]>
Commit | Line | Data |
---|---|---|
4f22f405 RS |
1 | /* |
2 | * Copyright 2002-2016 The OpenSSL Project Authors. All Rights Reserved. | |
3 | * | |
4 | * Licensed under the OpenSSL license (the "License"). You may not use | |
5 | * this file except in compliance with the License. You can obtain a copy | |
6 | * in the file LICENSE in the source distribution or at | |
7 | * https://www.openssl.org/source/license.html | |
8 | */ | |
9 | ||
41b76195 | 10 | #include "../bn_lcl.h" |
93c4ba07 | 11 | #if !(defined(__GNUC__) && __GNUC__>=2) |
0f113f3e | 12 | # include "../bn_asm.c" /* kind of dirty hack for Sun Studio */ |
c8a0d0aa | 13 | #else |
1d97c843 | 14 | /*- |
2f98abbc AP |
15 | * x86_64 BIGNUM accelerator version 0.1, December 2002. |
16 | * | |
17 | * Implemented by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL | |
18 | * project. | |
19 | * | |
20 | * Rights for redistribution and usage in source and binary forms are | |
21 | * granted according to the OpenSSL license. Warranty of any kind is | |
22 | * disclaimed. | |
23 | * | |
24 | * Q. Version 0.1? It doesn't sound like Andy, he used to assign real | |
25 | * versions, like 1.0... | |
26 | * A. Well, that's because this code is basically a quick-n-dirty | |
27 | * proof-of-concept hack. As you can see it's implemented with | |
28 | * inline assembler, which means that you're bound to GCC and that | |
1809e858 | 29 | * there might be enough room for further improvement. |
2f98abbc AP |
30 | * |
31 | * Q. Why inline assembler? | |
1809e858 AP |
32 | * A. x86_64 features own ABI which I'm not familiar with. This is |
33 | * why I decided to let the compiler take care of subroutine | |
34 | * prologue/epilogue as well as register allocation. For reference. | |
35 | * Win64 implements different ABI for AMD64, different from Linux. | |
2f98abbc AP |
36 | * |
37 | * Q. How much faster does it get? | |
1809e858 AP |
38 | * A. 'apps/openssl speed rsa dsa' output with no-asm: |
39 | * | |
0f113f3e MC |
40 | * sign verify sign/s verify/s |
41 | * rsa 512 bits 0.0006s 0.0001s 1683.8 18456.2 | |
42 | * rsa 1024 bits 0.0028s 0.0002s 356.0 6407.0 | |
43 | * rsa 2048 bits 0.0172s 0.0005s 58.0 1957.8 | |
44 | * rsa 4096 bits 0.1155s 0.0018s 8.7 555.6 | |
45 | * sign verify sign/s verify/s | |
46 | * dsa 512 bits 0.0005s 0.0006s 2100.8 1768.3 | |
47 | * dsa 1024 bits 0.0014s 0.0018s 692.3 559.2 | |
48 | * dsa 2048 bits 0.0049s 0.0061s 204.7 165.0 | |
1809e858 AP |
49 | * |
50 | * 'apps/openssl speed rsa dsa' output with this module: | |
51 | * | |
0f113f3e MC |
52 | * sign verify sign/s verify/s |
53 | * rsa 512 bits 0.0004s 0.0000s 2767.1 33297.9 | |
