TrustedFirmware Git Browser
Code Review Sign In
review.trustedfirmware.org / mirror / mbed-tls / refs/heads/polarssl-0.14 / . / include / polarssl / bn_mul.h
blob: a73d5fb0ff0c7d528e991e5416d430f6fcbaaed2 [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 *
Paul Bakker84f12b72010-07-18 10:13:04 +0000[diff] [blame]4 * Copyright (C) 2006-2010, Brainspark B.V.
Paul Bakkerb96f1542010-07-18 20:36:00 +0000[diff] [blame]5 *
6 * This file is part of PolarSSL (http://www.polarssl.org)
Paul Bakker84f12b72010-07-18 10:13:04 +0000[diff] [blame]7 * Lead Maintainer: Paul Bakker <polarssl_maintainer at polarssl.org>
Paul Bakkerb96f1542010-07-18 20:36:00 +0000[diff] [blame]8 *
Paul Bakker77b385e2009-07-28 17:23:11 +0000[diff] [blame]9 * All rights reserved.
Paul Bakkere0ccd0a2009-01-04 16:27:10 +0000[diff] [blame]10 *
Paul Bakkere0ccd0a2009-01-04 16:27:10 +0000[diff] [blame]11 * This program is free software; you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License as published by
13 * the Free Software Foundation; either version 2 of the License, or
14 * (at your option) any later version.
15 *
16 * This program is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19 * GNU General Public License for more details.
20 *
21 * You should have received a copy of the GNU General Public License along
22 * with this program; if not, write to the Free Software Foundation, Inc.,
23 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]24 */
25/*
26 * Multiply source vector [s] with b, add result
27 * to destination vector [d] and set carry c.
28 *
29 * Currently supports:
30 *
31 * . IA-32 (386+) . AMD64 / EM64T
32 * . IA-32 (SSE2) . Motorola 68000
33 * . PowerPC, 32-bit . MicroBlaze
34 * . PowerPC, 64-bit . TriCore
35 * . SPARC v8 . ARM v3+
36 * . Alpha . MIPS32
37 * . C, longlong . C, generic
38 */
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]39#ifndef POLARSSL_BN_MUL_H
40#define POLARSSL_BN_MUL_H
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]41
Paul Bakker8e831ed2009-01-03 21:24:11 +0000[diff] [blame]42#include "polarssl/config.h"
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]43
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]44#if defined(POLARSSL_HAVE_ASM)
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]45
46#if defined(__GNUC__)
47#if defined(__i386__)
48
Paul Bakkerc89cf7c2009-07-19 21:37:39 +0000[diff] [blame]49#define MULADDC_INIT \
50 asm( " \
51 movl %%ebx, %0; \
52 movl %5, %%esi; \
53 movl %6, %%edi; \
54 movl %7, %%ecx; \
55 movl %8, %%ebx; \
56 "
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]57
Paul Bakker40fe2b02009-07-08 19:49:01 +0000[diff] [blame]58#define MULADDC_CORE \
Paul Bakkerc89cf7c2009-07-19 21:37:39 +0000[diff] [blame]59 " \
Paul Bakker40fe2b02009-07-08 19:49:01 +0000[diff] [blame]60 lodsl; \
61 mull %%ebx; \
62 addl %%ecx, %%eax; \
63 adcl $0, %%edx; \
64 addl (%%edi), %%eax; \
65 adcl $0, %%edx; \
66 movl %%edx, %%ecx; \
