TrustedFirmware Git Browser
Code Review Sign In
review.trustedfirmware.org / mirror / mbed-tls / 5121ce5bdbc92e134a42220a1aa23f95037c7625 / . / include / polarssl / bn_mul.h
blob: 4213e74f186954eca37013859caab98f9d23a8ff [file] [log] [blame]
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame^]1/**
2 * \file bn_mul.h
3 */
4/*
5 * Multiply source vector [s] with b, add result
6 * to destination vector [d] and set carry c.
7 *
8 * Currently supports:
9 *
10 * . IA-32 (386+) . AMD64 / EM64T
11 * . IA-32 (SSE2) . Motorola 68000
12 * . PowerPC, 32-bit . MicroBlaze
13 * . PowerPC, 64-bit . TriCore
14 * . SPARC v8 . ARM v3+
15 * . Alpha . MIPS32
16 * . C, longlong . C, generic
17 */
18#ifndef XYSSL_BN_MUL_H
19#define XYSSL_BN_MUL_H
20
21#include "xyssl/config.h"
22
23#if defined(XYSSL_HAVE_ASM)
24
25#if defined(__GNUC__)
26#if defined(__i386__)
27
28#define MULADDC_INIT \
29 asm( "movl %%ebx, %0 " : "=m" (t)); \
30 asm( "movl %0, %%esi " :: "m" (s)); \
31 asm( "movl %0, %%edi " :: "m" (d)); \
32 asm( "movl %0, %%ecx " :: "m" (c)); \
33 asm( "movl %0, %%ebx " :: "m" (b));
34
35#define MULADDC_CORE \
36 asm( "lodsl " ); \
37 asm( "mull %ebx " ); \
38 asm( "addl %ecx, %eax " ); \
39 asm( "adcl $0, %edx " ); \
40 asm( "addl (%edi), %eax " ); \
41 asm( "adcl $0, %edx " ); \
42 asm( "movl %edx, %ecx " ); \
43 asm( "stosl " );
44
45#if defined(XYSSL_HAVE_SSE2)
46
47#define MULADDC_HUIT \
48 asm( "movd %ecx, %mm1 " ); \
49 asm( "movd %ebx, %mm0 " ); \
50 asm( "movd (%edi), %mm3 " ); \
51 asm( "paddq %mm3, %mm1 " ); \
52 asm( "movd (%esi), %mm2 " ); \
53 asm( "pmuludq %mm0, %mm2 " ); \
54 asm( "movd 4(%esi), %mm4 " ); \
55 asm( "pmuludq %mm0, %mm4 " ); \
56 asm( "movd 8(%esi), %mm6 " ); \
57 asm( "pmuludq %mm0, %mm6 " ); \
58 asm( "movd 12(%esi), %mm7 " ); \
59 asm( "pmuludq %mm0, %mm7 " ); \
60 asm( "paddq %mm2, %mm1 " ); \
61 asm( "movd 4(%edi), %mm3 " ); \
62 asm( "paddq %mm4, %mm3 " ); \
63 asm( "movd 8(%edi), %mm5 " ); \
64 asm( "paddq %mm6, %mm5 " ); \
65 asm( "movd 12(%edi), %mm4 " ); \
66 asm( "paddq %mm4, %mm7 " ); \
67 asm( "movd %mm1, (%edi) " ); \
68 asm( "movd 16(%esi), %mm2 " ); \
69 asm( "pmuludq %mm0, %mm2 " ); \
70 asm( "psrlq $32, %mm1 " ); \
71 asm( "movd 20(%esi), %mm4 " ); \
72 asm( "pmuludq %mm0, %mm4 " ); \
73 asm( "paddq %mm3, %mm1 " ); \
74 asm( "movd 24(%esi), %mm6 " ); \
75 asm( "pmuludq %mm0, %mm6 " ); \
76 asm( "movd %mm1, 4(%edi) " ); \
77 asm( "psrlq $32, %mm1 " ); \
78 asm( "movd 28(%esi), %mm3 " ); \
79 asm( "pmuludq %mm0, %mm3 " ); \
80 asm( "paddq %mm5, %mm1 " ); \
81 asm( "movd 16(%edi), %mm5 " ); \
82 asm( "paddq %mm5, %mm2 " ); \
83 asm( "movd %mm1, 8(%edi) " ); \
84 asm( "psrlq $32, %mm1 " ); \
