TrustedFirmware Git Browser
Code Review Sign In
review.trustedfirmware.org / mirror / mbed-tls / refs/heads/polarssl-0.10 / . / include / polarssl / bn_mul.h
blob: d4ea7bdb4601679fc7c6f369e1b513b6d3b6b2b5 [file] [log] [blame]
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]1/**
2 * \file bn_mul.h
Paul Bakkere0ccd0a2009-01-04 16:27:10 +0000[diff] [blame]3 *
4 * Based on XySSL: Copyright (C) 2006-2008 Christophe Devine
5 *
Paul Bakker27db1f52009-01-25 15:27:00 +0000[diff] [blame]6 * Copyright (C) 2009 Paul Bakker <polarssl_maintainer at polarssl dot org>
Paul Bakkere0ccd0a2009-01-04 16:27:10 +0000[diff] [blame]7 *
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
12 *
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
17 *
18 * You should have received a copy of the GNU General Public License along
19 * with this program; if not, write to the Free Software Foundation, Inc.,
20 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]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 */
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]36#ifndef POLARSSL_BN_MUL_H
37#define POLARSSL_BN_MUL_H
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]38
Paul Bakker8e831ed2009-01-03 21:24:11 +0000[diff] [blame]39#include "polarssl/config.h"
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]40
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]41#if defined(POLARSSL_HAVE_ASM)
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]42
43#if defined(__GNUC__)
44#if defined(__i386__)
45
46#define MULADDC_INIT \
47 asm( "movl %%ebx, %0 " : "=m" (t)); \
48 asm( "movl %0, %%esi " :: "m" (s)); \
49 asm( "movl %0, %%edi " :: "m" (d)); \
50 asm( "movl %0, %%ecx " :: "m" (c)); \
51 asm( "movl %0, %%ebx " :: "m" (b));
52
53#define MULADDC_CORE \
54 asm( "lodsl " ); \
55 asm( "mull %ebx " ); \
56 asm( "addl %ecx, %eax " ); \
57 asm( "adcl $0, %edx " ); \
58 asm( "addl (%edi), %eax " ); \
59 asm( "adcl $0, %edx " ); \
60 asm( "movl %edx, %ecx " ); \
61 asm( "stosl " );
62
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]63#if defined(POLARSSL_HAVE_SSE2)
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]64
65#define MULADDC_HUIT \
66 asm( "movd %ecx, %mm1 " ); \
67 asm( "movd %ebx, %mm0 " ); \
68 asm( "movd (%edi), %mm3 " ); \
69 asm( "paddq %mm3, %mm1 " ); \
70 asm( "movd (%esi), %mm2 " ); \
71 asm( "pmuludq %mm0, %mm2 " ); \
72 asm( "movd 4(%esi), %mm4 " ); \
73 asm( "pmuludq %mm0, %mm4 " ); \
74 asm( "movd 8(%esi), %mm6 " ); \
75 asm( "pmuludq %mm0, %mm6 " ); \
76 asm( "movd 12(%esi), %mm7 " ); \
77 asm( "pmuludq %mm0, %mm7 " ); \
78 asm( "paddq %mm2, %mm1 " ); \
79 asm( "movd 4(%edi), %mm3 " ); \
80 asm( "paddq %mm4, %mm3 " ); \
81 asm( "movd 8(%edi), %mm5 " ); \
82 asm( "paddq %mm6, %mm5 " ); \
83 asm( "movd 12(%edi), %mm4 " ); \
84 asm( "paddq %mm4, %mm7 " ); \