Paul Bakkerc89cf7c2009-07-19 21:37:39 +0000[diff] [blame]67 stosl; \
68 "
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]69
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]70#if defined(POLARSSL_HAVE_SSE2)
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]71
Paul Bakkerc89cf7c2009-07-19 21:37:39 +0000[diff] [blame]72#define MULADDC_HUIT \
73 " \
74 movd %%ecx, %%mm1; \
75 movd %%ebx, %%mm0; \
76 movd (%%edi), %%mm3; \
77 paddq %%mm3, %%mm1; \
78 movd (%%esi), %%mm2; \
79 pmuludq %%mm0, %%mm2; \
80 movd 4(%%esi), %%mm4; \
81 pmuludq %%mm0, %%mm4; \
82 movd 8(%%esi), %%mm6; \
83 pmuludq %%mm0, %%mm6; \
84 movd 12(%%esi), %%mm7; \
85 pmuludq %%mm0, %%mm7; \
86 paddq %%mm2, %%mm1; \
87 movd 4(%%edi), %%mm3; \
88 paddq %%mm4, %%mm3; \
89 movd 8(%%edi), %%mm5; \
90 paddq %%mm6, %%mm5; \
91 movd 12(%%edi), %%mm4; \
92 paddq %%mm4, %%mm7; \
93 movd %%mm1, (%%edi); \
94 movd 16(%%esi), %%mm2; \
95 pmuludq %%mm0, %%mm2; \
96 psrlq $32, %%mm1; \
97 movd 20(%%esi), %%mm4; \
98 pmuludq %%mm0, %%mm4; \
99 paddq %%mm3, %%mm1; \
100 movd 24(%%esi), %%mm6; \
101 pmuludq %%mm0, %%mm6; \
102 movd %%mm1, 4(%%edi); \
103 psrlq $32, %%mm1; \
104 movd 28(%%esi), %%mm3; \
105 pmuludq %%mm0, %%mm3; \
106 paddq %%mm5, %%mm1; \
107 movd 16(%%edi), %%mm5; \
108 paddq %%mm5, %%mm2; \
109 movd %%mm1, 8(%%edi); \
110 psrlq $32, %%mm1; \
111 paddq %%mm7, %%mm1; \
112 movd 20(%%edi), %%mm5; \
113 paddq %%mm5, %%mm4; \
114 movd %%mm1, 12(%%edi); \
115 psrlq $32, %%mm1; \
116 paddq %%mm2, %%mm1; \
117 movd 24(%%edi), %%mm5; \
118 paddq %%mm5, %%mm6; \
119 movd %%mm1, 16(%%edi); \
120 psrlq $32, %%mm1; \
121 paddq %%mm4, %%mm1; \
122 movd 28(%%edi), %%mm5; \
123 paddq %%mm5, %%mm3; \
124 movd %%mm1, 20(%%edi); \
125 psrlq $32, %%mm1; \
126 paddq %%mm6, %%mm1; \
127 movd %%mm1, 24(%%edi); \
128 psrlq $32, %%mm1; \
129 paddq %%mm3, %%mm1; \
130 movd %%mm1, 28(%%edi); \
131 addl $32, %%edi; \
132 addl $32, %%esi; \
133 psrlq $32, %%mm1; \
134 movd %%mm1, %%ecx; \
135 "
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]136
Paul Bakkerc89cf7c2009-07-19 21:37:39 +0000[diff] [blame]137#define MULADDC_STOP \
138 " \
139 emms; \
140 movl %4, %%ebx; \
141 movl %%ecx, %1; \
142 movl %%edi, %2; \
143 movl %%esi, %3; \
144 " \
145 : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \
146 : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \
147 : "eax", "ecx", "edx", "esi", "edi" \
148 );
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]149
150#else
151
Paul Bakkerc89cf7c2009-07-19 21:37:39 +0000[diff] [blame]152#define MULADDC_STOP \
153 " \
154 movl %4, %%ebx; \
155 movl %%ecx, %1; \
156 movl %%edi, %2; \
157 movl %%esi, %3; \
158 " \
159 : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \
160 : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \
161 : "eax", "ecx", "edx", "esi", "edi" \
162 );
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]163#endif /* SSE2 */
164#endif /* i386 */
165
