TrustedFirmware Git Browser
Code Review Sign In
review.trustedfirmware.org / mirror / mbed-tls / refs/heads/polarssl-0.13 / . / include / polarssl / bn_mul.h
blob: 4986b51fd0059d348ad98178323cf54518e73e41 [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 Bakker83ded912010-03-21 17:46:26 +0000[diff] [blame]4 * Copyright (C) 2006-2010, Paul Bakker <polarssl_maintainer at polarssl.org>
Paul Bakker77b385e2009-07-28 17:23:11 +0000[diff] [blame]5 * All rights reserved.
Paul Bakkere0ccd0a2009-01-04 16:27:10 +0000[diff] [blame]6 *
Paul Bakkere0ccd0a2009-01-04 16:27:10 +0000[diff] [blame]7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License along
18 * with this program; if not, write to the Free Software Foundation, Inc.,
19 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]20 */
21/*
22 * Multiply source vector [s] with b, add result
23 * to destination vector [d] and set carry c.
24 *
25 * Currently supports:
26 *
27 * . IA-32 (386+) . AMD64 / EM64T
28 * . IA-32 (SSE2) . Motorola 68000
29 * . PowerPC, 32-bit . MicroBlaze
30 * . PowerPC, 64-bit . TriCore
31 * . SPARC v8 . ARM v3+
32 * . Alpha . MIPS32
33 * . C, longlong . C, generic
34 */
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]35#ifndef POLARSSL_BN_MUL_H
36#define POLARSSL_BN_MUL_H
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]37
Paul Bakker8e831ed2009-01-03 21:24:11 +0000[diff] [blame]38#include "polarssl/config.h"
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]39
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]40#if defined(POLARSSL_HAVE_ASM)
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]41
42#if defined(__GNUC__)
43#if defined(__i386__)
44
Paul Bakkerc89cf7c2009-07-19 21:37:39 +0000[diff] [blame]45#define MULADDC_INIT \
46 asm( " \
47 movl %%ebx, %0; \
48 movl %5, %%esi; \
49 movl %6, %%edi; \
50 movl %7, %%ecx; \
51 movl %8, %%ebx; \
52 "
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]53
Paul Bakker40fe2b02009-07-08 19:49:01 +0000[diff] [blame]54#define MULADDC_CORE \
Paul Bakkerc89cf7c2009-07-19 21:37:39 +0000[diff] [blame]55 " \
Paul Bakker40fe2b02009-07-08 19:49:01 +0000[diff] [blame]56 lodsl; \
57 mull %%ebx; \
58 addl %%ecx, %%eax; \
59 adcl $0, %%edx; \
60 addl (%%edi), %%eax; \
61 adcl $0, %%edx; \
62 movl %%edx, %%ecx; \
Paul Bakkerc89cf7c2009-07-19 21:37:39 +0000[diff] [blame]63 stosl; \
64 "
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]65
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]66#if defined(POLARSSL_HAVE_SSE2)
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]67
Paul Bakkerc89cf7c2009-07-19 21:37:39 +0000[diff] [blame]68#define MULADDC_HUIT \
69 " \
70 movd %%ecx, %%mm1; \
71 movd %%ebx, %%mm0; \
72 movd (%%edi), %%mm3; \
73 paddq %%mm3, %%mm1; \
74 movd (%%esi), %%mm2; \
75 pmuludq %%mm0, %%mm2; \
76 movd 4(%%esi), %%mm4; \