85 asm( "movd %mm1, (%edi) " ); \
86 asm( "movd 16(%esi), %mm2 " ); \
87 asm( "pmuludq %mm0, %mm2 " ); \
88 asm( "psrlq $32, %mm1 " ); \
89 asm( "movd 20(%esi), %mm4 " ); \
90 asm( "pmuludq %mm0, %mm4 " ); \
91 asm( "paddq %mm3, %mm1 " ); \
92 asm( "movd 24(%esi), %mm6 " ); \
93 asm( "pmuludq %mm0, %mm6 " ); \
94 asm( "movd %mm1, 4(%edi) " ); \
95 asm( "psrlq $32, %mm1 " ); \
96 asm( "movd 28(%esi), %mm3 " ); \
97 asm( "pmuludq %mm0, %mm3 " ); \
98 asm( "paddq %mm5, %mm1 " ); \
99 asm( "movd 16(%edi), %mm5 " ); \
100 asm( "paddq %mm5, %mm2 " ); \
101 asm( "movd %mm1, 8(%edi) " ); \
102 asm( "psrlq $32, %mm1 " ); \
103 asm( "paddq %mm7, %mm1 " ); \
104 asm( "movd 20(%edi), %mm5 " ); \
105 asm( "paddq %mm5, %mm4 " ); \
106 asm( "movd %mm1, 12(%edi) " ); \
107 asm( "psrlq $32, %mm1 " ); \
108 asm( "paddq %mm2, %mm1 " ); \
109 asm( "movd 24(%edi), %mm5 " ); \
110 asm( "paddq %mm5, %mm6 " ); \
111 asm( "movd %mm1, 16(%edi) " ); \
112 asm( "psrlq $32, %mm1 " ); \
113 asm( "paddq %mm4, %mm1 " ); \
114 asm( "movd 28(%edi), %mm5 " ); \
115 asm( "paddq %mm5, %mm3 " ); \
116 asm( "movd %mm1, 20(%edi) " ); \
117 asm( "psrlq $32, %mm1 " ); \
118 asm( "paddq %mm6, %mm1 " ); \
119 asm( "movd %mm1, 24(%edi) " ); \
120 asm( "psrlq $32, %mm1 " ); \
121 asm( "paddq %mm3, %mm1 " ); \
122 asm( "movd %mm1, 28(%edi) " ); \
123 asm( "addl $32, %edi " ); \
124 asm( "addl $32, %esi " ); \
125 asm( "psrlq $32, %mm1 " ); \
126 asm( "movd %mm1, %ecx " );
127
128#define MULADDC_STOP \
129 asm( "emms " ); \
130 asm( "movl %0, %%ebx " :: "m" (t)); \
131 asm( "movl %%ecx, %0 " : "=m" (c)); \
132 asm( "movl %%edi, %0 " : "=m" (d)); \
133 asm( "movl %%esi, %0 " : "=m" (s) :: \
134 "eax", "ecx", "edx", "esi", "edi" );
135
136#else
137
138#define MULADDC_STOP \
139 asm( "movl %0, %%ebx " :: "m" (t)); \
140 asm( "movl %%ecx, %0 " : "=m" (c)); \
141 asm( "movl %%edi, %0 " : "=m" (d)); \
142 asm( "movl %%esi, %0 " : "=m" (s) :: \
143 "eax", "ecx", "edx", "esi", "edi" );
144
145#endif /* SSE2 */
146#endif /* i386 */
147
148#if defined(__amd64__) || defined (__x86_64__)
149
150#define MULADDC_INIT \
151 asm( "movq %0, %%rsi " :: "m" (s)); \
152 asm( "movq %0, %%rdi " :: "m" (d)); \
153 asm( "movq %0, %%rcx " :: "m" (c)); \
154 asm( "movq %0, %%rbx " :: "m" (b)); \
155 asm( "xorq %r8, %r8 " );
156
157#define MULADDC_CORE \
158 asm( "movq (%rsi),%rax " ); \
159 asm( "mulq %rbx " ); \
160 asm( "addq $8, %rsi " ); \
161 asm( "addq %rcx, %rax " ); \
162 asm( "movq %r8, %rcx " ); \
163 asm( "adcq $0, %rdx " ); \
164 asm( "nop " ); \
165 asm( "addq %rax, (%rdi) " ); \
166 asm( "adcq %rdx, %rcx " ); \
167 asm( "addq $8, %rdi " );
168
169#define MULADDC_STOP \
170 asm( "movq %%rcx, %0 " : "=m" (c)); \
171 asm( "movq %%rdi, %0 " : "=m" (d)); \
172 asm( "movq %%rsi, %0 " : "=m" (s) :: \
