mbed TLS v1.3.17
bn_mul.h
Go to the documentation of this file.
1 
24 /*
25  * Multiply source vector [s] with b, add result
26  * to destination vector [d] and set carry c.
27  *
28  * Currently supports:
29  *
30  * . IA-32 (386+) . AMD64 / EM64T
31  * . IA-32 (SSE2) . Motorola 68000
32  * . PowerPC, 32-bit . MicroBlaze
33  * . PowerPC, 64-bit . TriCore
34  * . SPARC v8 . ARM v3+
35  * . Alpha . MIPS32
36  * . C, longlong . C, generic
37  */
38 #ifndef POLARSSL_BN_MUL_H
39 #define POLARSSL_BN_MUL_H
40 
41 #include "bignum.h"
42 
43 #if defined(POLARSSL_HAVE_ASM)
44 
45 /* armcc5 --gnu defines __GNUC__ but doesn't support GNU's extended asm */
46 #if defined(__GNUC__) && \
47  ( !defined(__ARMCC_VERSION) || __ARMCC_VERSION >= 6000000 )
48 #if defined(__i386__)
49 
50 #define MULADDC_INIT \
51  asm( \
52  "movl %%ebx, %0 \n\t" \
53  "movl %5, %%esi \n\t" \
54  "movl %6, %%edi \n\t" \
55  "movl %7, %%ecx \n\t" \
56  "movl %8, %%ebx \n\t"
57 
58 #define MULADDC_CORE \
59  "lodsl \n\t" \
60  "mull %%ebx \n\t" \
61  "addl %%ecx, %%eax \n\t" \
62  "adcl $0, %%edx \n\t" \
63  "addl (%%edi), %%eax \n\t" \
64  "adcl $0, %%edx \n\t" \
65  "movl %%edx, %%ecx \n\t" \
66  "stosl \n\t"
67 
68 #if defined(POLARSSL_HAVE_SSE2)
69 
70 #define MULADDC_HUIT \
71  "movd %%ecx, %%mm1 \n\t" \
72  "movd %%ebx, %%mm0 \n\t" \
73  "movd (%%edi), %%mm3 \n\t" \
74  "paddq %%mm3, %%mm1 \n\t" \
75  "movd (%%esi), %%mm2 \n\t" \
76  "pmuludq %%mm0, %%mm2 \n\t" \
77  "movd 4(%%esi), %%mm4 \n\t" \
78  "pmuludq %%mm0, %%mm4 \n\t" \
79  "movd 8(%%esi), %%mm6 \n\t" \
80  "pmuludq %%mm0, %%mm6 \n\t" \
81  "movd 12(%%esi), %%mm7 \n\t" \
82  "pmuludq %%mm0, %%mm7 \n\t" \
83  "paddq %%mm2, %%mm1 \n\t" \
84  "movd 4(%%edi), %%mm3 \n\t" \
85  "paddq %%mm4, %%mm3 \n\t" \
86  "movd 8(%%edi), %%mm5 \n\t" \
87  "paddq %%mm6, %%mm5 \n\t" \
88  "movd 12(%%edi), %%mm4 \n\t" \
89  "paddq %%mm4, %%mm7 \n\t" \
90  "movd %%mm1, (%%edi) \n\t" \
91  "movd 16(%%esi), %%mm2 \n\t" \
92  "pmuludq %%mm0, %%mm2 \n\t" \
93  "psrlq $32, %%mm1 \n\t" \
94  "movd 20(%%esi), %%mm4 \n\t" \
95  "pmuludq %%mm0, %%mm4 \n\t" \
96  "paddq %%mm3, %%mm1 \n\t" \
97  "movd 24(%%esi), %%mm6 \n\t" \
98  "pmuludq %%mm0, %%mm6 \n\t" \
99  "movd %%mm1, 4(%%edi) \n\t" \
100  "psrlq $32, %%mm1 \n\t" \
101  "movd 28(%%esi), %%mm3 \n\t" \
102  "pmuludq %%mm0, %%mm3 \n\t" \
103  "paddq %%mm5, %%mm1 \n\t" \
104  "movd 16(%%edi), %%mm5 \n\t" \
105  "paddq %%mm5, %%mm2 \n\t" \
106  "movd %%mm1, 8(%%edi) \n\t" \
107  "psrlq $32, %%mm1 \n\t" \
108  "paddq %%mm7, %%mm1 \n\t" \
109  "movd 20(%%edi), %%mm5 \n\t" \
110  "paddq %%mm5, %%mm4 \n\t" \
111  "movd %%mm1, 12(%%edi) \n\t" \
112  "psrlq $32, %%mm1 \n\t" \
113  "paddq %%mm2, %%mm1 \n\t" \
114  "movd 24(%%edi), %%mm5 \n\t" \