85 asm( "paddq %mm7, %mm1 " ); \
86 asm( "movd 20(%edi), %mm5 " ); \
87 asm( "paddq %mm5, %mm4 " ); \
88 asm( "movd %mm1, 12(%edi) " ); \
89 asm( "psrlq $32, %mm1 " ); \
90 asm( "paddq %mm2, %mm1 " ); \
91 asm( "movd 24(%edi), %mm5 " ); \
92 asm( "paddq %mm5, %mm6 " ); \
93 asm( "movd %mm1, 16(%edi) " ); \
94 asm( "psrlq $32, %mm1 " ); \
95 asm( "paddq %mm4, %mm1 " ); \
96 asm( "movd 28(%edi), %mm5 " ); \
97 asm( "paddq %mm5, %mm3 " ); \
98 asm( "movd %mm1, 20(%edi) " ); \
99 asm( "psrlq $32, %mm1 " ); \
100 asm( "paddq %mm6, %mm1 " ); \
101 asm( "movd %mm1, 24(%edi) " ); \
102 asm( "psrlq $32, %mm1 " ); \
103 asm( "paddq %mm3, %mm1 " ); \
104 asm( "movd %mm1, 28(%edi) " ); \
105 asm( "addl $32, %edi " ); \
106 asm( "addl $32, %esi " ); \
107 asm( "psrlq $32, %mm1 " ); \
108 asm( "movd %mm1, %ecx " );
109
110#define MULADDC_STOP \
111 asm( "emms " ); \
112 asm( "movl %0, %%ebx " :: "m" (t)); \
113 asm( "movl %%ecx, %0 " : "=m" (c)); \
114 asm( "movl %%edi, %0 " : "=m" (d)); \
115 asm( "movl %%esi, %0 " : "=m" (s) :: \
116 "eax", "ecx", "edx", "esi", "edi" );
117
118#else
119
120#define MULADDC_STOP \
121 asm( "movl %0, %%ebx " :: "m" (t)); \
122 asm( "movl %%ecx, %0 " : "=m" (c)); \
123 asm( "movl %%edi, %0 " : "=m" (d)); \
124 asm( "movl %%esi, %0 " : "=m" (s) :: \
125 "eax", "ecx", "edx", "esi", "edi" );
126
127#endif /* SSE2 */
128#endif /* i386 */
129
130#if defined(__amd64__) || defined (__x86_64__)
131
132#define MULADDC_INIT \
133 asm( "movq %0, %%rsi " :: "m" (s)); \
134 asm( "movq %0, %%rdi " :: "m" (d)); \
135 asm( "movq %0, %%rcx " :: "m" (c)); \
136 asm( "movq %0, %%rbx " :: "m" (b)); \
137 asm( "xorq %r8, %r8 " );
138
139#define MULADDC_CORE \
140 asm( "movq (%rsi),%rax " ); \
141 asm( "mulq %rbx " ); \
142 asm( "addq $8, %rsi " ); \
143 asm( "addq %rcx, %rax " ); \
144 asm( "movq %r8, %rcx " ); \
145 asm( "adcq $0, %rdx " ); \
146 asm( "nop " ); \
147 asm( "addq %rax, (%rdi) " ); \
148 asm( "adcq %rdx, %rcx " ); \
149 asm( "addq $8, %rdi " );
150
151#define MULADDC_STOP \
152 asm( "movq %%rcx, %0 " : "=m" (c)); \
153 asm( "movq %%rdi, %0 " : "=m" (d)); \
154 asm( "movq %%rsi, %0 " : "=m" (s) :: \
155 "rax", "rcx", "rdx", "rbx", "rsi", "rdi", "r8" );
156
157#endif /* AMD64 */
158
159#if defined(__mc68020__) || defined(__mcpu32__)
160
161#define MULADDC_INIT \
162 asm( "movl %0, %%a2 " :: "m" (s)); \
163 asm( "movl %0, %%a3 " :: "m" (d)); \
164 asm( "movl %0, %%d3 " :: "m" (c)); \
165 asm( "movl %0, %%d2 " :: "m" (b)); \
166 asm( "moveq #0, %d0 " );
167
168#define MULADDC_CORE \
169 asm( "movel %a2@+, %d1 " ); \
170 asm( "mulul %d2, %d4:%d1 " ); \
171 asm( "addl %d3, %d1 " ); \
172 asm( "addxl %d0, %d4 " ); \