77 pmuludq %%mm0, %%mm4; \
78 movd 8(%%esi), %%mm6; \
79 pmuludq %%mm0, %%mm6; \
80 movd 12(%%esi), %%mm7; \
81 pmuludq %%mm0, %%mm7; \
82 paddq %%mm2, %%mm1; \
83 movd 4(%%edi), %%mm3; \
84 paddq %%mm4, %%mm3; \
85 movd 8(%%edi), %%mm5; \
86 paddq %%mm6, %%mm5; \
87 movd 12(%%edi), %%mm4; \
88 paddq %%mm4, %%mm7; \
89 movd %%mm1, (%%edi); \
90 movd 16(%%esi), %%mm2; \
91 pmuludq %%mm0, %%mm2; \
92 psrlq $32, %%mm1; \
93 movd 20(%%esi), %%mm4; \
94 pmuludq %%mm0, %%mm4; \
95 paddq %%mm3, %%mm1; \
96 movd 24(%%esi), %%mm6; \
97 pmuludq %%mm0, %%mm6; \
98 movd %%mm1, 4(%%edi); \
99 psrlq $32, %%mm1; \
100 movd 28(%%esi), %%mm3; \
101 pmuludq %%mm0, %%mm3; \
102 paddq %%mm5, %%mm1; \
103 movd 16(%%edi), %%mm5; \
104 paddq %%mm5, %%mm2; \
105 movd %%mm1, 8(%%edi); \
106 psrlq $32, %%mm1; \
107 paddq %%mm7, %%mm1; \
108 movd 20(%%edi), %%mm5; \
109 paddq %%mm5, %%mm4; \
110 movd %%mm1, 12(%%edi); \
111 psrlq $32, %%mm1; \
112 paddq %%mm2, %%mm1; \
113 movd 24(%%edi), %%mm5; \
114 paddq %%mm5, %%mm6; \
115 movd %%mm1, 16(%%edi); \
116 psrlq $32, %%mm1; \
117 paddq %%mm4, %%mm1; \
118 movd 28(%%edi), %%mm5; \
119 paddq %%mm5, %%mm3; \
120 movd %%mm1, 20(%%edi); \
121 psrlq $32, %%mm1; \
122 paddq %%mm6, %%mm1; \
123 movd %%mm1, 24(%%edi); \
124 psrlq $32, %%mm1; \
125 paddq %%mm3, %%mm1; \
126 movd %%mm1, 28(%%edi); \
127 addl $32, %%edi; \
128 addl $32, %%esi; \
129 psrlq $32, %%mm1; \
130 movd %%mm1, %%ecx; \
131 "
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]132
Paul Bakkerc89cf7c2009-07-19 21:37:39 +0000[diff] [blame]133#define MULADDC_STOP \
134 " \
135 emms; \
136 movl %4, %%ebx; \
137 movl %%ecx, %1; \
138 movl %%edi, %2; \
139 movl %%esi, %3; \
140 " \
141 : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \
142 : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \
143 : "eax", "ecx", "edx", "esi", "edi" \
144 );
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]145
146#else
147
Paul Bakkerc89cf7c2009-07-19 21:37:39 +0000[diff] [blame]148#define MULADDC_STOP \
149 " \
150 movl %4, %%ebx; \
151 movl %%ecx, %1; \
152 movl %%edi, %2; \
153 movl %%esi, %3; \
154 " \
155 : "=m" (t), "=m" (c), "=m" (d), "=m" (s) \
156 : "m" (t), "m" (s), "m" (d), "m" (c), "m" (b) \
157 : "eax", "ecx", "edx", "esi", "edi" \
158 );
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]159#endif /* SSE2 */
160#endif /* i386 */
161
162#if defined(__amd64__) || defined (__x86_64__)
163
164#define MULADDC_INIT \
165 asm( "movq %0, %%rsi " :: "m" (s)); \
166 asm( "movq %0, %%rdi " :: "m" (d)); \
167 asm( "movq %0, %%rcx " :: "m" (c)); \
168 asm( "movq %0, %%rbx " :: "m" (b)); \
169 asm( "xorq %r8, %r8 " );
170
171#define MULADDC_CORE \
172 asm( "movq (%rsi),%rax " ); \
173 asm( "mulq %rbx " ); \
174 asm( "addq $8, %rsi " ); \
175 asm( "addq %rcx, %rax " ); \
176 asm( "movq %r8, %rcx " ); \
177 asm( "adcq $0, %rdx " ); \