173 "rax", "rcx", "rdx", "rbx", "rsi", "rdi", "r8" );
174
175#endif /* AMD64 */
176
177#if defined(__mc68020__) || defined(__mcpu32__)
178
179#define MULADDC_INIT \
180 asm( "movl %0, %%a2 " :: "m" (s)); \
181 asm( "movl %0, %%a3 " :: "m" (d)); \
182 asm( "movl %0, %%d3 " :: "m" (c)); \
183 asm( "movl %0, %%d2 " :: "m" (b)); \
184 asm( "moveq #0, %d0 " );
185
186#define MULADDC_CORE \
187 asm( "movel %a2@+, %d1 " ); \
188 asm( "mulul %d2, %d4:%d1 " ); \
189 asm( "addl %d3, %d1 " ); \
190 asm( "addxl %d0, %d4 " ); \
191 asm( "moveq #0, %d3 " ); \
192 asm( "addl %d1, %a3@+ " ); \
193 asm( "addxl %d4, %d3 " );
194
195#define MULADDC_STOP \
196 asm( "movl %%d3, %0 " : "=m" (c)); \
197 asm( "movl %%a3, %0 " : "=m" (d)); \
198 asm( "movl %%a2, %0 " : "=m" (s) :: \
199 "d0", "d1", "d2", "d3", "d4", "a2", "a3" );
200
201#define MULADDC_HUIT \
202 asm( "movel %a2@+, %d1 " ); \
203 asm( "mulul %d2, %d4:%d1 " ); \
204 asm( "addxl %d3, %d1 " ); \
205 asm( "addxl %d0, %d4 " ); \
206 asm( "addl %d1, %a3@+ " ); \
207 asm( "movel %a2@+, %d1 " ); \
208 asm( "mulul %d2, %d3:%d1 " ); \
209 asm( "addxl %d4, %d1 " ); \
210 asm( "addxl %d0, %d3 " ); \
211 asm( "addl %d1, %a3@+ " ); \
212 asm( "movel %a2@+, %d1 " ); \
213 asm( "mulul %d2, %d4:%d1 " ); \
214 asm( "addxl %d3, %d1 " ); \
215 asm( "addxl %d0, %d4 " ); \
216 asm( "addl %d1, %a3@+ " ); \
217 asm( "movel %a2@+, %d1 " ); \
218 asm( "mulul %d2, %d3:%d1 " ); \
219 asm( "addxl %d4, %d1 " ); \
220 asm( "addxl %d0, %d3 " ); \
221 asm( "addl %d1, %a3@+ " ); \
222 asm( "movel %a2@+, %d1 " ); \
223 asm( "mulul %d2, %d4:%d1 " ); \
224 asm( "addxl %d3, %d1 " ); \
225 asm( "addxl %d0, %d4 " ); \
226 asm( "addl %d1, %a3@+ " ); \
227 asm( "movel %a2@+, %d1 " ); \
228 asm( "mulul %d2, %d3:%d1 " ); \
229 asm( "addxl %d4, %d1 " ); \
230 asm( "addxl %d0, %d3 " ); \
231 asm( "addl %d1, %a3@+ " ); \
232 asm( "movel %a2@+, %d1 " ); \
233 asm( "mulul %d2, %d4:%d1 " ); \
234 asm( "addxl %d3, %d1 " ); \
235 asm( "addxl %d0, %d4 " ); \
236 asm( "addl %d1, %a3@+ " ); \
237 asm( "movel %a2@+, %d1 " ); \
238 asm( "mulul %d2, %d3:%d1 " ); \
239 asm( "addxl %d4, %d1 " ); \
240 asm( "addxl %d0, %d3 " ); \
241 asm( "addl %d1, %a3@+ " ); \
242 asm( "addxl %d0, %d3 " );
243
244#endif /* MC68000 */
245
246#if defined(__powerpc__) || defined(__ppc__)
247#if defined(__powerpc64__) || defined(__ppc64__)
248
249#if defined(__MACH__) && defined(__APPLE__)
250
251#define MULADDC_INIT \
252 asm( "ld r3, %0 " :: "m" (s)); \
253 asm( "ld r4, %0 " :: "m" (d)); \
254 asm( "ld r5, %0 " :: "m" (c)); \
255 asm( "ld r6, %0 " :: "m" (b)); \
256 asm( "addi r3, r3, -8 " ); \
257 asm( "addi r4, r4, -8 " ); \
258 asm( "addic r5, r5, 0 " );
259
260#define MULADDC_CORE \
261 asm( "ldu r7, 8(r3) " ); \
262 asm( "mulld r8, r7, r6 " ); \
263 asm( "mulhdu r9, r7, r6 " ); \
264 asm( "adde r8, r8, r5 " ); \