173 asm( "moveq #0, %d3 " ); \
174 asm( "addl %d1, %a3@+ " ); \
175 asm( "addxl %d4, %d3 " );
176
177#define MULADDC_STOP \
178 asm( "movl %%d3, %0 " : "=m" (c)); \
179 asm( "movl %%a3, %0 " : "=m" (d)); \
180 asm( "movl %%a2, %0 " : "=m" (s) :: \
181 "d0", "d1", "d2", "d3", "d4", "a2", "a3" );
182
183#define MULADDC_HUIT \
184 asm( "movel %a2@+, %d1 " ); \
185 asm( "mulul %d2, %d4:%d1 " ); \
186 asm( "addxl %d3, %d1 " ); \
187 asm( "addxl %d0, %d4 " ); \
188 asm( "addl %d1, %a3@+ " ); \
189 asm( "movel %a2@+, %d1 " ); \
190 asm( "mulul %d2, %d3:%d1 " ); \
191 asm( "addxl %d4, %d1 " ); \
192 asm( "addxl %d0, %d3 " ); \
193 asm( "addl %d1, %a3@+ " ); \
194 asm( "movel %a2@+, %d1 " ); \
195 asm( "mulul %d2, %d4:%d1 " ); \
196 asm( "addxl %d3, %d1 " ); \
197 asm( "addxl %d0, %d4 " ); \
198 asm( "addl %d1, %a3@+ " ); \
199 asm( "movel %a2@+, %d1 " ); \
200 asm( "mulul %d2, %d3:%d1 " ); \
201 asm( "addxl %d4, %d1 " ); \
202 asm( "addxl %d0, %d3 " ); \
203 asm( "addl %d1, %a3@+ " ); \
204 asm( "movel %a2@+, %d1 " ); \
205 asm( "mulul %d2, %d4:%d1 " ); \
206 asm( "addxl %d3, %d1 " ); \
207 asm( "addxl %d0, %d4 " ); \
208 asm( "addl %d1, %a3@+ " ); \
209 asm( "movel %a2@+, %d1 " ); \
210 asm( "mulul %d2, %d3:%d1 " ); \
211 asm( "addxl %d4, %d1 " ); \
212 asm( "addxl %d0, %d3 " ); \
213 asm( "addl %d1, %a3@+ " ); \
214 asm( "movel %a2@+, %d1 " ); \
215 asm( "mulul %d2, %d4:%d1 " ); \
216 asm( "addxl %d3, %d1 " ); \
217 asm( "addxl %d0, %d4 " ); \
218 asm( "addl %d1, %a3@+ " ); \
219 asm( "movel %a2@+, %d1 " ); \
220 asm( "mulul %d2, %d3:%d1 " ); \
221 asm( "addxl %d4, %d1 " ); \
222 asm( "addxl %d0, %d3 " ); \
223 asm( "addl %d1, %a3@+ " ); \
224 asm( "addxl %d0, %d3 " );
225
226#endif /* MC68000 */
227
228#if defined(__powerpc__) || defined(__ppc__)
229#if defined(__powerpc64__) || defined(__ppc64__)
230
231#if defined(__MACH__) && defined(__APPLE__)
232
233#define MULADDC_INIT \
234 asm( "ld r3, %0 " :: "m" (s)); \
235 asm( "ld r4, %0 " :: "m" (d)); \
236 asm( "ld r5, %0 " :: "m" (c)); \
237 asm( "ld r6, %0 " :: "m" (b)); \
238 asm( "addi r3, r3, -8 " ); \
239 asm( "addi r4, r4, -8 " ); \
240 asm( "addic r5, r5, 0 " );
241
242#define MULADDC_CORE \
243 asm( "ldu r7, 8(r3) " ); \
244 asm( "mulld r8, r7, r6 " ); \
245 asm( "mulhdu r9, r7, r6 " ); \
246 asm( "adde r8, r8, r5 " ); \
247 asm( "ld r7, 8(r4) " ); \
248 asm( "addze r5, r9 " ); \
249 asm( "addc r8, r8, r7 " ); \
250 asm( "stdu r8, 8(r4) " );
251
252#define MULADDC_STOP \
253 asm( "addze r5, r5 " ); \
254 asm( "addi r4, r4, 8 " ); \
255 asm( "addi r3, r3, 8 " ); \
256 asm( "std r5, %0 " : "=m" (c)); \
257 asm( "std r4, %0 " : "=m" (d)); \
258 asm( "std r3, %0 " : "=m" (s) :: \