178 asm( "nop " ); \
179 asm( "addq %rax, (%rdi) " ); \
180 asm( "adcq %rdx, %rcx " ); \
181 asm( "addq $8, %rdi " );
182
183#define MULADDC_STOP \
184 asm( "movq %%rcx, %0 " : "=m" (c)); \
185 asm( "movq %%rdi, %0 " : "=m" (d)); \
186 asm( "movq %%rsi, %0 " : "=m" (s) :: \
187 "rax", "rcx", "rdx", "rbx", "rsi", "rdi", "r8" );
188
189#endif /* AMD64 */
190
191#if defined(__mc68020__) || defined(__mcpu32__)
192
193#define MULADDC_INIT \
194 asm( "movl %0, %%a2 " :: "m" (s)); \
195 asm( "movl %0, %%a3 " :: "m" (d)); \
196 asm( "movl %0, %%d3 " :: "m" (c)); \
197 asm( "movl %0, %%d2 " :: "m" (b)); \
198 asm( "moveq #0, %d0 " );
199
200#define MULADDC_CORE \
201 asm( "movel %a2@+, %d1 " ); \
202 asm( "mulul %d2, %d4:%d1 " ); \
203 asm( "addl %d3, %d1 " ); \
204 asm( "addxl %d0, %d4 " ); \
205 asm( "moveq #0, %d3 " ); \
206 asm( "addl %d1, %a3@+ " ); \
207 asm( "addxl %d4, %d3 " );
208
209#define MULADDC_STOP \
210 asm( "movl %%d3, %0 " : "=m" (c)); \
211 asm( "movl %%a3, %0 " : "=m" (d)); \
212 asm( "movl %%a2, %0 " : "=m" (s) :: \
213 "d0", "d1", "d2", "d3", "d4", "a2", "a3" );
214
215#define MULADDC_HUIT \
216 asm( "movel %a2@+, %d1 " ); \
217 asm( "mulul %d2, %d4:%d1 " ); \
218 asm( "addxl %d3, %d1 " ); \
219 asm( "addxl %d0, %d4 " ); \
220 asm( "addl %d1, %a3@+ " ); \
221 asm( "movel %a2@+, %d1 " ); \
222 asm( "mulul %d2, %d3:%d1 " ); \
223 asm( "addxl %d4, %d1 " ); \
224 asm( "addxl %d0, %d3 " ); \
225 asm( "addl %d1, %a3@+ " ); \
226 asm( "movel %a2@+, %d1 " ); \
227 asm( "mulul %d2, %d4:%d1 " ); \
228 asm( "addxl %d3, %d1 " ); \
229 asm( "addxl %d0, %d4 " ); \
230 asm( "addl %d1, %a3@+ " ); \
231 asm( "movel %a2@+, %d1 " ); \
232 asm( "mulul %d2, %d3:%d1 " ); \
233 asm( "addxl %d4, %d1 " ); \
234 asm( "addxl %d0, %d3 " ); \
235 asm( "addl %d1, %a3@+ " ); \
236 asm( "movel %a2@+, %d1 " ); \
237 asm( "mulul %d2, %d4:%d1 " ); \
238 asm( "addxl %d3, %d1 " ); \
239 asm( "addxl %d0, %d4 " ); \
240 asm( "addl %d1, %a3@+ " ); \
241 asm( "movel %a2@+, %d1 " ); \
242 asm( "mulul %d2, %d3:%d1 " ); \
243 asm( "addxl %d4, %d1 " ); \
244 asm( "addxl %d0, %d3 " ); \
245 asm( "addl %d1, %a3@+ " ); \
246 asm( "movel %a2@+, %d1 " ); \
247 asm( "mulul %d2, %d4:%d1 " ); \
248 asm( "addxl %d3, %d1 " ); \
249 asm( "addxl %d0, %d4 " ); \
250 asm( "addl %d1, %a3@+ " ); \
251 asm( "movel %a2@+, %d1 " ); \
252 asm( "mulul %d2, %d3:%d1 " ); \
253 asm( "addxl %d4, %d1 " ); \
254 asm( "addxl %d0, %d3 " ); \
255 asm( "addl %d1, %a3@+ " ); \
256 asm( "addxl %d0, %d3 " );
257
258#endif /* MC68000 */
259
260#if defined(__powerpc__) || defined(__ppc__)
261#if defined(__powerpc64__) || defined(__ppc64__)
262
263#if defined(__MACH__) && defined(__APPLE__)
264
265#define MULADDC_INIT \
266 asm( "ld r3, %0 " :: "m" (s)); \
267 asm( "ld r4, %0 " :: "m" (d)); \
268 asm( "ld r5, %0 " :: "m" (c)); \