54 | * rsa 1024 bits 0.0012s 0.0001s 867.4 14674.7 | |
55 | * rsa 2048 bits 0.0061s 0.0002s 164.0 5270.0 | |
56 | * rsa 4096 bits 0.0384s 0.0006s 26.1 1650.8 | |
57 | * sign verify sign/s verify/s | |
58 | * dsa 512 bits 0.0002s 0.0003s 4442.2 3786.3 | |
59 | * dsa 1024 bits 0.0005s 0.0007s 1835.1 1497.4 | |
60 | * dsa 2048 bits 0.0016s 0.0020s 620.4 504.6 | |
1809e858 AP |
61 | * |
62 | * For the reference. IA-32 assembler implementation performs | |
63 | * very much like 64-bit code compiled with no-asm on the same | |
64 | * machine. | |
2f98abbc AP |
65 | */ |
66 | ||
0f113f3e MC |
67 | # if defined(_WIN64) || !defined(__LP64__) |
68 | # define BN_ULONG unsigned long long | |
69 | # else | |
70 | # define BN_ULONG unsigned long | |
71 | # endif | |
2f98abbc | 72 | |
0f113f3e MC |
73 | # undef mul |
74 | # undef mul_add | |
41b76195 | 75 | |
1d97c843 | 76 | /*- |
053fa39a | 77 | * "m"(a), "+m"(r) is the way to favor DirectPath ยต-code; |
0f113f3e MC |
78 | * "g"(0) let the compiler to decide where does it |
79 | * want to keep the value of zero; | |
2f98abbc | 80 | */ |
0f113f3e MC |
81 | # define mul_add(r,a,word,carry) do { \ |
82 | register BN_ULONG high,low; \ | |
83 | asm ("mulq %3" \ | |
84 | : "=a"(low),"=d"(high) \ | |
85 | : "a"(word),"m"(a) \ | |
86 | : "cc"); \ | |
87 | asm ("addq %2,%0; adcq %3,%1" \ | |
88 | : "+r"(carry),"+d"(high)\ | |
89 | : "a"(low),"g"(0) \ | |
90 | : "cc"); \ | |
91 | asm ("addq %2,%0; adcq %3,%1" \ | |
92 | : "+m"(r),"+d"(high) \ | |
93 | : "r"(carry),"g"(0) \ | |
94 | : "cc"); \ | |
95 | carry=high; \ | |
96 | } while (0) | |
97 | ||
98 | # define mul(r,a,word,carry) do { \ | |
99 | register BN_ULONG high,low; \ | |
100 | asm ("mulq %3" \ | |
101 | : "=a"(low),"=d"(high) \ | |
102 | : "a"(word),"g"(a) \ | |
103 | : "cc"); \ | |
104 | asm ("addq %2,%0; adcq %3,%1" \ | |
105 | : "+r"(carry),"+d"(high)\ | |
106 | : "a"(low),"g"(0) \ | |
107 | : "cc"); \ | |
108 | (r)=carry, carry=high; \ | |
109 | } while (0) | |
c27310f9 | 110 | # undef sqr |
0f113f3e MC |
111 | # define sqr(r0,r1,a) \ |
112 | asm ("mulq %2" \ | |
113 | : "=a"(r0),"=d"(r1) \ | |
114 | : "a"(a) \ | |
115 | : "cc"); | |
116 | ||
117 | BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num, | |
118 | BN_ULONG w) | |
119 | { | |
120 | BN_ULONG c1 = 0; | |
121 | ||
122 | if (num <= 0) | |
123 | return (c1); | |
124 | ||
125 | while (num & ~3) { | |
126 | mul_add(rp[0], ap[0], w, c1); | |
127 | mul_add(rp[1], ap[1], w, c1); | |
128 | mul_add(rp[2], ap[2], w, c1); | |
129 | mul_add(rp[3], ap[3], w, c1); | |
130 | ap += 4; | |
131 | rp += 4; | |
132 | num -= 4; | |
133 | } | |
134 | if (num) { | |