259 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
260
261#else
262
263#define MULADDC_INIT \
264 asm( "ld %%r3, %0 " :: "m" (s)); \
265 asm( "ld %%r4, %0 " :: "m" (d)); \
266 asm( "ld %%r5, %0 " :: "m" (c)); \
267 asm( "ld %%r6, %0 " :: "m" (b)); \
268 asm( "addi %r3, %r3, -8 " ); \
269 asm( "addi %r4, %r4, -8 " ); \
270 asm( "addic %r5, %r5, 0 " );
271
272#define MULADDC_CORE \
273 asm( "ldu %r7, 8(%r3) " ); \
274 asm( "mulld %r8, %r7, %r6 " ); \
275 asm( "mulhdu %r9, %r7, %r6 " ); \
276 asm( "adde %r8, %r8, %r5 " ); \
277 asm( "ld %r7, 8(%r4) " ); \
278 asm( "addze %r5, %r9 " ); \
279 asm( "addc %r8, %r8, %r7 " ); \
280 asm( "stdu %r8, 8(%r4) " );
281
282#define MULADDC_STOP \
283 asm( "addze %r5, %r5 " ); \
284 asm( "addi %r4, %r4, 8 " ); \
285 asm( "addi %r3, %r3, 8 " ); \
286 asm( "std %%r5, %0 " : "=m" (c)); \
287 asm( "std %%r4, %0 " : "=m" (d)); \
288 asm( "std %%r3, %0 " : "=m" (s) :: \
289 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
290
291#endif
292
293#else /* PPC32 */
294
295#if defined(__MACH__) && defined(__APPLE__)
296
297#define MULADDC_INIT \
298 asm( "lwz r3, %0 " :: "m" (s)); \
299 asm( "lwz r4, %0 " :: "m" (d)); \
300 asm( "lwz r5, %0 " :: "m" (c)); \
301 asm( "lwz r6, %0 " :: "m" (b)); \
302 asm( "addi r3, r3, -4 " ); \
303 asm( "addi r4, r4, -4 " ); \
304 asm( "addic r5, r5, 0 " );
305
306#define MULADDC_CORE \
307 asm( "lwzu r7, 4(r3) " ); \
308 asm( "mullw r8, r7, r6 " ); \
309 asm( "mulhwu r9, r7, r6 " ); \
310 asm( "adde r8, r8, r5 " ); \
311 asm( "lwz r7, 4(r4) " ); \
312 asm( "addze r5, r9 " ); \
313 asm( "addc r8, r8, r7 " ); \
314 asm( "stwu r8, 4(r4) " );
315
316#define MULADDC_STOP \
317 asm( "addze r5, r5 " ); \
318 asm( "addi r4, r4, 4 " ); \
319 asm( "addi r3, r3, 4 " ); \
320 asm( "stw r5, %0 " : "=m" (c)); \
321 asm( "stw r4, %0 " : "=m" (d)); \
322 asm( "stw r3, %0 " : "=m" (s) :: \
323 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
324
325#else
326
327#define MULADDC_INIT \
328 asm( "lwz %%r3, %0 " :: "m" (s)); \
329 asm( "lwz %%r4, %0 " :: "m" (d)); \
330 asm( "lwz %%r5, %0 " :: "m" (c)); \
331 asm( "lwz %%r6, %0 " :: "m" (b)); \
332 asm( "addi %r3, %r3, -4 " ); \
333 asm( "addi %r4, %r4, -4 " ); \
334 asm( "addic %r5, %r5, 0 " );
335
336#define MULADDC_CORE \
337 asm( "lwzu %r7, 4(%r3) " ); \
338 asm( "mullw %r8, %r7, %r6 " ); \
339 asm( "mulhwu %r9, %r7, %r6 " ); \
340 asm( "adde %r8, %r8, %r5 " ); \
341 asm( "lwz %r7, 4(%r4) " ); \
342 asm( "addze %r5, %r9 " ); \
343 asm( "addc %r8, %r8, %r7 " ); \
344 asm( "stwu %r8, 4(%r4) " );
345
346#define MULADDC_STOP \
347 asm( "addze %r5, %r5 " ); \
348 asm( "addi %r4, %r4, 4 " ); \
349 asm( "addi %r3, %r3, 4 " ); \
350 asm( "stw %%r5, %0 " : "=m" (c)); \