265 asm( "ld r7, 8(r4) " ); \
266 asm( "addze r5, r9 " ); \
267 asm( "addc r8, r8, r7 " ); \
268 asm( "stdu r8, 8(r4) " );
269
270#define MULADDC_STOP \
271 asm( "addze r5, r5 " ); \
272 asm( "addi r4, r4, 8 " ); \
273 asm( "addi r3, r3, 8 " ); \
274 asm( "std r5, %0 " : "=m" (c)); \
275 asm( "std r4, %0 " : "=m" (d)); \
276 asm( "std r3, %0 " : "=m" (s) :: \
277 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
278
279#else
280
281#define MULADDC_INIT \
282 asm( "ld %%r3, %0 " :: "m" (s)); \
283 asm( "ld %%r4, %0 " :: "m" (d)); \
284 asm( "ld %%r5, %0 " :: "m" (c)); \
285 asm( "ld %%r6, %0 " :: "m" (b)); \
286 asm( "addi %r3, %r3, -8 " ); \
287 asm( "addi %r4, %r4, -8 " ); \
288 asm( "addic %r5, %r5, 0 " );
289
290#define MULADDC_CORE \
291 asm( "ldu %r7, 8(%r3) " ); \
292 asm( "mulld %r8, %r7, %r6 " ); \
293 asm( "mulhdu %r9, %r7, %r6 " ); \
294 asm( "adde %r8, %r8, %r5 " ); \
295 asm( "ld %r7, 8(%r4) " ); \
296 asm( "addze %r5, %r9 " ); \
297 asm( "addc %r8, %r8, %r7 " ); \
298 asm( "stdu %r8, 8(%r4) " );
299
300#define MULADDC_STOP \
301 asm( "addze %r5, %r5 " ); \
302 asm( "addi %r4, %r4, 8 " ); \
303 asm( "addi %r3, %r3, 8 " ); \
304 asm( "std %%r5, %0 " : "=m" (c)); \
305 asm( "std %%r4, %0 " : "=m" (d)); \
306 asm( "std %%r3, %0 " : "=m" (s) :: \
307 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
308
309#endif
310
311#else /* PPC32 */
312
313#if defined(__MACH__) && defined(__APPLE__)
314
315#define MULADDC_INIT \
316 asm( "lwz r3, %0 " :: "m" (s)); \
317 asm( "lwz r4, %0 " :: "m" (d)); \
318 asm( "lwz r5, %0 " :: "m" (c)); \
319 asm( "lwz r6, %0 " :: "m" (b)); \
320 asm( "addi r3, r3, -4 " ); \
321 asm( "addi r4, r4, -4 " ); \
322 asm( "addic r5, r5, 0 " );
323
324#define MULADDC_CORE \
325 asm( "lwzu r7, 4(r3) " ); \
326 asm( "mullw r8, r7, r6 " ); \
327 asm( "mulhwu r9, r7, r6 " ); \
328 asm( "adde r8, r8, r5 " ); \
329 asm( "lwz r7, 4(r4) " ); \
330 asm( "addze r5, r9 " ); \
331 asm( "addc r8, r8, r7 " ); \
332 asm( "stwu r8, 4(r4) " );
333
334#define MULADDC_STOP \
335 asm( "addze r5, r5 " ); \
336 asm( "addi r4, r4, 4 " ); \
337 asm( "addi r3, r3, 4 " ); \
338 asm( "stw r5, %0 " : "=m" (c)); \
339 asm( "stw r4, %0 " : "=m" (d)); \
340 asm( "stw r3, %0 " : "=m" (s) :: \
341 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
342
343#else
344
345#define MULADDC_INIT \
346 asm( "lwz %%r3, %0 " :: "m" (s)); \
347 asm( "lwz %%r4, %0 " :: "m" (d)); \
348 asm( "lwz %%r5, %0 " :: "m" (c)); \
349 asm( "lwz %%r6, %0 " :: "m" (b)); \
350 asm( "addi %r3, %r3, -4 " ); \
351 asm( "addi %r4, %r4, -4 " ); \
352 asm( "addic %r5, %r5, 0 " );
353
354#define MULADDC_CORE \
355 asm( "lwzu %r7, 4(%r3) " ); \
356 asm( "mullw %r8, %r7, %r6 " ); \
357 asm( "mulhwu %r9, %r7, %r6 " ); \
358 asm( "adde %r8, %r8, %r5 " ); \
359 asm( "lwz %r7, 4(%r4) " ); \
360 asm( "addze %r5, %r9 " ); \