115  "paddq %%mm5, %%mm6 \n\t" \
116  "movd %%mm1, 16(%%edi) \n\t" \
117  "psrlq $32, %%mm1 \n\t" \
118  "paddq %%mm4, %%mm1 \n\t" \
119  "movd 28(%%edi), %%mm5 \n\t" \
120  "paddq %%mm5, %%mm3 \n\t" \
121  "movd %%mm1, 20(%%edi) \n\t" \
122  "psrlq $32, %%mm1 \n\t" \
123  "paddq %%mm6, %%mm1 \n\t" \
124  "movd %%mm1, 24(%%edi) \n\t" \
125  "psrlq $32, %%mm1 \n\t" \
126  "paddq %%mm3, %%mm1 \n\t" \
127  "movd %%mm1, 28(%%edi) \n\t" \
128  "addl $32, %%edi \n\t" \
129  "addl $32, %%esi \n\t" \
130  "psrlq $32, %%mm1 \n\t" \
131  "movd %%mm1, %%ecx \n\t"
132 
133 #define MULADDC_STOP \
134  "emms \n\t" \
135  "movl %4, %%ebx \n\t" \
136  "movl %%ecx, %1 \n\t" \
137  "movl %%edi, %2 \n\t" \
138  "movl %%esi, %3 \n\t" \
139  : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \
140  : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \
141  : "eax", "ecx", "edx", "esi", "edi" \
142  );
143 
144 #else
145 
146 #define MULADDC_STOP \
147  "movl %4, %%ebx \n\t" \
148  "movl %%ecx, %1 \n\t" \
149  "movl %%edi, %2 \n\t" \
150  "movl %%esi, %3 \n\t" \
151  : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \
152  : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \
153  : "eax", "ecx", "edx", "esi", "edi" \
154  );
155 #endif /* SSE2 */
156 #endif /* i386 */
157 
158 #if defined(__amd64__) || defined (__x86_64__)
159 
160 #define MULADDC_INIT \
161  asm( \
162  "movq %3, %%rsi \n\t" \
163  "movq %4, %%rdi \n\t" \
164  "movq %5, %%rcx \n\t" \
165  "movq %6, %%rbx \n\t" \
166  "xorq %%r8, %%r8 \n\t"
167 
168 #define MULADDC_CORE \
169  "movq (%%rsi), %%rax \n\t" \
170  "mulq %%rbx \n\t" \
171  "addq $8, %%rsi \n\t" \
172  "addq %%rcx, %%rax \n\t" \
173  "movq %%r8, %%rcx \n\t" \
174  "adcq $0, %%rdx \n\t" \
175  "nop \n\t" \
176  "addq %%rax, (%%rdi) \n\t" \
177  "adcq %%rdx, %%rcx \n\t" \
178  "addq $8, %%rdi \n\t"
179 
180 #define MULADDC_STOP \
181  "movq %%rcx, %0 \n\t" \
182  "movq %%rdi, %1 \n\t" \
183  "movq %%rsi, %2 \n\t" \
184  : "=m" (c), "=m" (d), "=m" (s) \
185  : "m" (s), "m" (d), "m" (c), "m" (b) \
186  : "rax", "rcx", "rdx", "rbx", "rsi", "rdi", "r8" \
187  );
188 
189 #endif /* AMD64 */
190 
191 #if defined(__mc68020__) || defined(__mcpu32__)
192 
193 #define MULADDC_INIT \
194  asm( \
195  "movl %3, %%a2 \n\t" \
196  "movl %4, %%a3 \n\t" \
197  "movl %5, %%d3 \n\t" \
198  "movl %6, %%d2 \n\t" \
199  "moveq #0, %%d0 \n\t"
200 
201 #define MULADDC_CORE \
202  "movel %%a2@+, %%d1 \n\t" \
203  "mulul %%d2, %%d4:%%d1 \n\t" \
204  "addl %%d3, %%d1 \n\t" \
205  "addxl %%d0, %%d4 \n\t" \
206  "moveq #0, %%d3 \n\t" \
207  "addl %%d1, %%a3@+ \n\t" \
208  "addxl %%d4, %%d3 \n\t"
209 
210 #define MULADDC_STOP \
211  "movl %%d3, %0 \n\t" \
212  "movl %%a3, %1 \n\t" \