166#if defined(__amd64__) || defined (__x86_64__)
167
168#define MULADDC_INIT \
169 asm( "movq %0, %%rsi " :: "m" (s)); \
170 asm( "movq %0, %%rdi " :: "m" (d)); \
171 asm( "movq %0, %%rcx " :: "m" (c)); \
172 asm( "movq %0, %%rbx " :: "m" (b)); \
173 asm( "xorq %r8, %r8 " );
174
175#define MULADDC_CORE \
176 asm( "movq (%rsi),%rax " ); \
177 asm( "mulq %rbx " ); \
178 asm( "addq $8, %rsi " ); \
179 asm( "addq %rcx, %rax " ); \
180 asm( "movq %r8, %rcx " ); \
181 asm( "adcq $0, %rdx " ); \
182 asm( "nop " ); \
183 asm( "addq %rax, (%rdi) " ); \
184 asm( "adcq %rdx, %rcx " ); \
185 asm( "addq $8, %rdi " );
186
187#define MULADDC_STOP \
188 asm( "movq %%rcx, %0 " : "=m" (c)); \
189 asm( "movq %%rdi, %0 " : "=m" (d)); \
190 asm( "movq %%rsi, %0 " : "=m" (s) :: \
191 "rax", "rcx", "rdx", "rbx", "rsi", "rdi", "r8" );
192
193#endif /* AMD64 */
194
195#if defined(__mc68020__) || defined(__mcpu32__)
196
197#define MULADDC_INIT \
198 asm( "movl %0, %%a2 " :: "m" (s)); \
199 asm( "movl %0, %%a3 " :: "m" (d)); \
200 asm( "movl %0, %%d3 " :: "m" (c)); \
201 asm( "movl %0, %%d2 " :: "m" (b)); \
202 asm( "moveq #0, %d0 " );
203
204#define MULADDC_CORE \
205 asm( "movel %a2@+, %d1 " ); \
206 asm( "mulul %d2, %d4:%d1 " ); \
207 asm( "addl %d3, %d1 " ); \
208 asm( "addxl %d0, %d4 " ); \
209 asm( "moveq #0, %d3 " ); \
210 asm( "addl %d1, %a3@+ " ); \
211 asm( "addxl %d4, %d3 " );
212
213#define MULADDC_STOP \
214 asm( "movl %%d3, %0 " : "=m" (c)); \
215 asm( "movl %%a3, %0 " : "=m" (d)); \
216 asm( "movl %%a2, %0 " : "=m" (s) :: \
217 "d0", "d1", "d2", "d3", "d4", "a2", "a3" );
218
219#define MULADDC_HUIT \
220 asm( "movel %a2@+, %d1 " ); \
221 asm( "mulul %d2, %d4:%d1 " ); \
222 asm( "addxl %d3, %d1 " ); \
223 asm( "addxl %d0, %d4 " ); \
224 asm( "addl %d1, %a3@+ " ); \
225 asm( "movel %a2@+, %d1 " ); \
226 asm( "mulul %d2, %d3:%d1 " ); \
227 asm( "addxl %d4, %d1 " ); \
228 asm( "addxl %d0, %d3 " ); \
229 asm( "addl %d1, %a3@+ " ); \
230 asm( "movel %a2@+, %d1 " ); \
231 asm( "mulul %d2, %d4:%d1 " ); \
232 asm( "addxl %d3, %d1 " ); \
233 asm( "addxl %d0, %d4 " ); \
234 asm( "addl %d1, %a3@+ " ); \
235 asm( "movel %a2@+, %d1 " ); \
236 asm( "mulul %d2, %d3:%d1 " ); \
237 asm( "addxl %d4, %d1 " ); \
238 asm( "addxl %d0, %d3 " ); \
239 asm( "addl %d1, %a3@+ " ); \
240 asm( "movel %a2@+, %d1 " ); \
241 asm( "mulul %d2, %d4:%d1 " ); \
242 asm( "addxl %d3, %d1 " ); \
243 asm( "addxl %d0, %d4 " ); \
244 asm( "addl %d1, %a3@+ " ); \
245 asm( "movel %a2@+, %d1 " ); \
246 asm( "mulul %d2, %d3:%d1 " ); \
247 asm( "addxl %d4, %d1 " ); \
248 asm( "addxl %d0, %d3 " ); \
249 asm( "addl %d1, %a3@+ " ); \
250 asm( "movel %a2@+, %d1 " ); \
251 asm( "mulul %d2, %d4:%d1 " ); \
252 asm( "addxl %d3, %d1 " ); \
253 asm( "addxl %d0, %d4 " ); \
254 asm( "addl %d1, %a3@+ " ); \
255 asm( "movel %a2@+, %d1 " ); \
256 asm( "mulul %d2, %d3:%d1 " ); \
257 asm( "addxl %d4, %d1 " ); \
258 asm( "addxl %d0, %d3 " ); \
259 asm( "addl %d1, %a3@+ " ); \
260 asm( "addxl %d0, %d3 " );