361 asm( "addc %r8, %r8, %r7 " ); \
362 asm( "stwu %r8, 4(%r4) " );
363
364#define MULADDC_STOP \
365 asm( "addze %r5, %r5 " ); \
366 asm( "addi %r4, %r4, 4 " ); \
367 asm( "addi %r3, %r3, 4 " ); \
368 asm( "stw %%r5, %0 " : "=m" (c)); \
369 asm( "stw %%r4, %0 " : "=m" (d)); \
370 asm( "stw %%r3, %0 " : "=m" (s) :: \
371 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
372
373#endif
374
375#endif /* PPC32 */
376#endif /* PPC64 */
377
378#if defined(__sparc__)
379
380#define MULADDC_INIT \
381 asm( "ld %0, %%o0 " :: "m" (s)); \
382 asm( "ld %0, %%o1 " :: "m" (d)); \
383 asm( "ld %0, %%o2 " :: "m" (c)); \
384 asm( "ld %0, %%o3 " :: "m" (b));
385
386#define MULADDC_CORE \
387 asm( "ld [%o0], %o4 " ); \
388 asm( "inc 4, %o0 " ); \
389 asm( "ld [%o1], %o5 " ); \
390 asm( "umul %o3, %o4, %o4 " ); \
391 asm( "addcc %o4, %o2, %o4 " ); \
392 asm( "rd %y, %g1 " ); \
393 asm( "addx %g1, 0, %g1 " ); \
394 asm( "addcc %o4, %o5, %o4 " ); \
395 asm( "st %o4, [%o1] " ); \
396 asm( "addx %g1, 0, %o2 " ); \
397 asm( "inc 4, %o1 " );
398
399#define MULADDC_STOP \
400 asm( "st %%o2, %0 " : "=m" (c)); \
401 asm( "st %%o1, %0 " : "=m" (d)); \
402 asm( "st %%o0, %0 " : "=m" (s) :: \
403 "g1", "o0", "o1", "o2", "o3", "o4", "o5" );
404
405#endif /* SPARCv8 */
406
407#if defined(__microblaze__) || defined(microblaze)
408
409#define MULADDC_INIT \
410 asm( "lwi r3, %0 " :: "m" (s)); \
411 asm( "lwi r4, %0 " :: "m" (d)); \
412 asm( "lwi r5, %0 " :: "m" (c)); \
413 asm( "lwi r6, %0 " :: "m" (b)); \
414 asm( "andi r7, r6, 0xffff" ); \
415 asm( "bsrli r6, r6, 16 " );
416
417#define MULADDC_CORE \
418 asm( "lhui r8, r3, 0 " ); \
419 asm( "addi r3, r3, 2 " ); \
420 asm( "lhui r9, r3, 0 " ); \
421 asm( "addi r3, r3, 2 " ); \
422 asm( "mul r10, r9, r6 " ); \
423 asm( "mul r11, r8, r7 " ); \
424 asm( "mul r12, r9, r7 " ); \
425 asm( "mul r13, r8, r6 " ); \
426 asm( "bsrli r8, r10, 16 " ); \
427 asm( "bsrli r9, r11, 16 " ); \
428 asm( "add r13, r13, r8 " ); \
429 asm( "add r13, r13, r9 " ); \
430 asm( "bslli r10, r10, 16 " ); \
431 asm( "bslli r11, r11, 16 " ); \
432 asm( "add r12, r12, r10 " ); \
433 asm( "addc r13, r13, r0 " ); \
434 asm( "add r12, r12, r11 " ); \
435 asm( "addc r13, r13, r0 " ); \
436 asm( "lwi r10, r4, 0 " ); \
437 asm( "add r12, r12, r10 " ); \
438 asm( "addc r13, r13, r0 " ); \
439 asm( "add r12, r12, r5 " ); \
440 asm( "addc r5, r13, r0 " ); \
441 asm( "swi r12, r4, 0 " ); \
442 asm( "addi r4, r4, 4 " );
443
444#define MULADDC_STOP \
445 asm( "swi r5, %0 " : "=m" (c)); \
446 asm( "swi r4, %0 " : "=m" (d)); \
447 asm( "swi r3, %0 " : "=m" (s) :: \
448 "r3", "r4" , "r5" , "r6" , "r7" , "r8" , \
449 "r9", "r10", "r11", "r12", "r13" );
450
451#endif /* MicroBlaze */
452
453#if defined(__tricore__)
454
455#define MULADDC_INIT \
456 asm( "ld.a %%a2, %0 " :: "m" (s)); \