213  "movl %%a2, %2 \n\t" \
214  : "=m" (c), "=m" (d), "=m" (s) \
215  : "m" (s), "m" (d), "m" (c), "m" (b) \
216  : "d0", "d1", "d2", "d3", "d4", "a2", "a3" \
217  );
218 
219 #define MULADDC_HUIT \
220  "movel %%a2@+, %%d1 \n\t" \
221  "mulul %%d2, %%d4:%%d1 \n\t" \
222  "addxl %%d3, %%d1 \n\t" \
223  "addxl %%d0, %%d4 \n\t" \
224  "addl %%d1, %%a3@+ \n\t" \
225  "movel %%a2@+, %%d1 \n\t" \
226  "mulul %%d2, %%d3:%%d1 \n\t" \
227  "addxl %%d4, %%d1 \n\t" \
228  "addxl %%d0, %%d3 \n\t" \
229  "addl %%d1, %%a3@+ \n\t" \
230  "movel %%a2@+, %%d1 \n\t" \
231  "mulul %%d2, %%d4:%%d1 \n\t" \
232  "addxl %%d3, %%d1 \n\t" \
233  "addxl %%d0, %%d4 \n\t" \
234  "addl %%d1, %%a3@+ \n\t" \
235  "movel %%a2@+, %%d1 \n\t" \
236  "mulul %%d2, %%d3:%%d1 \n\t" \
237  "addxl %%d4, %%d1 \n\t" \
238  "addxl %%d0, %%d3 \n\t" \
239  "addl %%d1, %%a3@+ \n\t" \
240  "movel %%a2@+, %%d1 \n\t" \
241  "mulul %%d2, %%d4:%%d1 \n\t" \
242  "addxl %%d3, %%d1 \n\t" \
243  "addxl %%d0, %%d4 \n\t" \
244  "addl %%d1, %%a3@+ \n\t" \
245  "movel %%a2@+, %%d1 \n\t" \
246  "mulul %%d2, %%d3:%%d1 \n\t" \
247  "addxl %%d4, %%d1 \n\t" \
248  "addxl %%d0, %%d3 \n\t" \
249  "addl %%d1, %%a3@+ \n\t" \
250  "movel %%a2@+, %%d1 \n\t" \
251  "mulul %%d2, %%d4:%%d1 \n\t" \
252  "addxl %%d3, %%d1 \n\t" \
253  "addxl %%d0, %%d4 \n\t" \
254  "addl %%d1, %%a3@+ \n\t" \
255  "movel %%a2@+, %%d1 \n\t" \
256  "mulul %%d2, %%d3:%%d1 \n\t" \
257  "addxl %%d4, %%d1 \n\t" \
258  "addxl %%d0, %%d3 \n\t" \
259  "addl %%d1, %%a3@+ \n\t" \
260  "addxl %%d0, %%d3 \n\t"
261 
262 #endif /* MC68000 */
263 
264 #if defined(__powerpc64__) || defined(__ppc64__)
265 
266 #if defined(__MACH__) && defined(__APPLE__)
267 
268 #define MULADDC_INIT \
269  asm( \
270  "ld r3, %3 \n\t" \
271  "ld r4, %4 \n\t" \
272  "ld r5, %5 \n\t" \
273  "ld r6, %6 \n\t" \
274  "addi r3, r3, -8 \n\t" \
275  "addi r4, r4, -8 \n\t" \
276  "addic r5, r5, 0 \n\t"
277 
278 #define MULADDC_CORE \
279  "ldu r7, 8(r3) \n\t" \
280  "mulld r8, r7, r6 \n\t" \
281  "mulhdu r9, r7, r6 \n\t" \
282  "adde r8, r8, r5 \n\t" \
283  "ld r7, 8(r4) \n\t" \
284  "addze r5, r9 \n\t" \
285  "addc r8, r8, r7 \n\t" \
286  "stdu r8, 8(r4) \n\t"
287 
288 #define MULADDC_STOP \
289  "addze r5, r5 \n\t" \
290  "addi r4, r4, 8 \n\t" \
291  "addi r3, r3, 8 \n\t" \
292  "std r5, %0 \n\t" \
293  "std r4, %1 \n\t" \
294  "std r3, %2 \n\t" \
295  : "=m" (c), "=m" (d), "=m" (s) \
296  : "m" (s), "m" (d), "m" (c), "m" (b) \
297  : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \
298  );
299 
300 
301 #else /* __MACH__ && __APPLE__ */
302 
303 #define MULADDC_INIT \
304  asm( \
305  "ld %%r3, %3 \n\t" \
306  "ld %%r4, %4 \n\t" \
307  "ld %%r5, %5 \n\t" \
308  "ld %%r6, %6 \n\t" \
309  "addi %%r3, %%r3, -8 \n\t" \