261
262#endif /* MC68000 */
263
264#if defined(__powerpc__) || defined(__ppc__)
265#if defined(__powerpc64__) || defined(__ppc64__)
266
267#if defined(__MACH__) && defined(__APPLE__)
268
269#define MULADDC_INIT \
270 asm( "ld r3, %0 " :: "m" (s)); \
271 asm( "ld r4, %0 " :: "m" (d)); \
272 asm( "ld r5, %0 " :: "m" (c)); \
273 asm( "ld r6, %0 " :: "m" (b)); \
274 asm( "addi r3, r3, -8 " ); \
275 asm( "addi r4, r4, -8 " ); \
276 asm( "addic r5, r5, 0 " );
277
278#define MULADDC_CORE \
279 asm( "ldu r7, 8(r3) " ); \
280 asm( "mulld r8, r7, r6 " ); \
281 asm( "mulhdu r9, r7, r6 " ); \
282 asm( "adde r8, r8, r5 " ); \
283 asm( "ld r7, 8(r4) " ); \
284 asm( "addze r5, r9 " ); \
285 asm( "addc r8, r8, r7 " ); \
286 asm( "stdu r8, 8(r4) " );
287
288#define MULADDC_STOP \
289 asm( "addze r5, r5 " ); \
290 asm( "addi r4, r4, 8 " ); \
291 asm( "addi r3, r3, 8 " ); \
292 asm( "std r5, %0 " : "=m" (c)); \
293 asm( "std r4, %0 " : "=m" (d)); \
294 asm( "std r3, %0 " : "=m" (s) :: \
295 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
296
297#else
298
299#define MULADDC_INIT \
300 asm( "ld %%r3, %0 " :: "m" (s)); \
301 asm( "ld %%r4, %0 " :: "m" (d)); \
302 asm( "ld %%r5, %0 " :: "m" (c)); \
303 asm( "ld %%r6, %0 " :: "m" (b)); \
304 asm( "addi %r3, %r3, -8 " ); \
305 asm( "addi %r4, %r4, -8 " ); \
306 asm( "addic %r5, %r5, 0 " );
307
308#define MULADDC_CORE \
309 asm( "ldu %r7, 8(%r3) " ); \
310 asm( "mulld %r8, %r7, %r6 " ); \
311 asm( "mulhdu %r9, %r7, %r6 " ); \
312 asm( "adde %r8, %r8, %r5 " ); \
313 asm( "ld %r7, 8(%r4) " ); \
314 asm( "addze %r5, %r9 " ); \
315 asm( "addc %r8, %r8, %r7 " ); \
316 asm( "stdu %r8, 8(%r4) " );
317
318#define MULADDC_STOP \
319 asm( "addze %r5, %r5 " ); \
320 asm( "addi %r4, %r4, 8 " ); \
321 asm( "addi %r3, %r3, 8 " ); \
322 asm( "std %%r5, %0 " : "=m" (c)); \
323 asm( "std %%r4, %0 " : "=m" (d)); \
324 asm( "std %%r3, %0 " : "=m" (s) :: \
325 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
326
327#endif
328
329#else /* PPC32 */
330
331#if defined(__MACH__) && defined(__APPLE__)
332
333#define MULADDC_INIT \
334 asm( "lwz r3, %0 " :: "m" (s)); \
335 asm( "lwz r4, %0 " :: "m" (d)); \
336 asm( "lwz r5, %0 " :: "m" (c)); \
337 asm( "lwz r6, %0 " :: "m" (b)); \
338 asm( "addi r3, r3, -4 " ); \
339 asm( "addi r4, r4, -4 " ); \
340 asm( "addic r5, r5, 0 " );
341
342#define MULADDC_CORE \
343 asm( "lwzu r7, 4(r3) " ); \
344 asm( "mullw r8, r7, r6 " ); \
345 asm( "mulhwu r9, r7, r6 " ); \
346 asm( "adde r8, r8, r5 " ); \
347 asm( "lwz r7, 4(r4) " ); \
348 asm( "addze r5, r9 " ); \
349 asm( "addc r8, r8, r7 " ); \
350 asm( "stwu r8, 4(r4) " );
351
352#define MULADDC_STOP \
353 asm( "addze r5, r5 " ); \
354 asm( "addi r4, r4, 4 " ); \
355 asm( "addi r3, r3, 4 " ); \
356 asm( "stw r5, %0 " : "=m" (c)); \
357 asm( "stw r4, %0 " : "=m" (d)); \
358 asm( "stw r3, %0 " : "=m" (s) :: \
359 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
360
361#else
362