135 | mul_add(rp[0], ap[0], w, c1); | |
136 | if (--num == 0) | |
137 | return c1; | |
138 | mul_add(rp[1], ap[1], w, c1); | |
139 | if (--num == 0) | |
140 | return c1; | |
141 | mul_add(rp[2], ap[2], w, c1); | |
142 | return c1; | |
143 | } | |
144 | ||
145 | return (c1); | |
146 | } | |
2f98abbc | 147 | |
41b76195 | 148 | BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w) |
0f113f3e MC |
149 | { |
150 | BN_ULONG c1 = 0; | |
151 | ||
152 | if (num <= 0) | |
153 | return (c1); | |
154 | ||
155 | while (num & ~3) { | |
156 | mul(rp[0], ap[0], w, c1); | |
157 | mul(rp[1], ap[1], w, c1); | |
158 | mul(rp[2], ap[2], w, c1); | |
159 | mul(rp[3], ap[3], w, c1); | |
160 | ap += 4; | |
161 | rp += 4; | |
162 | num -= 4; | |
163 | } | |
164 | if (num) { | |
165 | mul(rp[0], ap[0], w, c1); | |
166 | if (--num == 0) | |
167 | return c1; | |
168 | mul(rp[1], ap[1], w, c1); | |
169 | if (--num == 0) | |
170 | return c1; | |
171 | mul(rp[2], ap[2], w, c1); | |
172 | } | |
173 | return (c1); | |
174 | } | |
2f98abbc | 175 | |
41b76195 | 176 | void bn_sqr_words(BN_ULONG *r, const BN_ULONG *a, int n) |
0f113f3e MC |
177 | { |
178 | if (n <= 0) | |
179 | return; | |
180 | ||
181 | while (n & ~3) { | |
182 | sqr(r[0], r[1], a[0]); | |
183 | sqr(r[2], r[3], a[1]); | |
184 | sqr(r[4], r[5], a[2]); | |
185 | sqr(r[6], r[7], a[3]); | |
186 | a += 4; | |
187 | r += 8; | |
188 | n -= 4; | |
189 | } | |
190 | if (n) { | |
191 | sqr(r[0], r[1], a[0]); | |
192 | if (--n == 0) | |
193 | return; | |
194 | sqr(r[2], r[3], a[1]); | |
195 | if (--n == 0) | |
196 | return; | |
197 | sqr(r[4], r[5], a[2]); | |
198 | } | |
199 | } | |
2f98abbc AP |
200 | |
201 | BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d) | |
0f113f3e MC |
202 | { |
203 | BN_ULONG ret, waste; | |
2f98abbc | 204 | |
0f113f3e MC |
205 | asm("divq %4":"=a"(ret), "=d"(waste) |
206 | : "a"(l), "d"(h), "g"(d) | |
207 | : "cc"); | |
2f98abbc | 208 | |
0f113f3e | 209 | return ret; |
2f98abbc AP |
210 | } |
211 | ||
0f113f3e MC |
212 | BN_ULONG bn_add_words(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, |
213 | int n) | |
214 | { | |
215 | BN_ULONG ret; | |
216 | size_t i = 0; | |
217 | ||
218 | if (n <= 0) | |
219 | return 0; | |
220 | ||
221 | asm volatile (" subq %0,%0 \n" /* clear carry */ | |
222 | " jmp 1f \n" | |
223 | ".p2align 4 \n" | |
224 | "1: movq (%4,%2,8),%0 \n" | |
225 | " adcq (%5,%2,8),%0 \n" | |
226 | " movq %0,(%3,%2,8) \n" | |
227 | " lea 1(%2),%2 \n" | |
228 | " loop 1b \n" | |
229 | " sbbq %0,%0 \n":"=&r" (ret), "+c"(n), | |
230 | "+r"(i) | |
231 | :"r"(rp), "r"(ap), "r"(bp) | |
232 | :"cc", "memory"); | |
233 | ||