310  "addi %%r4, %%r4, -8 \n\t" \
311  "addic %%r5, %%r5, 0 \n\t"
312 
313 #define MULADDC_CORE \
314  "ldu %%r7, 8(%%r3) \n\t" \
315  "mulld %%r8, %%r7, %%r6 \n\t" \
316  "mulhdu %%r9, %%r7, %%r6 \n\t" \
317  "adde %%r8, %%r8, %%r5 \n\t" \
318  "ld %%r7, 8(%%r4) \n\t" \
319  "addze %%r5, %%r9 \n\t" \
320  "addc %%r8, %%r8, %%r7 \n\t" \
321  "stdu %%r8, 8(%%r4) \n\t"
322 
323 #define MULADDC_STOP \
324  "addze %%r5, %%r5 \n\t" \
325  "addi %%r4, %%r4, 8 \n\t" \
326  "addi %%r3, %%r3, 8 \n\t" \
327  "std %%r5, %0 \n\t" \
328  "std %%r4, %1 \n\t" \
329  "std %%r3, %2 \n\t" \
330  : "=m" (c), "=m" (d), "=m" (s) \
331  : "m" (s), "m" (d), "m" (c), "m" (b) \
332  : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \
333  );
334 
335 #endif /* __MACH__ && __APPLE__ */
336 
337 #elif defined(__powerpc__) || defined(__ppc__) /* end PPC64/begin PPC32 */
338 
339 #if defined(__MACH__) && defined(__APPLE__)
340 
341 #define MULADDC_INIT \
342  asm( \
343  "lwz r3, %3 \n\t" \
344  "lwz r4, %4 \n\t" \
345  "lwz r5, %5 \n\t" \
346  "lwz r6, %6 \n\t" \
347  "addi r3, r3, -4 \n\t" \
348  "addi r4, r4, -4 \n\t" \
349  "addic r5, r5, 0 \n\t"
350 
351 #define MULADDC_CORE \
352  "lwzu r7, 4(r3) \n\t" \
353  "mullw r8, r7, r6 \n\t" \
354  "mulhwu r9, r7, r6 \n\t" \
355  "adde r8, r8, r5 \n\t" \
356  "lwz r7, 4(r4) \n\t" \
357  "addze r5, r9 \n\t" \
358  "addc r8, r8, r7 \n\t" \
359  "stwu r8, 4(r4) \n\t"
360 
361 #define MULADDC_STOP \
362  "addze r5, r5 \n\t" \
363  "addi r4, r4, 4 \n\t" \
364  "addi r3, r3, 4 \n\t" \
365  "stw r5, %0 \n\t" \
366  "stw r4, %1 \n\t" \
367  "stw r3, %2 \n\t" \
368  : "=m" (c), "=m" (d), "=m" (s) \
369  : "m" (s), "m" (d), "m" (c), "m" (b) \
370  : "r3", "r4", "r5", "r6", "r7", "r8", "r9" \
371  );
372 
373 #else /* __MACH__ && __APPLE__ */
374 
375 #define MULADDC_INIT \
376  asm( \
377  "lwz %%r3, %3 \n\t" \
378  "lwz %%r4, %4 \n\t" \
379  "lwz %%r5, %5 \n\t" \
380  "lwz %%r6, %6 \n\t" \
381  "addi %%r3, %%r3, -4 \n\t" \
382  "addi %%r4, %%r4, -4 \n\t" \
383  "addic %%r5, %%r5, 0 \n\t"
384 
385 #define MULADDC_CORE \
386  "lwzu %%r7, 4(%%r3) \n\t" \
387  "mullw %%r8, %%r7, %%r6 \n\t" \
388  "mulhwu %%r9, %%r7, %%r6 \n\t" \
389  "adde %%r8, %%r8, %%r5 \n\t" \
390  "lwz %%r7, 4(%%r4) \n\t" \
391  "addze %%r5, %%r9 \n\t" \
392  "addc %%r8, %%r8, %%r7 \n\t" \
393  "stwu %%r8, 4(%%r4) \n\t"
394 
395 #define MULADDC_STOP \
396  "addze %%r5, %%r5 \n\t" \
397  "addi %%r4, %%r4, 4 \n\t" \
398  "addi %%r3, %%r3, 4 \n\t" \
399  "stw %%r5, %0 \n\t" \
400  "stw %%r4, %1 \n\t" \
401  "stw %%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 #endif /* PPC32 */
410 
411 /*
412  * The Sparc(64) assembly is reported to be broken.
413  * Disable it for now, until we're able to fix it.