351 asm( "stw %%r4, %0 " : "=m" (d)); \
352 asm( "stw %%r3, %0 " : "=m" (s) :: \
353 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
354
355#endif
356
357#endif /* PPC32 */
358#endif /* PPC64 */
359
360#if defined(__sparc__)
361
362#define MULADDC_INIT \
363 asm( "ld %0, %%o0 " :: "m" (s)); \
364 asm( "ld %0, %%o1 " :: "m" (d)); \
365 asm( "ld %0, %%o2 " :: "m" (c)); \
366 asm( "ld %0, %%o3 " :: "m" (b));
367
368#define MULADDC_CORE \
369 asm( "ld [%o0], %o4 " ); \
370 asm( "inc 4, %o0 " ); \
371 asm( "ld [%o1], %o5 " ); \
372 asm( "umul %o3, %o4, %o4 " ); \
373 asm( "addcc %o4, %o2, %o4 " ); \
374 asm( "rd %y, %g1 " ); \
375 asm( "addx %g1, 0, %g1 " ); \
376 asm( "addcc %o4, %o5, %o4 " ); \
377 asm( "st %o4, [%o1] " ); \
378 asm( "addx %g1, 0, %o2 " ); \
379 asm( "inc 4, %o1 " );
380
381#define MULADDC_STOP \
382 asm( "st %%o2, %0 " : "=m" (c)); \
383 asm( "st %%o1, %0 " : "=m" (d)); \
384 asm( "st %%o0, %0 " : "=m" (s) :: \
385 "g1", "o0", "o1", "o2", "o3", "o4", "o5" );
386
387#endif /* SPARCv8 */
388
389#if defined(__microblaze__) || defined(microblaze)
390
391#define MULADDC_INIT \
392 asm( "lwi r3, %0 " :: "m" (s)); \
393 asm( "lwi r4, %0 " :: "m" (d)); \
394 asm( "lwi r5, %0 " :: "m" (c)); \
395 asm( "lwi r6, %0 " :: "m" (b)); \
396 asm( "andi r7, r6, 0xffff" ); \
397 asm( "bsrli r6, r6, 16 " );
398
399#define MULADDC_CORE \
400 asm( "lhui r8, r3, 0 " ); \
401 asm( "addi r3, r3, 2 " ); \
402 asm( "lhui r9, r3, 0 " ); \
403 asm( "addi r3, r3, 2 " ); \
404 asm( "mul r10, r9, r6 " ); \
405 asm( "mul r11, r8, r7 " ); \
406 asm( "mul r12, r9, r7 " ); \
407 asm( "mul r13, r8, r6 " ); \
408 asm( "bsrli r8, r10, 16 " ); \
409 asm( "bsrli r9, r11, 16 " ); \
410 asm( "add r13, r13, r8 " ); \
411 asm( "add r13, r13, r9 " ); \
412 asm( "bslli r10, r10, 16 " ); \
413 asm( "bslli r11, r11, 16 " ); \
414 asm( "add r12, r12, r10 " ); \
415 asm( "addc r13, r13, r0 " ); \
416 asm( "add r12, r12, r11 " ); \
417 asm( "addc r13, r13, r0 " ); \
418 asm( "lwi r10, r4, 0 " ); \
419 asm( "add r12, r12, r10 " ); \
420 asm( "addc r13, r13, r0 " ); \
421 asm( "add r12, r12, r5 " ); \
422 asm( "addc r5, r13, r0 " ); \
423 asm( "swi r12, r4, 0 " ); \
424 asm( "addi r4, r4, 4 " );
425
426#define MULADDC_STOP \
427 asm( "swi r5, %0 " : "=m" (c)); \
428 asm( "swi r4, %0 " : "=m" (d)); \
429 asm( "swi r3, %0 " : "=m" (s) :: \
430 "r3", "r4" , "r5" , "r6" , "r7" , "r8" , \
431 "r9", "r10", "r11", "r12", "r13" );
432
433#endif /* MicroBlaze */
434
435#if defined(__tricore__)
436
437#define MULADDC_INIT \
438 asm( "ld.a %%a2, %0 " :: "m" (s)); \
439 asm( "ld.a %%a3, %0 " :: "m" (d)); \
440 asm( "ld.w %%d4, %0 " :: "m" (c)); \