234 | return ret & 1; | |
2f98abbc AP |
235 | } |
236 | ||
0f113f3e MC |
237 | # ifndef SIMICS |
238 | BN_ULONG bn_sub_words(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, | |
239 | int n) | |
240 | { | |
241 | BN_ULONG ret; | |
242 | size_t i = 0; | |
243 | ||
244 | if (n <= 0) | |
245 | return 0; | |
246 | ||
247 | asm volatile (" subq %0,%0 \n" /* clear borrow */ | |
248 | " jmp 1f \n" | |
249 | ".p2align 4 \n" | |
250 | "1: movq (%4,%2,8),%0 \n" | |
251 | " sbbq (%5,%2,8),%0 \n" | |
252 | " movq %0,(%3,%2,8) \n" | |
253 | " lea 1(%2),%2 \n" | |
254 | " loop 1b \n" | |
255 | " sbbq %0,%0 \n":"=&r" (ret), "+c"(n), | |
256 | "+r"(i) | |
257 | :"r"(rp), "r"(ap), "r"(bp) | |
258 | :"cc", "memory"); | |
259 | ||
260 | return ret & 1; | |
2f98abbc | 261 | } |
0f113f3e | 262 | # else |
2f98abbc | 263 | /* Simics 1.4<7 has buggy sbbq:-( */ |
0f113f3e | 264 | # define BN_MASK2 0xffffffffffffffffL |
2f98abbc | 265 | BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n) |
0f113f3e MC |
266 | { |
267 | BN_ULONG t1, t2; | |
268 | int c = 0; | |
269 | ||
270 | if (n <= 0) | |
271 | return ((BN_ULONG)0); | |
272 | ||
273 | for (;;) { | |
274 | t1 = a[0]; | |
275 | t2 = b[0]; | |
276 | r[0] = (t1 - t2 - c) & BN_MASK2; | |
277 | if (t1 != t2) | |
278 | c = (t1 < t2); | |
279 | if (--n <= 0) | |
280 | break; | |
281 | ||
282 | t1 = a[1]; | |
283 | t2 = b[1]; | |
284 | r[1] = (t1 - t2 - c) & BN_MASK2; | |
285 | if (t1 != t2) | |
286 | c = (t1 < t2); | |
287 | if (--n <= 0) | |
288 | break; | |
289 | ||
290 | t1 = a[2]; | |
291 | t2 = b[2]; | |
292 | r[2] = (t1 - t2 - c) & BN_MASK2; | |
293 | if (t1 != t2) | |
294 | c = (t1 < t2); | |
295 | if (--n <= 0) | |
296 | break; | |
297 | ||
298 | t1 = a[3]; | |
299 | t2 = b[3]; | |
300 | r[3] = (t1 - t2 - c) & BN_MASK2; | |
301 | if (t1 != t2) | |
302 | c = (t1 < t2); | |
303 | if (--n <= 0) | |
304 | break; | |
305 | ||
306 | a += 4; | |
307 | b += 4; | |
308 | r += 4; | |
309 | } | |
310 | return (c); | |
311 | } | |
312 | # endif | |
2f98abbc AP |
313 | |
314 | /* mul_add_c(a,b,c0,c1,c2) -- c+=a*b for three word number c=(c2,c1,c0) */ | |
315 | /* mul_add_c2(a,b,c0,c1,c2) -- c+=2*a*b for three word number c=(c2,c1,c0) */ | |
316 | /* sqr_add_c(a,i,c0,c1,c2) -- c+=a[i]^2 for three word number c=(c2,c1,c0) */ | |
0f113f3e MC |
317 | /* |
318 | * sqr_add_c2(a,i,c0,c1,c2) -- c+=2*a[i]*a[j] for three word number | |
319 | * c=(c2,c1,c0) | |
320 | */ | |
2f98abbc | 321 | |
a7a44ba5 AP |
322 | /* |
323 | * Keep in mind that carrying into high part of multiplication result | |
324 | * can not overflow, because it cannot be all-ones. | |
325 | */ | |