363#define MULADDC_INIT \
364 asm( "lwz %%r3, %0 " :: "m" (s)); \
365 asm( "lwz %%r4, %0 " :: "m" (d)); \
366 asm( "lwz %%r5, %0 " :: "m" (c)); \
367 asm( "lwz %%r6, %0 " :: "m" (b)); \
368 asm( "addi %r3, %r3, -4 " ); \
369 asm( "addi %r4, %r4, -4 " ); \
370 asm( "addic %r5, %r5, 0 " );
371
372#define MULADDC_CORE \
373 asm( "lwzu %r7, 4(%r3) " ); \
374 asm( "mullw %r8, %r7, %r6 " ); \
375 asm( "mulhwu %r9, %r7, %r6 " ); \
376 asm( "adde %r8, %r8, %r5 " ); \
377 asm( "lwz %r7, 4(%r4) " ); \
378 asm( "addze %r5, %r9 " ); \
379 asm( "addc %r8, %r8, %r7 " ); \
380 asm( "stwu %r8, 4(%r4) " );
381
382#define MULADDC_STOP \
383 asm( "addze %r5, %r5 " ); \
384 asm( "addi %r4, %r4, 4 " ); \
385 asm( "addi %r3, %r3, 4 " ); \
386 asm( "stw %%r5, %0 " : "=m" (c)); \
387 asm( "stw %%r4, %0 " : "=m" (d)); \
388 asm( "stw %%r3, %0 " : "=m" (s) :: \
389 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
390
391#endif
392
393#endif /* PPC32 */
394#endif /* PPC64 */
395
396#if defined(__sparc__)
397
398#define MULADDC_INIT \
399 asm( "ld %0, %%o0 " :: "m" (s)); \
400 asm( "ld %0, %%o1 " :: "m" (d)); \
401 asm( "ld %0, %%o2 " :: "m" (c)); \
402 asm( "ld %0, %%o3 " :: "m" (b));
403
404#define MULADDC_CORE \
405 asm( "ld [%o0], %o4 " ); \
406 asm( "inc 4, %o0 " ); \
407 asm( "ld [%o1], %o5 " ); \
408 asm( "umul %o3, %o4, %o4 " ); \
409 asm( "addcc %o4, %o2, %o4 " ); \
410 asm( "rd %y, %g1 " ); \
411 asm( "addx %g1, 0, %g1 " ); \
412 asm( "addcc %o4, %o5, %o4 " ); \
413 asm( "st %o4, [%o1] " ); \
414 asm( "addx %g1, 0, %o2 " ); \
415 asm( "inc 4, %o1 " );
416
417#define MULADDC_STOP \
418 asm( "st %%o2, %0 " : "=m" (c)); \
419 asm( "st %%o1, %0 " : "=m" (d)); \
420 asm( "st %%o0, %0 " : "=m" (s) :: \
421 "g1", "o0", "o1", "o2", "o3", "o4", "o5" );
422
423#endif /* SPARCv8 */
424
425#if defined(__microblaze__) || defined(microblaze)
426
427#define MULADDC_INIT \
428 asm( "lwi r3, %0 " :: "m" (s)); \
429 asm( "lwi r4, %0 " :: "m" (d)); \
430 asm( "lwi r5, %0 " :: "m" (c)); \
431 asm( "lwi r6, %0 " :: "m" (b)); \
432 asm( "andi r7, r6, 0xffff" ); \
433 asm( "bsrli r6, r6, 16 " );
434
435#define MULADDC_CORE \
436 asm( "lhui r8, r3, 0 " ); \
437 asm( "addi r3, r3, 2 " ); \
438 asm( "lhui r9, r3, 0 " ); \
439 asm( "addi r3, r3, 2 " ); \
440 asm( "mul r10, r9, r6 " ); \
441 asm( "mul r11, r8, r7 " ); \
442 asm( "mul r12, r9, r7 " ); \
443 asm( "mul r13, r8, r6 " ); \
444 asm( "bsrli r8, r10, 16 " ); \
445 asm( "bsrli r9, r11, 16 " ); \
446 asm( "add r13, r13, r8 " ); \
447 asm( "add r13, r13, r9 " ); \
448 asm( "bslli r10, r10, 16 " ); \
449 asm( "bslli r11, r11, 16 " ); \
450 asm( "add r12, r12, r10 " ); \
451 asm( "addc r13, r13, r0 " ); \
452 asm( "add r12, r12, r11 " ); \
453 asm( "addc r13, r13, r0 " ); \
454 asm( "lwi r10, r4, 0 " ); \
455 asm( "add r12, r12, r10 " ); \
456 asm( "addc r13, r13, r0 " ); \
457 asm( "add r12, r12, r5 " ); \
458 asm( "addc r5, r13, r0 " ); \
459 asm( "swi r12, r4, 0 " ); \