414  */
415 #if 0 && defined(__sparc__)
416 #if defined(__sparc64__)
417 
418 #define MULADDC_INIT \
419  asm( \
420  "ldx %3, %%o0 \n\t" \
421  "ldx %4, %%o1 \n\t" \
422  "ld %5, %%o2 \n\t" \
423  "ld %6, %%o3 \n\t"
424 
425 #define MULADDC_CORE \
426  "ld [%%o0], %%o4 \n\t" \
427  "inc 4, %%o0 \n\t" \
428  "ld [%%o1], %%o5 \n\t" \
429  "umul %%o3, %%o4, %%o4 \n\t" \
430  "addcc %%o4, %%o2, %%o4 \n\t" \
431  "rd %%y, %%g1 \n\t" \
432  "addx %%g1, 0, %%g1 \n\t" \
433  "addcc %%o4, %%o5, %%o4 \n\t" \
434  "st %%o4, [%%o1] \n\t" \
435  "addx %%g1, 0, %%o2 \n\t" \
436  "inc 4, %%o1 \n\t"
437 
438  #define MULADDC_STOP \
439  "st %%o2, %0 \n\t" \
440  "stx %%o1, %1 \n\t" \
441  "stx %%o0, %2 \n\t" \
442  : "=m" (c), "=m" (d), "=m" (s) \
443  : "m" (s), "m" (d), "m" (c), "m" (b) \
444  : "g1", "o0", "o1", "o2", "o3", "o4", \
445  "o5" \
446  );
447 
448 #else /* __sparc64__ */
449 
450 #define MULADDC_INIT \
451  asm( \
452  "ld %3, %%o0 \n\t" \
453  "ld %4, %%o1 \n\t" \
454  "ld %5, %%o2 \n\t" \
455  "ld %6, %%o3 \n\t"
456 
457 #define MULADDC_CORE \
458  "ld [%%o0], %%o4 \n\t" \
459  "inc 4, %%o0 \n\t" \
460  "ld [%%o1], %%o5 \n\t" \
461  "umul %%o3, %%o4, %%o4 \n\t" \
462  "addcc %%o4, %%o2, %%o4 \n\t" \
463  "rd %%y, %%g1 \n\t" \
464  "addx %%g1, 0, %%g1 \n\t" \
465  "addcc %%o4, %%o5, %%o4 \n\t" \
466  "st %%o4, [%%o1] \n\t" \
467  "addx %%g1, 0, %%o2 \n\t" \
468  "inc 4, %%o1 \n\t"
469 
470 #define MULADDC_STOP \
471  "st %%o2, %0 \n\t" \
472  "st %%o1, %1 \n\t" \
473  "st %%o0, %2 \n\t" \
474  : "=m" (c), "=m" (d), "=m" (s) \
475  : "m" (s), "m" (d), "m" (c), "m" (b) \
476  : "g1", "o0", "o1", "o2", "o3", "o4", \
477  "o5" \
478  );
479 
480 #endif /* __sparc64__ */
481 #endif /* __sparc__ */
482 
483 #if defined(__microblaze__) || defined(microblaze)
484 
485 #define MULADDC_INIT \
486  asm( \
487  "lwi r3, %3 \n\t" \
488  "lwi r4, %4 \n\t" \
489  "lwi r5, %5 \n\t" \
490  "lwi r6, %6 \n\t" \
491  "andi r7, r6, 0xffff \n\t" \
492  "bsrli r6, r6, 16 \n\t"
493 
494 #define MULADDC_CORE \
495  "lhui r8, r3, 0 \n\t" \
496  "addi r3, r3, 2 \n\t" \
497  "lhui r9, r3, 0 \n\t" \
498  "addi r3, r3, 2 \n\t" \
499  "mul r10, r9, r6 \n\t" \
500  "mul r11, r8, r7 \n\t" \
501  "mul r12, r9, r7 \n\t" \
502  "mul r13, r8, r6 \n\t" \
503  "bsrli r8, r10, 16 \n\t" \
504  "bsrli r9, r11, 16 \n\t" \
505  "add r13, r13, r8 \n\t" \
506  "add r13, r13, r9 \n\t" \
507  "bslli r10, r10, 16 \n\t" \
508  "bslli r11, r11, 16 \n\t" \
509  "add r12, r12, r10 \n\t" \
510  "addc r13, r13, r0 \n\t" \
511  "add r12, r12, r11 \n\t" \
512  "addc r13, r13, r0 \n\t" \
513  "lwi r10, r4, 0 \n\t" \
514  "add r12, r12, r10 \n\t" \
515  "addc r13, r13, r0 \n\t" \
516  "add r12, r12, r5 \n\t" \
517  "addc r5, r13, r0 \n\t" \
518  "swi r12, r4, 0 \n\t" \
519  "addi r4, r4, 4 \n\t"
520 
521 #define MULADDC_STOP \
522  "swi r5, %0 \n\t" \
523  "swi r4, %1 \n\t" \
524  "swi r3, %2 \n\t" \
525  : "=m" (c), "=m" (d), "=m" (s) \
526  : "m" (s), "m" (d), "m" (c), "m" (b) \
527  : "r3", "r4" "r5", "r6", "r7", "r8", \
528  "r9", "r10", "r11", "r12", "r13" \
529  );
530 
531 #endif /* MicroBlaze */
532 
533 #if defined(__tricore__)
534 
535 #define MULADDC_INIT \
536  asm( \
537  "ld.a %%a2, %3 \n\t" \
538  "ld.a %%a3, %4 \n\t" \
539  "ld.w %%d4, %5 \n\t" \
540  "ld.w %%d1, %6 \n\t" \
541  "xor %%d5, %%d5 \n\t"
542 
543 #define MULADDC_CORE \
544  "ld.w %%d0, [%%a2+] \n\t" \
545  "madd.u %%e2, %%e4, %%d0, %%d1 \n\t" \
546  "ld.w %%d0, [%%a3] \n\t" \
547  "addx %%d2, %%d2, %%d0 \n\t" \
548  "addc %%d3, %%d3, 0 \n\t" \
549  "mov %%d4, %%d3 \n\t" \
550  "st.w [%%a3+], %%d2 \n\t"
551 
552 #define MULADDC_STOP \
553  "st.w %0, %%d4 \n\t" \
554  "st.a %1, %%a3 \n\t" \
555  "st.a %2, %%a2 \n\t" \
556  : "=m" (c), "=m" (d), "=m" (s) \
557  : "m" (s), "m" (d), "m" (c), "m" (b) \
558  : "d0", "d1", "e2", "d4", "a2", "a3" \
559  );
560 
561 #endif /* TriCore */
562 
563 /*
564  * gcc -O0 by default uses r7 for the frame pointer, so it complains about our
565  * use of r7 below, unless -fomit-frame-pointer is passed. Unfortunately,
566  * passing that option is not easy when building with yotta.