269 asm( "ld r6, %0 " :: "m" (b)); \
270 asm( "addi r3, r3, -8 " ); \
271 asm( "addi r4, r4, -8 " ); \
272 asm( "addic r5, r5, 0 " );
273
274#define MULADDC_CORE \
275 asm( "ldu r7, 8(r3) " ); \
276 asm( "mulld r8, r7, r6 " ); \
277 asm( "mulhdu r9, r7, r6 " ); \
278 asm( "adde r8, r8, r5 " ); \
279 asm( "ld r7, 8(r4) " ); \
280 asm( "addze r5, r9 " ); \
281 asm( "addc r8, r8, r7 " ); \
282 asm( "stdu r8, 8(r4) " );
283
284#define MULADDC_STOP \
285 asm( "addze r5, r5 " ); \
286 asm( "addi r4, r4, 8 " ); \
287 asm( "addi r3, r3, 8 " ); \
288 asm( "std r5, %0 " : "=m" (c)); \
289 asm( "std r4, %0 " : "=m" (d)); \
290 asm( "std r3, %0 " : "=m" (s) :: \
291 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
292
293#else
294
295#define MULADDC_INIT \
296 asm( "ld %%r3, %0 " :: "m" (s)); \
297 asm( "ld %%r4, %0 " :: "m" (d)); \
298 asm( "ld %%r5, %0 " :: "m" (c)); \
299 asm( "ld %%r6, %0 " :: "m" (b)); \
300 asm( "addi %r3, %r3, -8 " ); \
301 asm( "addi %r4, %r4, -8 " ); \
302 asm( "addic %r5, %r5, 0 " );
303
304#define MULADDC_CORE \
305 asm( "ldu %r7, 8(%r3) " ); \
306 asm( "mulld %r8, %r7, %r6 " ); \
307 asm( "mulhdu %r9, %r7, %r6 " ); \
308 asm( "adde %r8, %r8, %r5 " ); \
309 asm( "ld %r7, 8(%r4) " ); \
310 asm( "addze %r5, %r9 " ); \
311 asm( "addc %r8, %r8, %r7 " ); \
312 asm( "stdu %r8, 8(%r4) " );
313
314#define MULADDC_STOP \
315 asm( "addze %r5, %r5 " ); \
316 asm( "addi %r4, %r4, 8 " ); \
317 asm( "addi %r3, %r3, 8 " ); \
318 asm( "std %%r5, %0 " : "=m" (c)); \
319 asm( "std %%r4, %0 " : "=m" (d)); \
320 asm( "std %%r3, %0 " : "=m" (s) :: \
321 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
322
323#endif
324
325#else /* PPC32 */
326
327#if defined(__MACH__) && defined(__APPLE__)
328
329#define MULADDC_INIT \
330 asm( "lwz r3, %0 " :: "m" (s)); \
331 asm( "lwz r4, %0 " :: "m" (d)); \
332 asm( "lwz r5, %0 " :: "m" (c)); \
333 asm( "lwz r6, %0 " :: "m" (b)); \
334 asm( "addi r3, r3, -4 " ); \
335 asm( "addi r4, r4, -4 " ); \
336 asm( "addic r5, r5, 0 " );
337
338#define MULADDC_CORE \
339 asm( "lwzu r7, 4(r3) " ); \
340 asm( "mullw r8, r7, r6 " ); \
341 asm( "mulhwu r9, r7, r6 " ); \
342 asm( "adde r8, r8, r5 " ); \
343 asm( "lwz r7, 4(r4) " ); \
344 asm( "addze r5, r9 " ); \
345 asm( "addc r8, r8, r7 " ); \
346 asm( "stwu r8, 4(r4) " );
347
348#define MULADDC_STOP \
349 asm( "addze r5, r5 " ); \
350 asm( "addi r4, r4, 4 " ); \
351 asm( "addi r3, r3, 4 " ); \
352 asm( "stw r5, %0 " : "=m" (c)); \
353 asm( "stw r4, %0 " : "=m" (d)); \
354 asm( "stw r3, %0 " : "=m" (s) :: \
355 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
356
357#else
358
359#define MULADDC_INIT \
360 asm( "lwz %%r3, %0 " :: "m" (s)); \
361 asm( "lwz %%r4, %0 " :: "m" (d)); \
362 asm( "lwz %%r5, %0 " :: "m" (c)); \
363 asm( "lwz %%r6, %0 " :: "m" (b)); \
364 asm( "addi %r3, %r3, -4 " ); \
365 asm( "addi %r4, %r4, -4 " ); \