441 asm( "ld.w %%d1, %0 " :: "m" (b)); \
442 asm( "xor %d5, %d5 " );
443
444#define MULADDC_CORE \
445 asm( "ld.w %d0, [%a2+] " ); \
446 asm( "madd.u %e2, %e4, %d0, %d1 " ); \
447 asm( "ld.w %d0, [%a3] " ); \
448 asm( "addx %d2, %d2, %d0 " ); \
449 asm( "addc %d3, %d3, 0 " ); \
450 asm( "mov %d4, %d3 " ); \
451 asm( "st.w [%a3+], %d2 " );
452
453#define MULADDC_STOP \
454 asm( "st.w %0, %%d4 " : "=m" (c)); \
455 asm( "st.a %0, %%a3 " : "=m" (d)); \
456 asm( "st.a %0, %%a2 " : "=m" (s) :: \
457 "d0", "d1", "e2", "d4", "a2", "a3" );
458
459#endif /* TriCore */
460
461#if defined(__arm__)
462
463#define MULADDC_INIT \
464 asm( "ldr r0, %0 " :: "m" (s)); \
465 asm( "ldr r1, %0 " :: "m" (d)); \
466 asm( "ldr r2, %0 " :: "m" (c)); \
467 asm( "ldr r3, %0 " :: "m" (b));
468
469#define MULADDC_CORE \
470 asm( "ldr r4, [r0], #4 " ); \
471 asm( "mov r5, #0 " ); \
472 asm( "ldr r6, [r1] " ); \
473 asm( "umlal r2, r5, r3, r4 " ); \
474 asm( "adds r7, r6, r2 " ); \
475 asm( "adc r2, r5, #0 " ); \
476 asm( "str r7, [r1], #4 " );
477
478#define MULADDC_STOP \
479 asm( "str r2, %0 " : "=m" (c)); \
480 asm( "str r1, %0 " : "=m" (d)); \
481 asm( "str r0, %0 " : "=m" (s) :: \
482 "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7" );
483
484#endif /* ARMv3 */
485
486#if defined(__alpha__)
487
488#define MULADDC_INIT \
489 asm( "ldq $1, %0 " :: "m" (s)); \
490 asm( "ldq $2, %0 " :: "m" (d)); \
491 asm( "ldq $3, %0 " :: "m" (c)); \
492 asm( "ldq $4, %0 " :: "m" (b));
493
494#define MULADDC_CORE \
495 asm( "ldq $6, 0($1) " ); \
496 asm( "addq $1, 8, $1 " ); \
497 asm( "mulq $6, $4, $7 " ); \
498 asm( "umulh $6, $4, $6 " ); \
499 asm( "addq $7, $3, $7 " ); \
500 asm( "cmpult $7, $3, $3 " ); \
501 asm( "ldq $5, 0($2) " ); \
502 asm( "addq $7, $5, $7 " ); \
503 asm( "cmpult $7, $5, $5 " ); \
504 asm( "stq $7, 0($2) " ); \
505 asm( "addq $2, 8, $2 " ); \
506 asm( "addq $6, $3, $3 " ); \
507 asm( "addq $5, $3, $3 " );
508
509#define MULADDC_STOP \
510 asm( "stq $3, %0 " : "=m" (c)); \
511 asm( "stq $2, %0 " : "=m" (d)); \
512 asm( "stq $1, %0 " : "=m" (s) :: \
513 "$1", "$2", "$3", "$4", "$5", "$6", "$7" );
514
515#endif /* Alpha */
516
517#if defined(__mips__)
518
519#define MULADDC_INIT \
520 asm( "lw $10, %0 " :: "m" (s)); \
521 asm( "lw $11, %0 " :: "m" (d)); \
522 asm( "lw $12, %0 " :: "m" (c)); \
523 asm( "lw $13, %0 " :: "m" (b));
524
525#define MULADDC_CORE \
526 asm( "lw $14, 0($10) " ); \
527 asm( "multu $13, $14 " ); \
528 asm( "addi $10, $10, 4 " ); \
529 asm( "mflo $14 " ); \
530 asm( "mfhi $9 " ); \
531 asm( "addu $14, $12, $14 " ); \
532 asm( "lw $15, 0($11) " ); \
533 asm( "sltu $12, $14, $12 " ); \
534 asm( "addu $15, $14, $15 " ); \