460 asm( "addi r4, r4, 4 " );
461
462#define MULADDC_STOP \
463 asm( "swi r5, %0 " : "=m" (c)); \
464 asm( "swi r4, %0 " : "=m" (d)); \
465 asm( "swi r3, %0 " : "=m" (s) :: \
466 "r3", "r4" , "r5" , "r6" , "r7" , "r8" , \
467 "r9", "r10", "r11", "r12", "r13" );
468
469#endif /* MicroBlaze */
470
471#if defined(__tricore__)
472
473#define MULADDC_INIT \
474 asm( "ld.a %%a2, %0 " :: "m" (s)); \
475 asm( "ld.a %%a3, %0 " :: "m" (d)); \
476 asm( "ld.w %%d4, %0 " :: "m" (c)); \
477 asm( "ld.w %%d1, %0 " :: "m" (b)); \
478 asm( "xor %d5, %d5 " );
479
480#define MULADDC_CORE \
481 asm( "ld.w %d0, [%a2+] " ); \
482 asm( "madd.u %e2, %e4, %d0, %d1 " ); \
483 asm( "ld.w %d0, [%a3] " ); \
484 asm( "addx %d2, %d2, %d0 " ); \
485 asm( "addc %d3, %d3, 0 " ); \
486 asm( "mov %d4, %d3 " ); \
487 asm( "st.w [%a3+], %d2 " );
488
489#define MULADDC_STOP \
490 asm( "st.w %0, %%d4 " : "=m" (c)); \
491 asm( "st.a %0, %%a3 " : "=m" (d)); \
492 asm( "st.a %0, %%a2 " : "=m" (s) :: \
493 "d0", "d1", "e2", "d4", "a2", "a3" );
494
495#endif /* TriCore */
496
497#if defined(__arm__)
498
499#define MULADDC_INIT \
500 asm( "ldr r0, %0 " :: "m" (s)); \
501 asm( "ldr r1, %0 " :: "m" (d)); \
502 asm( "ldr r2, %0 " :: "m" (c)); \
503 asm( "ldr r3, %0 " :: "m" (b));
504
505#define MULADDC_CORE \
506 asm( "ldr r4, [r0], #4 " ); \
507 asm( "mov r5, #0 " ); \
508 asm( "ldr r6, [r1] " ); \
509 asm( "umlal r2, r5, r3, r4 " ); \
510 asm( "adds r7, r6, r2 " ); \
511 asm( "adc r2, r5, #0 " ); \
512 asm( "str r7, [r1], #4 " );
513
514#define MULADDC_STOP \
515 asm( "str r2, %0 " : "=m" (c)); \
516 asm( "str r1, %0 " : "=m" (d)); \
517 asm( "str r0, %0 " : "=m" (s) :: \
518 "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7" );
519
520#endif /* ARMv3 */
521
522#if defined(__alpha__)
523
524#define MULADDC_INIT \
525 asm( "ldq $1, %0 " :: "m" (s)); \
526 asm( "ldq $2, %0 " :: "m" (d)); \
527 asm( "ldq $3, %0 " :: "m" (c)); \
528 asm( "ldq $4, %0 " :: "m" (b));
529
530#define MULADDC_CORE \
531 asm( "ldq $6, 0($1) " ); \
532 asm( "addq $1, 8, $1 " ); \
533 asm( "mulq $6, $4, $7 " ); \
534 asm( "umulh $6, $4, $6 " ); \
535 asm( "addq $7, $3, $7 " ); \
536 asm( "cmpult $7, $3, $3 " ); \
537 asm( "ldq $5, 0($2) " ); \
538 asm( "addq $7, $5, $7 " ); \
539 asm( "cmpult $7, $5, $5 " ); \
540 asm( "stq $7, 0($2) " ); \
541 asm( "addq $2, 8, $2 " ); \
542 asm( "addq $6, $3, $3 " ); \
543 asm( "addq $5, $3, $3 " );
544
545#define MULADDC_STOP \
546 asm( "stq $3, %0 " : "=m" (c)); \
547 asm( "stq $2, %0 " : "=m" (d)); \
548 asm( "stq $1, %0 " : "=m" (s) :: \
549 "$1", "$2", "$3", "$4", "$5", "$6", "$7" );
550
551#endif /* Alpha */
552
553#if defined(__mips__)
554
555#define MULADDC_INIT \
556 asm( "lw $10, %0 " :: "m" (s)); \
557 asm( "lw $11, %0 " :: "m" (d)); \
558 asm( "lw $12, %0 " :: "m" (c)); \
559 asm( "lw $13, %0 " :: "m" (b));
560
561#define MULADDC_CORE \
562 asm( "lw $14, 0($10) " ); \
563 asm( "multu $13, $14 " ); \
564 asm( "addi $10, $10, 4 " ); \