366 asm( "addic %r5, %r5, 0 " );
367
368#define MULADDC_CORE \
369 asm( "lwzu %r7, 4(%r3) " ); \
370 asm( "mullw %r8, %r7, %r6 " ); \
371 asm( "mulhwu %r9, %r7, %r6 " ); \
372 asm( "adde %r8, %r8, %r5 " ); \
373 asm( "lwz %r7, 4(%r4) " ); \
374 asm( "addze %r5, %r9 " ); \
375 asm( "addc %r8, %r8, %r7 " ); \
376 asm( "stwu %r8, 4(%r4) " );
377
378#define MULADDC_STOP \
379 asm( "addze %r5, %r5 " ); \
380 asm( "addi %r4, %r4, 4 " ); \
381 asm( "addi %r3, %r3, 4 " ); \
382 asm( "stw %%r5, %0 " : "=m" (c)); \
383 asm( "stw %%r4, %0 " : "=m" (d)); \
384 asm( "stw %%r3, %0 " : "=m" (s) :: \
385 "r3", "r4", "r5", "r6", "r7", "r8", "r9" );
386
387#endif
388
389#endif /* PPC32 */
390#endif /* PPC64 */
391
392#if defined(__sparc__)
393
394#define MULADDC_INIT \
395 asm( "ld %0, %%o0 " :: "m" (s)); \
396 asm( "ld %0, %%o1 " :: "m" (d)); \
397 asm( "ld %0, %%o2 " :: "m" (c)); \
398 asm( "ld %0, %%o3 " :: "m" (b));
399
400#define MULADDC_CORE \
401 asm( "ld [%o0], %o4 " ); \
402 asm( "inc 4, %o0 " ); \
403 asm( "ld [%o1], %o5 " ); \
404 asm( "umul %o3, %o4, %o4 " ); \
405 asm( "addcc %o4, %o2, %o4 " ); \
406 asm( "rd %y, %g1 " ); \
407 asm( "addx %g1, 0, %g1 " ); \
408 asm( "addcc %o4, %o5, %o4 " ); \
409 asm( "st %o4, [%o1] " ); \
410 asm( "addx %g1, 0, %o2 " ); \
411 asm( "inc 4, %o1 " );
412
413#define MULADDC_STOP \
414 asm( "st %%o2, %0 " : "=m" (c)); \
415 asm( "st %%o1, %0 " : "=m" (d)); \
416 asm( "st %%o0, %0 " : "=m" (s) :: \
417 "g1", "o0", "o1", "o2", "o3", "o4", "o5" );
418
419#endif /* SPARCv8 */
420
421#if defined(__microblaze__) || defined(microblaze)
422
423#define MULADDC_INIT \
424 asm( "lwi r3, %0 " :: "m" (s)); \
425 asm( "lwi r4, %0 " :: "m" (d)); \
426 asm( "lwi r5, %0 " :: "m" (c)); \
427 asm( "lwi r6, %0 " :: "m" (b)); \
428 asm( "andi r7, r6, 0xffff" ); \
429 asm( "bsrli r6, r6, 16 " );
430
431#define MULADDC_CORE \
432 asm( "lhui r8, r3, 0 " ); \
433 asm( "addi r3, r3, 2 " ); \
434 asm( "lhui r9, r3, 0 " ); \
435 asm( "addi r3, r3, 2 " ); \
436 asm( "mul r10, r9, r6 " ); \
437 asm( "mul r11, r8, r7 " ); \
438 asm( "mul r12, r9, r7 " ); \
439 asm( "mul r13, r8, r6 " ); \
440 asm( "bsrli r8, r10, 16 " ); \
441 asm( "bsrli r9, r11, 16 " ); \
442 asm( "add r13, r13, r8 " ); \
443 asm( "add r13, r13, r9 " ); \
444 asm( "bslli r10, r10, 16 " ); \
445 asm( "bslli r11, r11, 16 " ); \
446 asm( "add r12, r12, r10 " ); \
447 asm( "addc r13, r13, r0 " ); \
448 asm( "add r12, r12, r11 " ); \
449 asm( "addc r13, r13, r0 " ); \
450 asm( "lwi r10, r4, 0 " ); \
451 asm( "add r12, r12, r10 " ); \
452 asm( "addc r13, r13, r0 " ); \
453 asm( "add r12, r12, r5 " ); \
454 asm( "addc r5, r13, r0 " ); \
455 asm( "swi r12, r4, 0 " ); \
456 asm( "addi r4, r4, 4 " );
457
458#define MULADDC_STOP \
459 asm( "swi r5, %0 " : "=m" (c)); \
460 asm( "swi r4, %0 " : "=m" (d)); \