535 asm( "sltu $14, $15, $14 " ); \
536 asm( "addu $12, $12, $9 " ); \
537 asm( "sw $15, 0($11) " ); \
538 asm( "addu $12, $12, $14 " ); \
539 asm( "addi $11, $11, 4 " );
540
541#define MULADDC_STOP \
542 asm( "sw $12, %0 " : "=m" (c)); \
543 asm( "sw $11, %0 " : "=m" (d)); \
544 asm( "sw $10, %0 " : "=m" (s) :: \
545 "$9", "$10", "$11", "$12", "$13", "$14", "$15" );
546
547#endif /* MIPS */
548#endif /* GNUC */
549
550#if (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__)
551
552#define MULADDC_INIT \
553 __asm mov esi, s \
554 __asm mov edi, d \
555 __asm mov ecx, c \
556 __asm mov ebx, b
557
558#define MULADDC_CORE \
559 __asm lodsd \
560 __asm mul ebx \
561 __asm add eax, ecx \
562 __asm adc edx, 0 \
563 __asm add eax, [edi] \
564 __asm adc edx, 0 \
565 __asm mov ecx, edx \
566 __asm stosd
567
568#if defined(XYSSL_HAVE_SSE2)
569
570#define EMIT __asm _emit
571
572#define MULADDC_HUIT \
573 EMIT 0x0F EMIT 0x6E EMIT 0xC9 \
574 EMIT 0x0F EMIT 0x6E EMIT 0xC3 \
575 EMIT 0x0F EMIT 0x6E EMIT 0x1F \
576 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
577 EMIT 0x0F EMIT 0x6E EMIT 0x16 \
578 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
579 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x04 \
580 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
581 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x08 \
582 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
583 EMIT 0x0F EMIT 0x6E EMIT 0x7E EMIT 0x0C \
584 EMIT 0x0F EMIT 0xF4 EMIT 0xF8 \
585 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
586 EMIT 0x0F EMIT 0x6E EMIT 0x5F EMIT 0x04 \
587 EMIT 0x0F EMIT 0xD4 EMIT 0xDC \
588 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x08 \
589 EMIT 0x0F EMIT 0xD4 EMIT 0xEE \
590 EMIT 0x0F EMIT 0x6E EMIT 0x67 EMIT 0x0C \
591 EMIT 0x0F EMIT 0xD4 EMIT 0xFC \
592 EMIT 0x0F EMIT 0x7E EMIT 0x0F \
593 EMIT 0x0F EMIT 0x6E EMIT 0x56 EMIT 0x10 \
594 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
595 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
596 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x14 \
597 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
598 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
599 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x18 \
600 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
601 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x04 \
602 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
603 EMIT 0x0F EMIT 0x6E EMIT 0x5E EMIT 0x1C \
604 EMIT 0x0F EMIT 0xF4 EMIT 0xD8 \
605 EMIT 0x0F EMIT 0xD4 EMIT 0xCD \
606 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x10 \
607 EMIT 0x0F EMIT 0xD4 EMIT 0xD5 \
608 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x08 \