567  *
568  * On the other hand, -fomit-frame-pointer is implied by any -Ox options with
569  * x !=0, which we can detect using __OPTIMIZE__ (which is also defined by
570  * clang and armcc5 under the same conditions).
571  *
572  * So, only use the optimized assembly below for optimized build, which avoids
573  * the build error and is pretty reasonable anyway.
574  */
575 #if defined(__GNUC__) && !defined(__OPTIMIZE__)
576 #define CANNOT_USE_R7
577 #endif
578 
579 #if defined(__arm__) && !defined(CANNOT_USE_R7)
580 
581 #if defined(__thumb__) && !defined(__thumb2__)
582 
583 #define MULADDC_INIT \
584  asm( \
585  "ldr r0, %3 \n\t" \
586  "ldr r1, %4 \n\t" \
587  "ldr r2, %5 \n\t" \
588  "ldr r3, %6 \n\t" \
589  "lsr r7, r3, #16 \n\t" \
590  "mov r9, r7 \n\t" \
591  "lsl r7, r3, #16 \n\t" \
592  "lsr r7, r7, #16 \n\t" \
593  "mov r8, r7 \n\t"
594 
595 #define MULADDC_CORE \
596  "ldmia r0!, {r6} \n\t" \
597  "lsr r7, r6, #16 \n\t" \
598  "lsl r6, r6, #16 \n\t" \
599  "lsr r6, r6, #16 \n\t" \
600  "mov r4, r8 \n\t" \
601  "mul r4, r6 \n\t" \
602  "mov r3, r9 \n\t" \
603  "mul r6, r3 \n\t" \
604  "mov r5, r9 \n\t" \
605  "mul r5, r7 \n\t" \
606  "mov r3, r8 \n\t" \
607  "mul r7, r3 \n\t" \
608  "lsr r3, r6, #16 \n\t" \
609  "add r5, r5, r3 \n\t" \
610  "lsr r3, r7, #16 \n\t" \
611  "add r5, r5, r3 \n\t" \
612  "add r4, r4, r2 \n\t" \
613  "mov r2, #0 \n\t" \
614  "adc r5, r2 \n\t" \
615  "lsl r3, r6, #16 \n\t" \
616  "add r4, r4, r3 \n\t" \
617  "adc r5, r2 \n\t" \
618  "lsl r3, r7, #16 \n\t" \
619  "add r4, r4, r3 \n\t" \
620  "adc r5, r2 \n\t" \
621  "ldr r3, [r1] \n\t" \
622  "add r4, r4, r3 \n\t" \
623  "adc r2, r5 \n\t" \
624  "stmia r1!, {r4} \n\t"
625 
626 #define MULADDC_STOP \
627  "str r2, %0 \n\t" \
628  "str r1, %1 \n\t" \
629  "str r0, %2 \n\t" \
630  : "=m" (c), "=m" (d), "=m" (s) \
631  : "m" (s), "m" (d), "m" (c), "m" (b) \
632  : "r0", "r1", "r2", "r3", "r4", "r5", \
633  "r6", "r7", "r8", "r9", "cc" \
634  );
635 
636 #else
637 
638 #define MULADDC_INIT \
639  asm( \
640  "ldr r0, %3 \n\t" \
641  "ldr r1, %4 \n\t" \
642  "ldr r2, %5 \n\t" \
643  "ldr r3, %6 \n\t"
644 
645 #define MULADDC_CORE \
646  "ldr r4, [r0], #4 \n\t" \
647  "mov r5, #0 \n\t" \
648  "ldr r6, [r1] \n\t" \
649  "umlal r2, r5, r3, r4 \n\t" \
650  "adds r7, r6, r2 \n\t" \
651  "adc r2, r5, #0 \n\t" \
652  "str r7, [r1], #4 \n\t"
653 
654 #define MULADDC_STOP \
655  "str r2, %0 \n\t" \
656  "str r1, %1 \n\t" \
657  "str r0, %2 \n\t" \
658  : "=m" (c), "=m" (d), "=m" (s) \
659  : "m" (s), "m" (d), "m" (c), "m" (b) \
660  : "r0", "r1", "r2", "r3", "r4", "r5", \
661  "r6", "r7", "cc" \
662  );
663 
664 #endif /* Thumb */
665 
666 #endif /* ARMv3 */
667 
668 #if defined(__alpha__)
669 
670 #define MULADDC_INIT \
671  asm( \
672  "ldq $1, %3 \n\t" \
673  "ldq $2, %4 \n\t" \
674  "ldq $3, %5 \n\t" \
675  "ldq $4, %6 \n\t"
676 
677 #define MULADDC_CORE \
678  "ldq $6, 0($1) \n\t" \
679  "addq $1, 8, $1 \n\t" \
680  "mulq $6, $4, $7 \n\t" \
681  "umulh $6, $4, $6 \n\t" \
682  "addq $7, $3, $7 \n\t" \
683  "cmpult $7, $3, $3 \n\t" \
684  "ldq $5, 0($2) \n\t" \
685  "addq $7, $5, $7 \n\t" \
686  "cmpult $7, $5, $5 \n\t" \
687  "stq $7, 0($2) \n\t" \
688  "addq $2, 8, $2 \n\t" \