461 asm( "swi r3, %0 " : "=m" (s) :: \
462 "r3", "r4" , "r5" , "r6" , "r7" , "r8" , \
463 "r9", "r10", "r11", "r12", "r13" );
464
465#endif /* MicroBlaze */
466
467#if defined(__tricore__)
468
469#define MULADDC_INIT \
470 asm( "ld.a %%a2, %0 " :: "m" (s)); \
471 asm( "ld.a %%a3, %0 " :: "m" (d)); \
472 asm( "ld.w %%d4, %0 " :: "m" (c)); \
473 asm( "ld.w %%d1, %0 " :: "m" (b)); \
474 asm( "xor %d5, %d5 " );
475
476#define MULADDC_CORE \
477 asm( "ld.w %d0, [%a2+] " ); \
478 asm( "madd.u %e2, %e4, %d0, %d1 " ); \
479 asm( "ld.w %d0, [%a3] " ); \
480 asm( "addx %d2, %d2, %d0 " ); \
481 asm( "addc %d3, %d3, 0 " ); \
482 asm( "mov %d4, %d3 " ); \
483 asm( "st.w [%a3+], %d2 " );
484
485#define MULADDC_STOP \
486 asm( "st.w %0, %%d4 " : "=m" (c)); \
487 asm( "st.a %0, %%a3 " : "=m" (d)); \
488 asm( "st.a %0, %%a2 " : "=m" (s) :: \
489 "d0", "d1", "e2", "d4", "a2", "a3" );
490
491#endif /* TriCore */
492
493#if defined(__arm__)
494
495#define MULADDC_INIT \
496 asm( "ldr r0, %0 " :: "m" (s)); \
497 asm( "ldr r1, %0 " :: "m" (d)); \
498 asm( "ldr r2, %0 " :: "m" (c)); \
499 asm( "ldr r3, %0 " :: "m" (b));
500
501#define MULADDC_CORE \
502 asm( "ldr r4, [r0], #4 " ); \
503 asm( "mov r5, #0 " ); \
504 asm( "ldr r6, [r1] " ); \
505 asm( "umlal r2, r5, r3, r4 " ); \
506 asm( "adds r7, r6, r2 " ); \
507 asm( "adc r2, r5, #0 " ); \
508 asm( "str r7, [r1], #4 " );
509
510#define MULADDC_STOP \
511 asm( "str r2, %0 " : "=m" (c)); \
512 asm( "str r1, %0 " : "=m" (d)); \
513 asm( "str r0, %0 " : "=m" (s) :: \
514 "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7" );
515
516#endif /* ARMv3 */
517
518#if defined(__alpha__)
519
520#define MULADDC_INIT \
521 asm( "ldq $1, %0 " :: "m" (s)); \
522 asm( "ldq $2, %0 " :: "m" (d)); \
523 asm( "ldq $3, %0 " :: "m" (c)); \
524 asm( "ldq $4, %0 " :: "m" (b));
525
526#define MULADDC_CORE \
527 asm( "ldq $6, 0($1) " ); \
528 asm( "addq $1, 8, $1 " ); \
529 asm( "mulq $6, $4, $7 " ); \
530 asm( "umulh $6, $4, $6 " ); \
531 asm( "addq $7, $3, $7 " ); \
532 asm( "cmpult $7, $3, $3 " ); \
533 asm( "ldq $5, 0($2) " ); \
534 asm( "addq $7, $5, $7 " ); \
535 asm( "cmpult $7, $5, $5 " ); \
536 asm( "stq $7, 0($2) " ); \
537 asm( "addq $2, 8, $2 " ); \
538 asm( "addq $6, $3, $3 " ); \
539 asm( "addq $5, $3, $3 " );
540
541#define MULADDC_STOP \
542 asm( "stq $3, %0 " : "=m" (c)); \
543 asm( "stq $2, %0 " : "=m" (d)); \
544 asm( "stq $1, %0 " : "=m" (s) :: \
545 "$1", "$2", "$3", "$4", "$5", "$6", "$7" );
546
547#endif /* Alpha */
548
549#if defined(__mips__)
550
551#define MULADDC_INIT \
552 asm( "lw $10, %0 " :: "m" (s)); \
553 asm( "lw $11, %0 " :: "m" (d)); \
554 asm( "lw $12, %0 " :: "m" (c)); \
555 asm( "lw $13, %0 " :: "m" (b));
556
557#define MULADDC_CORE \
558 asm( "lw $14, 0($10) " ); \
559 asm( "multu $13, $14 " ); \
560 asm( "addi $10, $10, 4 " ); \
561 asm( "mflo $14 " ); \