0f113f3e | 326 | # if 0 |
2f98abbc | 327 | /* original macros are kept for reference purposes */ |
0f113f3e MC |
328 | # define mul_add_c(a,b,c0,c1,c2) do { \ |
329 | BN_ULONG ta = (a), tb = (b); \ | |
330 | BN_ULONG lo, hi; \ | |
331 | BN_UMULT_LOHI(lo,hi,ta,tb); \ | |
332 | c0 += lo; hi += (c0<lo)?1:0; \ | |
333 | c1 += hi; c2 += (c1<hi)?1:0; \ | |
334 | } while(0) | |
335 | ||
336 | # define mul_add_c2(a,b,c0,c1,c2) do { \ | |
337 | BN_ULONG ta = (a), tb = (b); \ | |
338 | BN_ULONG lo, hi, tt; \ | |
339 | BN_UMULT_LOHI(lo,hi,ta,tb); \ | |
340 | c0 += lo; tt = hi+((c0<lo)?1:0); \ | |
341 | c1 += tt; c2 += (c1<tt)?1:0; \ | |
342 | c0 += lo; hi += (c0<lo)?1:0; \ | |
343 | c1 += hi; c2 += (c1<hi)?1:0; \ | |
344 | } while(0) | |
345 | ||
346 | # define sqr_add_c(a,i,c0,c1,c2) do { \ | |
347 | BN_ULONG ta = (a)[i]; \ | |
348 | BN_ULONG lo, hi; \ | |
349 | BN_UMULT_LOHI(lo,hi,ta,ta); \ | |
350 | c0 += lo; hi += (c0<lo)?1:0; \ | |
351 | c1 += hi; c2 += (c1<hi)?1:0; \ | |
352 | } while(0) | |
353 | # else | |
354 | # define mul_add_c(a,b,c0,c1,c2) do { \ | |
355 | BN_ULONG t1,t2; \ | |
356 | asm ("mulq %3" \ | |
357 | : "=a"(t1),"=d"(t2) \ | |
358 | : "a"(a),"m"(b) \ | |
359 | : "cc"); \ | |
360 | asm ("addq %3,%0; adcq %4,%1; adcq %5,%2" \ | |
361 | : "+r"(c0),"+r"(c1),"+r"(c2) \ | |
362 | : "r"(t1),"r"(t2),"g"(0) \ | |
363 | : "cc"); \ | |
364 | } while (0) | |
365 | ||
366 | # define sqr_add_c(a,i,c0,c1,c2) do { \ | |
367 | BN_ULONG t1,t2; \ | |
368 | asm ("mulq %2" \ | |
369 | : "=a"(t1),"=d"(t2) \ | |
370 | : "a"(a[i]) \ | |
371 | : "cc"); \ | |
372 | asm ("addq %3,%0; adcq %4,%1; adcq %5,%2" \ | |
373 | : "+r"(c0),"+r"(c1),"+r"(c2) \ | |
374 | : "r"(t1),"r"(t2),"g"(0) \ | |
375 | : "cc"); \ | |
376 | } while (0) | |
377 | ||
378 | # define mul_add_c2(a,b,c0,c1,c2) do { \ | |
379 | BN_ULONG t1,t2; \ | |
380 | asm ("mulq %3" \ | |
381 | : "=a"(t1),"=d"(t2) \ | |
382 | : "a"(a),"m"(b) \ | |
383 | : "cc"); \ | |
384 | asm ("addq %3,%0; adcq %4,%1; adcq %5,%2" \ | |
385 | : "+r"(c0),"+r"(c1),"+r"(c2) \ | |
386 | : "r"(t1),"r"(t2),"g"(0) \ | |
387 | : "cc"); \ | |
388 | asm ("addq %3,%0; adcq %4,%1; adcq %5,%2" \ | |
389 | : "+r"(c0),"+r"(c1),"+r"(c2) \ | |
390 | : "r"(t1),"r"(t2),"g"(0) \ | |
391 | : "cc"); \ | |
392 | } while (0) | |
393 | # endif | |
394 | ||
395 | # define sqr_add_c2(a,i,j,c0,c1,c2) \ | |
396 | mul_add_c2((a)[i],(a)[j],c0,c1,c2) | |
2f98abbc AP |
397 | |
398 | void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b) | |
0f113f3e MC |
399 | { |
400 | BN_ULONG c1, c2, c3; | |
401 | ||
402 | c1 = 0; | |
403 | c2 = 0; | |
404 | c3 = 0; | |
405 | mul_add_c(a[0], b[0], c1, c2, c3); | |
406 | r[0] = c1; | |