565 asm( "mflo $14 " ); \
566 asm( "mfhi $9 " ); \
567 asm( "addu $14, $12, $14 " ); \
568 asm( "lw $15, 0($11) " ); \
569 asm( "sltu $12, $14, $12 " ); \
570 asm( "addu $15, $14, $15 " ); \
571 asm( "sltu $14, $15, $14 " ); \
572 asm( "addu $12, $12, $9 " ); \
573 asm( "sw $15, 0($11) " ); \
574 asm( "addu $12, $12, $14 " ); \
575 asm( "addi $11, $11, 4 " );
576
577#define MULADDC_STOP \
578 asm( "sw $12, %0 " : "=m" (c)); \
579 asm( "sw $11, %0 " : "=m" (d)); \
580 asm( "sw $10, %0 " : "=m" (s) :: \
581 "$9", "$10", "$11", "$12", "$13", "$14", "$15" );
582
583#endif /* MIPS */
584#endif /* GNUC */
585
586#if (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__)
587
588#define MULADDC_INIT \
589 __asm mov esi, s \
590 __asm mov edi, d \
591 __asm mov ecx, c \
592 __asm mov ebx, b
593
594#define MULADDC_CORE \
595 __asm lodsd \
596 __asm mul ebx \
597 __asm add eax, ecx \
598 __asm adc edx, 0 \
599 __asm add eax, [edi] \
600 __asm adc edx, 0 \
601 __asm mov ecx, edx \
602 __asm stosd
603
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]604#if defined(POLARSSL_HAVE_SSE2)
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]605
606#define EMIT __asm _emit
607
608#define MULADDC_HUIT \
609 EMIT 0x0F EMIT 0x6E EMIT 0xC9 \
610 EMIT 0x0F EMIT 0x6E EMIT 0xC3 \
611 EMIT 0x0F EMIT 0x6E EMIT 0x1F \
612 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
613 EMIT 0x0F EMIT 0x6E EMIT 0x16 \
614 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
615 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x04 \
616 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
617 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x08 \
618 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
619 EMIT 0x0F EMIT 0x6E EMIT 0x7E EMIT 0x0C \
620 EMIT 0x0F EMIT 0xF4 EMIT 0xF8 \
621 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
622 EMIT 0x0F EMIT 0x6E EMIT 0x5F EMIT 0x04 \
623 EMIT 0x0F EMIT 0xD4 EMIT 0xDC \
624 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x08 \
625 EMIT 0x0F EMIT 0xD4 EMIT 0xEE \
626 EMIT 0x0F EMIT 0x6E EMIT 0x67 EMIT 0x0C \
627 EMIT 0x0F EMIT 0xD4 EMIT 0xFC \
628 EMIT 0x0F EMIT 0x7E EMIT 0x0F \
629 EMIT 0x0F EMIT 0x6E EMIT 0x56 EMIT 0x10 \
630 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
631 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
632 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x14 \
633 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
634 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
635 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x18 \
636 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
637 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x04 \
638 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
639 EMIT 0x0F EMIT 0x6E EMIT 0x5E EMIT 0x1C \
640 EMIT 0x0F EMIT 0xF4 EMIT 0xD8 \
641 EMIT 0x0F EMIT 0xD4 EMIT 0xCD \
642 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x10 \
643 EMIT 0x0F EMIT 0xD4 EMIT 0xD5 \