562 asm( "mfhi $9 " ); \
563 asm( "addu $14, $12, $14 " ); \
564 asm( "lw $15, 0($11) " ); \
565 asm( "sltu $12, $14, $12 " ); \
566 asm( "addu $15, $14, $15 " ); \
567 asm( "sltu $14, $15, $14 " ); \
568 asm( "addu $12, $12, $9 " ); \
569 asm( "sw $15, 0($11) " ); \
570 asm( "addu $12, $12, $14 " ); \
571 asm( "addi $11, $11, 4 " );
572
573#define MULADDC_STOP \
574 asm( "sw $12, %0 " : "=m" (c)); \
575 asm( "sw $11, %0 " : "=m" (d)); \
576 asm( "sw $10, %0 " : "=m" (s) :: \
577 "$9", "$10", "$11", "$12", "$13", "$14", "$15" );
578
579#endif /* MIPS */
580#endif /* GNUC */
581
582#if (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__)
583
584#define MULADDC_INIT \
585 __asm mov esi, s \
586 __asm mov edi, d \
587 __asm mov ecx, c \
588 __asm mov ebx, b
589
590#define MULADDC_CORE \
591 __asm lodsd \
592 __asm mul ebx \
593 __asm add eax, ecx \
594 __asm adc edx, 0 \
595 __asm add eax, [edi] \
596 __asm adc edx, 0 \
597 __asm mov ecx, edx \
598 __asm stosd
599
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]600#if defined(POLARSSL_HAVE_SSE2)
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]601
602#define EMIT __asm _emit
603
604#define MULADDC_HUIT \
605 EMIT 0x0F EMIT 0x6E EMIT 0xC9 \
606 EMIT 0x0F EMIT 0x6E EMIT 0xC3 \
607 EMIT 0x0F EMIT 0x6E EMIT 0x1F \
608 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
609 EMIT 0x0F EMIT 0x6E EMIT 0x16 \
610 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
611 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x04 \
612 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
613 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x08 \
614 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
615 EMIT 0x0F EMIT 0x6E EMIT 0x7E EMIT 0x0C \
616 EMIT 0x0F EMIT 0xF4 EMIT 0xF8 \
617 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
618 EMIT 0x0F EMIT 0x6E EMIT 0x5F EMIT 0x04 \
619 EMIT 0x0F EMIT 0xD4 EMIT 0xDC \
620 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x08 \
621 EMIT 0x0F EMIT 0xD4 EMIT 0xEE \
622 EMIT 0x0F EMIT 0x6E EMIT 0x67 EMIT 0x0C \
623 EMIT 0x0F EMIT 0xD4 EMIT 0xFC \
624 EMIT 0x0F EMIT 0x7E EMIT 0x0F \
625 EMIT 0x0F EMIT 0x6E EMIT 0x56 EMIT 0x10 \
626 EMIT 0x0F EMIT 0xF4 EMIT 0xD0 \
627 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
628 EMIT 0x0F EMIT 0x6E EMIT 0x66 EMIT 0x14 \
629 EMIT 0x0F EMIT 0xF4 EMIT 0xE0 \
630 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
631 EMIT 0x0F EMIT 0x6E EMIT 0x76 EMIT 0x18 \
632 EMIT 0x0F EMIT 0xF4 EMIT 0xF0 \
633 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x04 \
634 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
635 EMIT 0x0F EMIT 0x6E EMIT 0x5E EMIT 0x1C \
636 EMIT 0x0F EMIT 0xF4 EMIT 0xD8 \
637 EMIT 0x0F EMIT 0xD4 EMIT 0xCD \
638 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x10 \
639 EMIT 0x0F EMIT 0xD4 EMIT 0xD5 \