407 | c1 = 0; | |
408 | mul_add_c(a[0], b[1], c2, c3, c1); | |
409 | mul_add_c(a[1], b[0], c2, c3, c1); | |
410 | r[1] = c2; | |
411 | c2 = 0; | |
412 | mul_add_c(a[2], b[0], c3, c1, c2); | |
413 | mul_add_c(a[1], b[1], c3, c1, c2); | |
414 | mul_add_c(a[0], b[2], c3, c1, c2); | |
415 | r[2] = c3; | |
416 | c3 = 0; | |
417 | mul_add_c(a[0], b[3], c1, c2, c3); | |
418 | mul_add_c(a[1], b[2], c1, c2, c3); | |
419 | mul_add_c(a[2], b[1], c1, c2, c3); | |
420 | mul_add_c(a[3], b[0], c1, c2, c3); | |
421 | r[3] = c1; | |
422 | c1 = 0; | |
423 | mul_add_c(a[4], b[0], c2, c3, c1); | |
424 | mul_add_c(a[3], b[1], c2, c3, c1); | |
425 | mul_add_c(a[2], b[2], c2, c3, c1); | |
426 | mul_add_c(a[1], b[3], c2, c3, c1); | |
427 | mul_add_c(a[0], b[4], c2, c3, c1); | |
428 | r[4] = c2; | |
429 | c2 = 0; | |
430 | mul_add_c(a[0], b[5], c3, c1, c2); | |
431 | mul_add_c(a[1], b[4], c3, c1, c2); | |
432 | mul_add_c(a[2], b[3], c3, c1, c2); | |
433 | mul_add_c(a[3], b[2], c3, c1, c2); | |
434 | mul_add_c(a[4], b[1], c3, c1, c2); | |
435 | mul_add_c(a[5], b[0], c3, c1, c2); | |
436 | r[5] = c3; | |
437 | c3 = 0; | |
438 | mul_add_c(a[6], b[0], c1, c2, c3); | |
439 | mul_add_c(a[5], b[1], c1, c2, c3); | |
440 | mul_add_c(a[4], b[2], c1, c2, c3); | |
441 | mul_add_c(a[3], b[3], c1, c2, c3); | |
442 | mul_add_c(a[2], b[4], c1, c2, c3); | |
443 | mul_add_c(a[1], b[5], c1, c2, c3); | |
444 | mul_add_c(a[0], b[6], c1, c2, c3); | |
445 | r[6] = c1; | |
446 | c1 = 0; | |
447 | mul_add_c(a[0], b[7], c2, c3, c1); | |
448 | mul_add_c(a[1], b[6], c2, c3, c1); | |
449 | mul_add_c(a[2], b[5], c2, c3, c1); | |
450 | mul_add_c(a[3], b[4], c2, c3, c1); | |
451 | mul_add_c(a[4], b[3], c2, c3, c1); | |
452 | mul_add_c(a[5], b[2], c2, c3, c1); | |
453 | mul_add_c(a[6], b[1], c2, c3, c1); | |
454 | mul_add_c(a[7], b[0], c2, c3, c1); | |
455 | r[7] = c2; | |
456 | c2 = 0; | |
457 | mul_add_c(a[7], b[1], c3, c1, c2); | |
458 | mul_add_c(a[6], b[2], c3, c1, c2); | |
459 | mul_add_c(a[5], b[3], c3, c1, c2); | |
460 | mul_add_c(a[4], b[4], c3, c1, c2); | |
461 | mul_add_c(a[3], b[5], c3, c1, c2); | |
462 | mul_add_c(a[2], b[6], c3, c1, c2); | |
463 | mul_add_c(a[1], b[7], c3, c1, c2); | |
464 | r[8] = c3; | |
465 | c3 = 0; | |
466 | mul_add_c(a[2], b[7], c1, c2, c3); | |
467 | mul_add_c(a[3], b[6], c1, c2, c3); | |
468 | mul_add_c(a[4], b[5], c1, c2, c3); | |
469 | mul_add_c(a[5], b[4], c1, c2, c3); | |
470 | mul_add_c(a[6], b[3], c1, c2, c3); | |
471 | mul_add_c(a[7], b[2], c1, c2, c3); | |
472 | r[9] = c1; | |
473 | c1 = 0; | |
474 | mul_add_c(a[7], b[3], c2, c3, c1); | |
475 | mul_add_c(a[6], b[4], c2, c3, c1); | |