457 asm( "ld.a %%a3, %0 " :: "m" (d)); \
458 asm( "ld.w %%d4, %0 " :: "m" (c)); \
459 asm( "ld.w %%d1, %0 " :: "m" (b)); \
460 asm( "xor %d5, %d5 " );
461
462#define MULADDC_CORE \
463 asm( "ld.w %d0, [%a2+] " ); \
464 asm( "madd.u %e2, %e4, %d0, %d1 " ); \
465 asm( "ld.w %d0, [%a3] " ); \
466 asm( "addx %d2, %d2, %d0 " ); \
467 asm( "addc %d3, %d3, 0 " ); \
468 asm( "mov %d4, %d3 " ); \
469 asm( "st.w [%a3+], %d2 " );
470
471#define MULADDC_STOP \
472 asm( "st.w %0, %%d4 " : "=m" (c)); \
473 asm( "st.a %0, %%a3 " : "=m" (d)); \
474 asm( "st.a %0, %%a2 " : "=m" (s) :: \
475 "d0", "d1", "e2", "d4", "a2", "a3" );
476
477#endif /* TriCore */
478
479#if defined(__arm__)
480
481#define MULADDC_INIT \
482 asm( "ldr r0, %0 " :: "m" (s)); \
483 asm( "ldr r1, %0 " :: "m" (d)); \
484 asm( "ldr r2, %0 " :: "m" (c)); \
485 asm( "ldr r3, %0 " :: "m" (b));
486
487#define MULADDC_CORE \
488 asm( "ldr r4, [r0], #4 " ); \
489 asm( "mov r5, #0 " ); \
490 asm( "ldr r6, [r1] " ); \
491 asm( "umlal r2, r5, r3, r4 " ); \
492 asm( "adds r7, r6, r2 " ); \
493 asm( "adc r2, r5, #0 " ); \
494 asm( "str r7, [r1], #4 " );
495
496#define MULADDC_STOP \
497 asm( "str r2, %0 " : "=m" (c)); \
498 asm( "str r1, %0 " : "=m" (d)); \
499 asm( "str r0, %0 " : "=m" (s) :: \
500 "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7" );
501
502#endif /* ARMv3 */
503
504#if defined(__alpha__)
505
506#define MULADDC_INIT \
507 asm( "ldq $1, %0 " :: "m" (s)); \
508 asm( "ldq $2, %0 " :: "m" (d)); \
509 asm( "ldq $3, %0 " :: "m" (c)); \
510 asm( "ldq $4, %0 " :: "m" (b));
511
512#define MULADDC_CORE \
513 asm( "ldq $6, 0($1) " ); \
514 asm( "addq $1, 8, $1 " ); \
515 asm( "mulq $6, $4, $7 " ); \
516 asm( "umulh $6, $4, $6 " ); \
517 asm( "addq $7, $3, $7 " ); \
518 asm( "cmpult $7, $3, $3 " ); \
519 asm( "ldq $5, 0($2) " ); \
520 asm( "addq $7, $5, $7 " ); \
521 asm( "cmpult $7, $5, $5 " ); \
522 asm( "stq $7, 0($2) " ); \
523 asm( "addq $2, 8, $2 " ); \
524 asm( "addq $6, $3, $3 " ); \
525 asm( "addq $5, $3, $3 " );
526
527#define MULADDC_STOP \
528 asm( "stq $3, %0 " : "=m" (c)); \
529 asm( "stq $2, %0 " : "=m" (d)); \
530 asm( "stq $1, %0 " : "=m" (s) :: \
531 "$1", "$2", "$3", "$4", "$5", "$6", "$7" );
532
533#endif /* Alpha */
534
535#if defined(__mips__)
536
537#define MULADDC_INIT \
538 asm( "lw $10, %0 " :: "m" (s)); \
539 asm( "lw $11, %0 " :: "m" (d)); \
540 asm( "lw $12, %0 " :: "m" (c)); \
541 asm( "lw $13, %0 " :: "m" (b));
542
543#define MULADDC_CORE \
544 asm( "lw $14, 0($10) " ); \
545 asm( "multu $13, $14 " ); \
546 asm( "addi $10, $10, 4 " ); \
547 asm( "mflo $14 " ); \
548 asm( "mfhi $9 " ); \
549 asm( "addu $14, $12, $14 " ); \
550 asm( "lw $15, 0($11) " ); \
551 asm( "sltu $12, $14, $12 " ); \
552 asm( "addu $15, $14, $15 " ); \
553 asm( "sltu $14, $15, $14 " ); \
554 asm( "addu $12, $12, $9 " ); \