689  "addq $6, $3, $3 \n\t" \
690  "addq $5, $3, $3 \n\t"
691 
692 #define MULADDC_STOP \
693  "stq $3, %0 \n\t" \
694  "stq $2, %1 \n\t" \
695  "stq $1, %2 \n\t" \
696  : "=m" (c), "=m" (d), "=m" (s) \
697  : "m" (s), "m" (d), "m" (c), "m" (b) \
698  : "$1", "$2", "$3", "$4", "$5", "$6", "$7" \
699  );
700 #endif /* Alpha */
701 
702 #if defined(__mips__) && !defined(__mips64)
703 
704 #define MULADDC_INIT \
705  asm( \
706  "lw $10, %3 \n\t" \
707  "lw $11, %4 \n\t" \
708  "lw $12, %5 \n\t" \
709  "lw $13, %6 \n\t"
710 
711 #define MULADDC_CORE \
712  "lw $14, 0($10) \n\t" \
713  "multu $13, $14 \n\t" \
714  "addi $10, $10, 4 \n\t" \
715  "mflo $14 \n\t" \
716  "mfhi $9 \n\t" \
717  "addu $14, $12, $14 \n\t" \
718  "lw $15, 0($11) \n\t" \
719  "sltu $12, $14, $12 \n\t" \
720  "addu $15, $14, $15 \n\t" \
721  "sltu $14, $15, $14 \n\t" \
722  "addu $12, $12, $9 \n\t" \
723  "sw $15, 0($11) \n\t" \
724  "addu $12, $12, $14 \n\t" \
725  "addi $11, $11, 4 \n\t"
726 
727 #define MULADDC_STOP \
728  "sw $12, %0 \n\t" \
729  "sw $11, %1 \n\t" \
730  "sw $10, %2 \n\t" \
731  : "=m" (c), "=m" (d), "=m" (s) \
732  : "m" (s), "m" (d), "m" (c), "m" (b) \
733  : "$9", "$10", "$11", "$12", "$13", "$14", "$15" \
734  );
735 
736 #endif /* MIPS */
737 #endif /* GNUC */
738 
739 #if (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__)
740 
741 #define MULADDC_INIT \
742  __asm mov esi, s \
743  __asm mov edi, d \
744  __asm mov ecx, c \
745  __asm mov ebx, b
746 
747 #define MULADDC_CORE \
748  __asm lodsd \
749  __asm mul ebx \
750  __asm add eax, ecx \
751  __asm adc edx, 0 \
752  __asm add eax, [edi] \
753  __asm adc edx, 0 \
754  __asm mov ecx, edx \
755  __asm stosd
756 
757 #if defined(POLARSSL_HAVE_SSE2)
758 
759 #define EMIT __asm _emit
760 
761 #define MULADDC_HUIT \
762  EMIT 0x0F EMIT 0x6E EMIT 0xC9 \
763  EMIT 0x0F EMIT 0x6E EMIT 0xC3 \
764  EMIT 0x0F EMIT 0x6E EMIT 0x1F \
765  EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
766  EMIT 0x0F EMIT 0x6E EMIT 0x16 \
767  EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
768  EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x04 \
769  EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
770  EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x08 \
771  EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
772  EMIT 0x0F EMIT 0x6E EMIT 0x7E EMIT 0x0C \
773  EMIT 0x0F EMIT 0xF4 EMIT 0xF8 \
774  EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
775  EMIT 0x0F EMIT 0x6E EMIT 0x5F EMIT 0x04 \
776  EMIT 0x0F EMIT 0xD4 EMIT 0xDC \
777  EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x08 \
778  EMIT 0x0F EMIT 0xD4 EMIT 0xEE \
779  EMIT 0x0F EMIT 0x6E EMIT 0x67 EMIT 0x0C \
780  EMIT 0x0F EMIT 0xD4 EMIT 0xFC \
781  EMIT 0x0F EMIT 0x7E EMIT 0x0F \
782  EMIT 0x0F EMIT 0x6E EMIT 0x56 EMIT 0x10 \
783  EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
784  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
785  EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x14 \
786  EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
787  EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
788  EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x18 \
789  EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