476 | mul_add_c(a[5], b[5], c2, c3, c1); | |
477 | mul_add_c(a[4], b[6], c2, c3, c1); | |
478 | mul_add_c(a[3], b[7], c2, c3, c1); | |
479 | r[10] = c2; | |
480 | c2 = 0; | |
481 | mul_add_c(a[4], b[7], c3, c1, c2); | |
482 | mul_add_c(a[5], b[6], c3, c1, c2); | |
483 | mul_add_c(a[6], b[5], c3, c1, c2); | |
484 | mul_add_c(a[7], b[4], c3, c1, c2); | |
485 | r[11] = c3; | |
486 | c3 = 0; | |
487 | mul_add_c(a[7], b[5], c1, c2, c3); | |
488 | mul_add_c(a[6], b[6], c1, c2, c3); | |
489 | mul_add_c(a[5], b[7], c1, c2, c3); | |
490 | r[12] = c1; | |
491 | c1 = 0; | |
492 | mul_add_c(a[6], b[7], c2, c3, c1); | |
493 | mul_add_c(a[7], b[6], c2, c3, c1); | |
494 | r[13] = c2; | |
495 | c2 = 0; | |
496 | mul_add_c(a[7], b[7], c3, c1, c2); | |
497 | r[14] = c3; | |
498 | r[15] = c1; | |
499 | } | |
2f98abbc AP |
500 | |
501 | void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b) | |
0f113f3e MC |
502 | { |
503 | BN_ULONG c1, c2, c3; | |
504 | ||
505 | c1 = 0; | |
506 | c2 = 0; | |
507 | c3 = 0; | |
508 | mul_add_c(a[0], b[0], c1, c2, c3); | |
509 | r[0] = c1; | |
510 | c1 = 0; | |
511 | mul_add_c(a[0], b[1], c2, c3, c1); | |
512 | mul_add_c(a[1], b[0], c2, c3, c1); | |
513 | r[1] = c2; | |
514 | c2 = 0; | |
515 | mul_add_c(a[2], b[0], c3, c1, c2); | |
516 | mul_add_c(a[1], b[1], c3, c1, c2); | |
517 | mul_add_c(a[0], b[2], c3, c1, c2); | |
518 | r[2] = c3; | |
519 | c3 = 0; | |
520 | mul_add_c(a[0], b[3], c1, c2, c3); | |
521 | mul_add_c(a[1], b[2], c1, c2, c3); | |
522 | mul_add_c(a[2], b[1], c1, c2, c3); | |
523 | mul_add_c(a[3], b[0], c1, c2, c3); | |
524 | r[3] = c1; | |
525 | c1 = 0; | |
526 | mul_add_c(a[3], b[1], c2, c3, c1); | |
527 | mul_add_c(a[2], b[2], c2, c3, c1); | |
528 | mul_add_c(a[1], b[3], c2, c3, c1); | |
529 | r[4] = c2; | |
530 | c2 = 0; | |
531 | mul_add_c(a[2], b[3], c3, c1, c2); | |
532 | mul_add_c(a[3], b[2], c3, c1, c2); | |
533 | r[5] = c3; | |
534 | c3 = 0; | |
535 | mul_add_c(a[3], b[3], c1, c2, c3); | |
536 | r[6] = c1; | |
537 | r[7] = c2; | |
538 | } | |
2f98abbc | 539 | |
41b76195 | 540 | void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a) |
0f113f3e MC |
541 | { |
542 | BN_ULONG c1, c2, c3; | |
543 | ||
544 | c1 = 0; | |
545 | c2 = 0; | |
546 | c3 = 0; | |
547 | sqr_add_c(a, 0, c1, c2, c3); | |
548 | r[0] = c1; | |
549 | c1 = 0; | |
550 | sqr_add_c2(a, 1, 0, c2, c3, c1); | |
551 | r[1] = c2; | |
552 | c2 = 0; | |
553 | sqr_add_c(a, 1, c3, c1, c2); | |
554 | sqr_add_c2(a, 2, 0, c3, c1, c2); | |
555 | r[2] = c3; | |
556 | c3 = 0; | |
557 | sqr_add_c2(a, 3, 0, c1, c2, c3); | |
558 | sqr_add_c2(a, 2, 1, c1, c2, c3); | |
559 | r[3] = c1; | |
560 | c1 = 0; | |