640 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x08 \
641 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
642 EMIT 0x0F EMIT 0xD4 EMIT 0xCF \
643 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x14 \
644 EMIT 0x0F EMIT 0xD4 EMIT 0xE5 \
645 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x0C \
646 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
647 EMIT 0x0F EMIT 0xD4 EMIT 0xCA \
648 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x18 \
649 EMIT 0x0F EMIT 0xD4 EMIT 0xF5 \
650 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x10 \
651 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
652 EMIT 0x0F EMIT 0xD4 EMIT 0xCC \
653 EMIT 0x0F EMIT 0x6E EMIT 0x6F EMIT 0x1C \
654 EMIT 0x0F EMIT 0xD4 EMIT 0xDD \
655 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x14 \
656 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
657 EMIT 0x0F EMIT 0xD4 EMIT 0xCE \
658 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x18 \
659 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
660 EMIT 0x0F EMIT 0xD4 EMIT 0xCB \
661 EMIT 0x0F EMIT 0x7E EMIT 0x4F EMIT 0x1C \
662 EMIT 0x83 EMIT 0xC7 EMIT 0x20 \
663 EMIT 0x83 EMIT 0xC6 EMIT 0x20 \
664 EMIT 0x0F EMIT 0x73 EMIT 0xD1 EMIT 0x20 \
665 EMIT 0x0F EMIT 0x7E EMIT 0xC9
666
667#define MULADDC_STOP \
668 EMIT 0x0F EMIT 0x77 \
669 __asm mov c, ecx \
670 __asm mov d, edi \
671 __asm mov s, esi \
672
673#else
674
675#define MULADDC_STOP \
676 __asm mov c, ecx \
677 __asm mov d, edi \
678 __asm mov s, esi \
679
680#endif /* SSE2 */
681#endif /* MSVC */
682
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]683#endif /* POLARSSL_HAVE_ASM */
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]684
685#if !defined(MULADDC_CORE)
Paul Bakker40e46942009-01-03 21:51:57 +0000[diff] [blame]686#if defined(POLARSSL_HAVE_LONGLONG)
Paul Bakker5121ce52009-01-03 21:22:43 +0000[diff] [blame]687
688#define MULADDC_INIT \
689{ \
690 t_dbl r; \
691 t_int r0, r1;
692
693#define MULADDC_CORE \
694 r = *(s++) * (t_dbl) b; \
695 r0 = r; \
696 r1 = r >> biL; \
697 r0 += c; r1 += (r0 < c); \
698 r0 += *d; r1 += (r0 < *d); \
699 c = r1; *(d++) = r0;
700
701#define MULADDC_STOP \
702}
703
704#else
705#define MULADDC_INIT \
706{ \
707 t_int s0, s1, b0, b1; \
708 t_int r0, r1, rx, ry; \
709 b0 = ( b << biH ) >> biH; \
710 b1 = ( b >> biH );
711
712#define MULADDC_CORE \
713 s0 = ( *s << biH ) >> biH; \
714 s1 = ( *s >> biH ); s++; \
715 rx = s0 * b1; r0 = s0 * b0; \
716 ry = s1 * b0; r1 = s1 * b1; \
717 r1 += ( rx >> biH ); \
718 r1 += ( ry >> biH ); \
719 rx <<= biH; ry <<= biH; \
720 r0 += rx; r1 += (r0 < rx); \
721 r0 += ry; r1 += (r0 < ry); \
722 r0 += c; r1 += (r0 < c); \
723 r0 += *d; r1 += (r0 < *d); \
724 c = r1; *(d++) = r0;
725
726#define MULADDC_STOP \
727}
728
729#endif /* C (generic) */
730#endif /* C (longlong) */
731
732#endif /* bn_mul.h */