790  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x04 \
791  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
792  EMIT 0x0F EMIT 0x6E EMIT 0x5E EMIT 0x1C \
793  EMIT 0x0F EMIT 0xF4 EMIT 0xD8 \
794  EMIT 0x0F EMIT 0xD4 EMIT 0xCD \
795  EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x10 \
796  EMIT 0x0F EMIT 0xD4 EMIT 0xD5 \
797  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x08 \
798  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
799  EMIT 0x0F EMIT 0xD4 EMIT 0xCF \
800  EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x14 \
801  EMIT 0x0F EMIT 0xD4 EMIT 0xE5 \
802  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x0C \
803  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
804  EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
805  EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x18 \
806  EMIT 0x0F EMIT 0xD4 EMIT 0xF5 \
807  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x10 \
808  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
809  EMIT 0x0F EMIT 0xD4 EMIT 0xCC \
810  EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x1C \
811  EMIT 0x0F EMIT 0xD4 EMIT 0xDD \
812  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x14 \
813  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
814  EMIT 0x0F EMIT 0xD4 EMIT 0xCE \
815  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x18 \
816  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
817  EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
818  EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x1C \
819  EMIT 0x83 EMIT 0xC7 EMIT 0x20 \
820  EMIT 0x83 EMIT 0xC6 EMIT 0x20 \
821  EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
822  EMIT 0x0F EMIT 0x7E EMIT 0xC9
823 
824 #define MULADDC_STOP \
825  EMIT 0x0F EMIT 0x77 \
826  __asm mov c, ecx \
827  __asm mov d, edi \
828  __asm mov s, esi \
829 
830 #else
831 
832 #define MULADDC_STOP \
833  __asm mov c, ecx \
834  __asm mov d, edi \
835  __asm mov s, esi \
836 
837 #endif /* SSE2 */
838 #endif /* MSVC */
839 
840 #endif /* POLARSSL_HAVE_ASM */
841 
842 #if !defined(MULADDC_CORE)
843 #if defined(POLARSSL_HAVE_UDBL)
844 
845 #define MULADDC_INIT \
846 { \
847  t_udbl r; \
848  t_uint r0, r1;
849 
850 #define MULADDC_CORE \
851  r = *(s++) * (t_udbl) b; \
852  r0 = (t_uint) r; \
853  r1 = (t_uint)( r >> biL ); \
854  r0 += c; r1 += (r0 < c); \
855  r0 += *d; r1 += (r0 < *d); \
856  c = r1; *(d++) = r0;
857 
858 #define MULADDC_STOP \
859 }
860 
861 #else
862 #define MULADDC_INIT \
863 { \
864  t_uint s0, s1, b0, b1; \
865  t_uint r0, r1, rx, ry; \
866  b0 = ( b << biH ) >> biH; \
867  b1 = ( b >> biH );
868 
869 #define MULADDC_CORE \
870  s0 = ( *s << biH ) >> biH; \
871  s1 = ( *s >> biH ); s++; \
872  rx = s0 * b1; r0 = s0 * b0; \
873  ry = s1 * b0; r1 = s1 * b1; \
874  r1 += ( rx >> biH ); \
875  r1 += ( ry >> biH ); \
876  rx <<= biH; ry <<= biH; \
877  r0 += rx; r1 += (r0 < rx); \
878  r0 += ry; r1 += (r0 < ry); \
879  r0 += c; r1 += (r0 < c); \
880  r0 += *d; r1 += (r0 < *d); \
881  c = r1; *(d++) = r0;
882 
883 #define MULADDC_STOP \
884 }
885 
886 #endif /* C (generic) */
887 #endif /* C (longlong) */
888 
889 #endif /* bn_mul.h */
Multi-precision integer library.