555 asm( "sw $15, 0($11) " ); \
556 asm( "addu $12, $12, $14 " ); \
557 asm( "addi $11, $11, 4 " );
558
559#define MULADDC_STOP \
560 asm( "sw $12, %0 " : "=m" (c)); \
561 asm( "sw $11, %0 " : "=m" (d)); \
562 asm( "sw $10, %0 " : "=m" (s) :: \
563 "$9", "$10", "$11", "$12", "$13", "$14", "$15" );
564
565#endif /* MIPS */
566#endif /* GNUC */
567
568#if (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__)
569
570#define MULADDC_INIT \
571 __asm mov esi, s \
572 __asm mov edi, d \
573 __asm mov ecx, c \
574 __asm mov ebx, b
575
576#define MULADDC_CORE \
577 __asm lodsd \
578 __asm mul ebx \
579 __asm add eax, ecx \
580 __asm adc edx, 0 \
581 __asm add eax, [edi] \
582 __asm adc edx, 0 \
583 __asm mov ecx, edx \
584 __asm stosd
585
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]586#if defined(POLARSSL_HAVE_SSE2)
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]587
588#define EMIT __asm _emit
589
590#define MULADDC_HUIT \
591 EMIT 0x0F EMIT 0x6E EMIT 0xC9 \
592 EMIT 0x0F EMIT 0x6E EMIT 0xC3 \
593 EMIT 0x0F EMIT 0x6E EMIT 0x1F \
594 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
595 EMIT 0x0F EMIT 0x6E EMIT 0x16 \
596 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
597 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x04 \
598 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
599 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x08 \
600 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
601 EMIT 0x0F EMIT 0x6E EMIT 0x7E EMIT 0x0C \
602 EMIT 0x0F EMIT 0xF4 EMIT 0xF8 \
603 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
604 EMIT 0x0F EMIT 0x6E EMIT 0x5F EMIT 0x04 \
605 EMIT 0x0F EMIT 0xD4 EMIT 0xDC \
606 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x08 \
607 EMIT 0x0F EMIT 0xD4 EMIT 0xEE \
608 EMIT 0x0F EMIT 0x6E EMIT 0x67 EMIT 0x0C \
609 EMIT 0x0F EMIT 0xD4 EMIT 0xFC \
610 EMIT 0x0F EMIT 0x7E EMIT 0x0F \
611 EMIT 0x0F EMIT 0x6E EMIT 0x56 EMIT 0x10 \
612 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
613 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
614 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x14 \
615 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
616 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
617 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x18 \
618 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
619 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x04 \
620 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
621 EMIT 0x0F EMIT 0x6E EMIT 0x5E EMIT 0x1C \
622 EMIT 0x0F EMIT 0xF4 EMIT 0xD8 \
623 EMIT 0x0F EMIT 0xD4 EMIT 0xCD \
624 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x10 \
625 EMIT 0x0F EMIT 0xD4 EMIT 0xD5 \
626 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x08 \
627 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