644 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x08 \
645 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
646 EMIT 0x0F EMIT 0xD4 EMIT 0xCF \
647 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x14 \
648 EMIT 0x0F EMIT 0xD4 EMIT 0xE5 \
649 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x0C \
650 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
651 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
652 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x18 \
653 EMIT 0x0F EMIT 0xD4 EMIT 0xF5 \
654 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x10 \
655 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
656 EMIT 0x0F EMIT 0xD4 EMIT 0xCC \
657 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x1C \
658 EMIT 0x0F EMIT 0xD4 EMIT 0xDD \
659 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x14 \
660 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
661 EMIT 0x0F EMIT 0xD4 EMIT 0xCE \
662 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x18 \
663 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
664 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
665 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x1C \
666 EMIT 0x83 EMIT 0xC7 EMIT 0x20 \
667 EMIT 0x83 EMIT 0xC6 EMIT 0x20 \
668 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
669 EMIT 0x0F EMIT 0x7E EMIT 0xC9
670
671#define MULADDC_STOP \
672 EMIT 0x0F EMIT 0x77 \
673 __asm mov c, ecx \
674 __asm mov d, edi \
675 __asm mov s, esi \
676
677#else
678
679#define MULADDC_STOP \
680 __asm mov c, ecx \
681 __asm mov d, edi \
682 __asm mov s, esi \
683
684#endif /* SSE2 */
685#endif /* MSVC */
686
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]687#endif /* POLARSSL_HAVE_ASM */
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]688
689#if !defined(MULADDC_CORE)
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]690#if defined(POLARSSL_HAVE_LONGLONG)
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]691
692#define MULADDC_INIT \
693{ \
694 t_dbl r; \
695 t_int r0, r1;
696
697#define MULADDC_CORE \
698 r = *(s++) * (t_dbl) b; \
699 r0 = r; \
700 r1 = r >> biL; \
701 r0 += c; r1 += (r0 < c); \
702 r0 += *d; r1 += (r0 < *d); \
703 c = r1; *(d++) = r0;
704
705#define MULADDC_STOP \
706}
707
708#else
709#define MULADDC_INIT \
710{ \
711 t_int s0, s1, b0, b1; \
712 t_int r0, r1, rx, ry; \
713 b0 = ( b << biH ) >> biH; \
714 b1 = ( b >> biH );
715
716#define MULADDC_CORE \
717 s0 = ( *s << biH ) >> biH; \
718 s1 = ( *s >> biH ); s++; \
719 rx = s0 * b1; r0 = s0 * b0; \
720 ry = s1 * b0; r1 = s1 * b1; \
721 r1 += ( rx >> biH ); \
722 r1 += ( ry >> biH ); \
723 rx <<= biH; ry <<= biH; \
724 r0 += rx; r1 += (r0 < rx); \
725 r0 += ry; r1 += (r0 < ry); \
726 r0 += c; r1 += (r0 < c); \
727 r0 += *d; r1 += (r0 < *d); \
728 c = r1; *(d++) = r0;
729
730#define MULADDC_STOP \
731}
732
733#endif /* C (generic) */
734#endif /* C (longlong) */
735
736#endif /* bn_mul.h */