF4  EMIT 0xF8             \
863     EMIT 0x0F  EMIT 0xD4  EMIT 0xCA             \
864     EMIT 0x0F  EMIT 0x6E  EMIT 0x5F  EMIT 0x04  \
865     EMIT 0x0F  EMIT 0xD4  EMIT 0xDC             \
866     EMIT 0x0F  EMIT 0x6E  EMIT 0x6F  EMIT 0x08  \
867     EMIT 0x0F  EMIT 0xD4  EMIT 0xEE             \
868     EMIT 0x0F  EMIT 0x6E  EMIT 0x67  EMIT 0x0C  \
869     EMIT 0x0F  EMIT 0xD4  EMIT 0xFC             \
870     EMIT 0x0F  EMIT 0x7E  EMIT 0x0F             \
871     EMIT 0x0F  EMIT 0x6E  EMIT 0x56  EMIT 0x10  \
872     EMIT 0x0F  EMIT 0xF4  EMIT 0xD0             \
873     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
874     EMIT 0x0F  EMIT 0x6E  EMIT 0x66  EMIT 0x14  \
875     EMIT 0x0F  EMIT 0xF4  EMIT 0xE0             \
876     EMIT 0x0F  EMIT 0xD4  EMIT 0xCB             \
877     EMIT 0x0F  EMIT 0x6E  EMIT 0x76  EMIT 0x18  \
878     EMIT 0x0F  EMIT 0xF4  EMIT 0xF0             \
879     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x04  \
880     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
881     EMIT 0x0F  EMIT 0x6E  EMIT 0x5E  EMIT 0x1C  \
882     EMIT 0x0F  EMIT 0xF4  EMIT 0xD8             \
883     EMIT 0x0F  EMIT 0xD4  EMIT 0xCD             \
884     EMIT 0x0F  EMIT 0x6E  EMIT 0x6F  EMIT 0x10  \
885     EMIT 0x0F  EMIT 0xD4  EMIT 0xD5             \
886     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x08  \
887     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
888     EMIT 0x0F  EMIT 0xD4  EMIT 0xCF             \
889     EMIT 0x0F  EMIT 0x6E  EMIT 0x6F  EMIT 0x14  \
890     EMIT 0x0F  EMIT 0xD4  EMIT 0xE5             \
891     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x0C  \
892     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
893     EMIT 0x0F  EMIT 0xD4  EMIT 0xCA             \
894     EMIT 0x0F  EMIT 0x6E  EMIT 0x6F  EMIT 0x18  \
895     EMIT 0x0F  EMIT 0xD4  EMIT 0xF5             \
896     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x10  \
897     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
898     EMIT 0x0F  EMIT 0xD4  EMIT 0xCC             \
899     EMIT 0x0F  EMIT 0x6E  EMIT 0x6F  EMIT 0x1C  \
900     EMIT 0x0F  EMIT 0xD4  EMIT 0xDD             \
901     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x14  \
902     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
903     EMIT 0x0F  EMIT 0xD4  EMIT 0xCE             \
904     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x18  \
905     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
906     EMIT 0x0F  EMIT 0xD4  EMIT 0xCB             \
907     EMIT 0x0F  EMIT 0x7E  EMIT 0x4F  EMIT 0x1C  \
908     EMIT 0x83  EMIT 0xC7  EMIT 0x20             \
909     EMIT 0x83  EMIT 0xC6  EMIT 0x20             \
910     EMIT 0x0F  EMIT 0x73  EMIT 0xD1  EMIT 0x20  \
911     EMIT 0x0F  EMIT 0x7E  EMIT 0xC9
912 
913 #define MULADDC_STOP                            \
914     EMIT 0x0F  EMIT 0x77                        \
915     __asm   mov     c, ecx                      \
916     __asm   mov     d, edi                      \
917     __asm   mov     s, esi                      \
918 
919 #else
920 
921 #define MULADDC_STOP                            \
922     __asm   mov     c, ecx                      \
923     __asm   mov     d, edi                      \
924     __asm   mov     s, esi                      \
925 
926 #endif /* SSE2 */
927 #endif /* MSVC */
928 
929 #endif /* MBEDTLS_HAVE_ASM */
930 
931 #if !defined(MULADDC_CORE)
932 #if defined(MBEDTLS_HAVE_UDBL)
933 
934 #define MULADDC_INIT                    \
935 {                                       \
936     mbedtls_t_udbl r;                           \
937     mbedtls_mpi_uint r0, r1;
938 
939 #define MULADDC_CORE                    \
940     r   = *(s++) * (mbedtls_t_udbl) b;          \
941     r0  = (mbedtls_mpi_uint) r;                   \
942     r1  = (mbedtls_mpi_uint)( r >> biL );         \
943     r0 += c;  r1 += (r0 <  c);          \
944     r0 += *d; r1 += (r0 < *d);          \
945     c = r1; *(d++) = r0;
946 
947 #define MULADDC_STOP                    \
948 }
949 
950 #else
951 #define MULADDC_INIT                    \
952 {                                       \
953     mbedtls_mpi_uint s0, s1, b0, b1;              \
954     mbedtls_mpi_uint r0, r1, rx, ry;              \
955     b0 = ( b << biH ) >> biH;           \
956     b1 = ( b >> biH );
957 
958 #define MULADDC_CORE                    \
959     s0 = ( *s << biH ) >> biH;          \
960     s1 = ( *s >> biH ); s++;            \
961     rx = s0 * b1; r0 = s0 * b0;         \
962     ry = s1 * b0; r1 = s1 * b1;         \
963     r1 += ( rx >> biH );                \
964     r1 += ( ry >> biH );                \
965     rx <<= biH; ry <<= biH;             \
966     r0 += rx; r1 += (r0 < rx);          \
967     r0 += ry; r1 += (r0 < ry);          \
968     r0 +=  c; r1 += (r0 <  c);          \
969     r0 += *d; r1 += (r0 < *d);          \
970     c = r1; *(d++) = r0;
971 
972 #define MULADDC_STOP                    \
973 }
974 
975 #endif /* C (generic)  */
976 #endif /* C (longlong) */
977 
978 #endif /* bn_mul.h */
979