628 EMIT 0x0F EMIT 0xD4 EMIT 0xCF \
629 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x14 \
630 EMIT 0x0F EMIT 0xD4 EMIT 0xE5 \
631 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x0C \
632 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
633 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
634 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x18 \
635 EMIT 0x0F EMIT 0xD4 EMIT 0xF5 \
636 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x10 \
637 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
638 EMIT 0x0F EMIT 0xD4 EMIT 0xCC \
639 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x1C \
640 EMIT 0x0F EMIT 0xD4 EMIT 0xDD \
641 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x14 \
642 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
643 EMIT 0x0F EMIT 0xD4 EMIT 0xCE \
644 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x18 \
645 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
646 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
647 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x1C \
648 EMIT 0x83 EMIT 0xC7 EMIT 0x20 \
649 EMIT 0x83 EMIT 0xC6 EMIT 0x20 \
650 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
651 EMIT 0x0F EMIT 0x7E EMIT 0xC9
652
653#define MULADDC_STOP \
654 EMIT 0x0F EMIT 0x77 \
655 __asm mov c, ecx \
656 __asm mov d, edi \
657 __asm mov s, esi \
658
659#else
660
661#define MULADDC_STOP \
662 __asm mov c, ecx \
663 __asm mov d, edi \
664 __asm mov s, esi \
665
666#endif /* SSE2 */
667#endif /* MSVC */
668
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]669#endif /* POLARSSL_HAVE_ASM */
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]670
671#if !defined(MULADDC_CORE)
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]672#if defined(POLARSSL_HAVE_LONGLONG)
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]673
674#define MULADDC_INIT \
675{ \
676 t_dbl r; \
677 t_int r0, r1;
678
679#define MULADDC_CORE \
680 r = *(s++) * (t_dbl) b; \
681 r0 = r; \
682 r1 = r >> biL; \
683 r0 += c; r1 += (r0 < c); \
684 r0 += *d; r1 += (r0 < *d); \
685 c = r1; *(d++) = r0;
686
687#define MULADDC_STOP \
688}
689
690#else
691#define MULADDC_INIT \
692{ \
693 t_int s0, s1, b0, b1; \
694 t_int r0, r1, rx, ry; \
695 b0 = ( b << biH ) >> biH; \
696 b1 = ( b >> biH );
697
698#define MULADDC_CORE \
699 s0 = ( *s << biH ) >> biH; \
700 s1 = ( *s >> biH ); s++; \
701 rx = s0 * b1; r0 = s0 * b0; \
702 ry = s1 * b0; r1 = s1 * b1; \
703 r1 += ( rx >> biH ); \
704 r1 += ( ry >> biH ); \
705 rx <<= biH; ry <<= biH; \
706 r0 += rx; r1 += (r0 < rx); \
707 r0 += ry; r1 += (r0 < ry); \
708 r0 += c; r1 += (r0 < c); \
709 r0 += *d; r1 += (r0 < *d); \
710 c = r1; *(d++) = r0;
711
712#define MULADDC_STOP \
713}
714
715#endif /* C (generic) */
716#endif /* C (longlong) */
717
718#endif /* bn_mul.h */