609 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
610 EMIT 0x0F EMIT 0xD4 EMIT 0xCF \
611 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x14 \
612 EMIT 0x0F EMIT 0xD4 EMIT 0xE5 \
613 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x0C \
614 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
615 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
616 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x18 \
617 EMIT 0x0F EMIT 0xD4 EMIT 0xF5 \
618 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x10 \
619 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
620 EMIT 0x0F EMIT 0xD4 EMIT 0xCC \
621 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x1C \
622 EMIT 0x0F EMIT 0xD4 EMIT 0xDD \
623 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x14 \
624 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
625 EMIT 0x0F EMIT 0xD4 EMIT 0xCE \
626 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x18 \
627 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
628 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
629 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x1C \
630 EMIT 0x83 EMIT 0xC7 EMIT 0x20 \
631 EMIT 0x83 EMIT 0xC6 EMIT 0x20 \
632 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
633 EMIT 0x0F EMIT 0x7E EMIT 0xC9
634
635#define MULADDC_STOP \
636 EMIT 0x0F EMIT 0x77 \
637 __asm mov c, ecx \
638 __asm mov d, edi \
639 __asm mov s, esi \
640
641#else
642
643#define MULADDC_STOP \
644 __asm mov c, ecx \
645 __asm mov d, edi \
646 __asm mov s, esi \
647
648#endif /* SSE2 */
649#endif /* MSVC */
650
651#endif /* XYSSL_HAVE_ASM */
652
653#if !defined(MULADDC_CORE)
654#if defined(XYSSL_HAVE_LONGLONG)
655
656#define MULADDC_INIT \
657{ \
658 t_dbl r; \
659 t_int r0, r1;
660
661#define MULADDC_CORE \
662 r = *(s++) * (t_dbl) b; \
663 r0 = r; \
664 r1 = r >> biL; \
665 r0 += c; r1 += (r0 < c); \
666 r0 += *d; r1 += (r0 < *d); \
667 c = r1; *(d++) = r0;
668
669#define MULADDC_STOP \
670}
671
672#else
673#define MULADDC_INIT \
674{ \
675 t_int s0, s1, b0, b1; \
676 t_int r0, r1, rx, ry; \
677 b0 = ( b << biH ) >> biH; \
678 b1 = ( b >> biH );
679
680#define MULADDC_CORE \
681 s0 = ( *s << biH ) >> biH; \
682 s1 = ( *s >> biH ); s++; \
683 rx = s0 * b1; r0 = s0 * b0; \
684 ry = s1 * b0; r1 = s1 * b1; \
685 r1 += ( rx >> biH ); \
686 r1 += ( ry >> biH ); \
687 rx <<= biH; ry <<= biH; \
688 r0 += rx; r1 += (r0 < rx); \
689 r0 += ry; r1 += (r0 < ry); \
690 r0 += c; r1 += (r0 < c); \
691 r0 += *d; r1 += (r0 < *d); \
692 c = r1; *(d++) = r0;
693
694#define MULADDC_STOP \
695}
696
697#endif /* C (generic) */
698#endif /* C (longlong) */
699
700#endif /* bn_mul.h */