561 | sqr_add_c(a, 2, c2, c3, c1); | |
562 | sqr_add_c2(a, 3, 1, c2, c3, c1); | |
563 | sqr_add_c2(a, 4, 0, c2, c3, c1); | |
564 | r[4] = c2; | |
565 | c2 = 0; | |
566 | sqr_add_c2(a, 5, 0, c3, c1, c2); | |
567 | sqr_add_c2(a, 4, 1, c3, c1, c2); | |
568 | sqr_add_c2(a, 3, 2, c3, c1, c2); | |
569 | r[5] = c3; | |
570 | c3 = 0; | |
571 | sqr_add_c(a, 3, c1, c2, c3); | |
572 | sqr_add_c2(a, 4, 2, c1, c2, c3); | |
573 | sqr_add_c2(a, 5, 1, c1, c2, c3); | |
574 | sqr_add_c2(a, 6, 0, c1, c2, c3); | |
575 | r[6] = c1; | |
576 | c1 = 0; | |
577 | sqr_add_c2(a, 7, 0, c2, c3, c1); | |
578 | sqr_add_c2(a, 6, 1, c2, c3, c1); | |
579 | sqr_add_c2(a, 5, 2, c2, c3, c1); | |
580 | sqr_add_c2(a, 4, 3, c2, c3, c1); | |
581 | r[7] = c2; | |
582 | c2 = 0; | |
583 | sqr_add_c(a, 4, c3, c1, c2); | |
584 | sqr_add_c2(a, 5, 3, c3, c1, c2); | |
585 | sqr_add_c2(a, 6, 2, c3, c1, c2); | |
586 | sqr_add_c2(a, 7, 1, c3, c1, c2); | |
587 | r[8] = c3; | |
588 | c3 = 0; | |
589 | sqr_add_c2(a, 7, 2, c1, c2, c3); | |
590 | sqr_add_c2(a, 6, 3, c1, c2, c3); | |
591 | sqr_add_c2(a, 5, 4, c1, c2, c3); | |
592 | r[9] = c1; | |
593 | c1 = 0; | |
594 | sqr_add_c(a, 5, c2, c3, c1); | |
595 | sqr_add_c2(a, 6, 4, c2, c3, c1); | |
596 | sqr_add_c2(a, 7, 3, c2, c3, c1); | |
597 | r[10] = c2; | |
598 | c2 = 0; | |
599 | sqr_add_c2(a, 7, 4, c3, c1, c2); | |
600 | sqr_add_c2(a, 6, 5, c3, c1, c2); | |
601 | r[11] = c3; | |
602 | c3 = 0; | |
603 | sqr_add_c(a, 6, c1, c2, c3); | |
604 | sqr_add_c2(a, 7, 5, c1, c2, c3); | |
605 | r[12] = c1; | |
606 | c1 = 0; | |
607 | sqr_add_c2(a, 7, 6, c2, c3, c1); | |
608 | r[13] = c2; | |
609 | c2 = 0; | |
610 | sqr_add_c(a, 7, c3, c1, c2); | |
611 | r[14] = c3; | |
612 | r[15] = c1; | |
613 | } | |
2f98abbc | 614 | |
41b76195 | 615 | void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a) |
0f113f3e MC |
616 | { |
617 | BN_ULONG c1, c2, c3; | |
618 | ||
619 | c1 = 0; | |
620 | c2 = 0; | |
621 | c3 = 0; | |
622 | sqr_add_c(a, 0, c1, c2, c3); | |
623 | r[0] = c1; | |
624 | c1 = 0; | |
625 | sqr_add_c2(a, 1, 0, c2, c3, c1); | |
626 | r[1] = c2; | |
627 | c2 = 0; | |
628 | sqr_add_c(a, 1, c3, c1, c2); | |
629 | sqr_add_c2(a, 2, 0, c3, c1, c2); | |
630 | r[2] = c3; | |
631 | c3 = 0; | |
632 | sqr_add_c2(a, 3, 0, c1, c2, c3); | |
633 | sqr_add_c2(a, 2, 1, c1, c2, c3); | |
634 | r[3] = c1; | |
635 | c1 = 0; | |
636 | sqr_add_c(a, 2, c2, c3, c1); | |
637 | sqr_add_c2(a, 3, 1, c2, c3, c1); | |
638 | r[4] = c2; | |
639 | c2 = 0; | |
640 | sqr_add_c2(a, 3, 2, c3, c1, c2); | |
641 | r[5] = c3; | |
642 | c3 = 0; | |
643 | sqr_add_c(a, 3, c1, c2, c3); | |
644 | r[6] = c1; | |
645 | r[7] = c2; | |
646 | } | |
c8a0d0aa | 647 | #endif |