]>
git.ipfire.org Git - thirdparty/openssl.git/blob - crypto/bn/bn_asm.c
1 /* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
4 * This package is an SSL implementation written
5 * by Eric Young (eay@cryptsoft.com).
6 * The implementation was written so as to conform with Netscapes SSL.
8 * This library is free for commercial and non-commercial use as long as
9 * the following conditions are aheared to. The following conditions
10 * apply to all code found in this distribution, be it the RC4, RSA,
11 * lhash, DES, etc., code; not just the SSL code. The SSL documentation
12 * included with this distribution is covered by the same copyright terms
13 * except that the holder is Tim Hudson (tjh@cryptsoft.com).
15 * Copyright remains Eric Young's, and as such any Copyright notices in
16 * the code are not to be removed.
17 * If this package is used in a product, Eric Young should be given attribution
18 * as the author of the parts of the library used.
19 * This can be in the form of a textual message at program startup or
20 * in documentation (online or textual) provided with the package.
22 * Redistribution and use in source and binary forms, with or without
23 * modification, are permitted provided that the following conditions
25 * 1. Redistributions of source code must retain the copyright
26 * notice, this list of conditions and the following disclaimer.
27 * 2. Redistributions in binary form must reproduce the above copyright
28 * notice, this list of conditions and the following disclaimer in the
29 * documentation and/or other materials provided with the distribution.
30 * 3. All advertising materials mentioning features or use of this software
31 * must display the following acknowledgement:
32 * "This product includes cryptographic software written by
33 * Eric Young (eay@cryptsoft.com)"
34 * The word 'cryptographic' can be left out if the rouines from the library
35 * being used are not cryptographic related :-).
36 * 4. If you include any Windows specific code (or a derivative thereof) from
37 * the apps directory (application code) you must include an acknowledgement:
38 * "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
40 * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
41 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
42 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
43 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
44 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
45 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
46 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
47 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
48 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
49 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
52 * The licence and distribution terms for any publically available version or
53 * derivative of this code cannot be changed. i.e. this code cannot simply be
54 * copied and put under another distribution licence
55 * [including the GNU Public Licence.]
59 #include <openssl/crypto.h>
60 #include "internal/cryptlib.h"
63 #if defined(BN_LLONG) || defined(BN_UMULT_HIGH)
65 BN_ULONG
bn_mul_add_words(BN_ULONG
*rp
, const BN_ULONG
*ap
, int num
,
74 # ifndef OPENSSL_SMALL_FOOTPRINT
76 mul_add(rp
[0], ap
[0], w
, c1
);
77 mul_add(rp
[1], ap
[1], w
, c1
);
78 mul_add(rp
[2], ap
[2], w
, c1
);
79 mul_add(rp
[3], ap
[3], w
, c1
);
86 mul_add(rp
[0], ap
[0], w
, c1
);
95 BN_ULONG
bn_mul_words(BN_ULONG
*rp
, const BN_ULONG
*ap
, int num
, BN_ULONG w
)
103 # ifndef OPENSSL_SMALL_FOOTPRINT
105 mul(rp
[0], ap
[0], w
, c1
);
106 mul(rp
[1], ap
[1], w
, c1
);
107 mul(rp
[2], ap
[2], w
, c1
);
108 mul(rp
[3], ap
[3], w
, c1
);
115 mul(rp
[0], ap
[0], w
, c1
);
123 void bn_sqr_words(BN_ULONG
*r
, const BN_ULONG
*a
, int n
)
129 # ifndef OPENSSL_SMALL_FOOTPRINT
131 sqr(r
[0], r
[1], a
[0]);
132 sqr(r
[2], r
[3], a
[1]);
133 sqr(r
[4], r
[5], a
[2]);
134 sqr(r
[6], r
[7], a
[3]);
141 sqr(r
[0], r
[1], a
[0]);
148 #else /* !(defined(BN_LLONG) ||
149 * defined(BN_UMULT_HIGH)) */
151 BN_ULONG
bn_mul_add_words(BN_ULONG
*rp
, const BN_ULONG
*ap
, int num
,
159 return ((BN_ULONG
)0);
164 # ifndef OPENSSL_SMALL_FOOTPRINT
166 mul_add(rp
[0], ap
[0], bl
, bh
, c
);
167 mul_add(rp
[1], ap
[1], bl
, bh
, c
);
168 mul_add(rp
[2], ap
[2], bl
, bh
, c
);
169 mul_add(rp
[3], ap
[3], bl
, bh
, c
);
176 mul_add(rp
[0], ap
[0], bl
, bh
, c
);
184 BN_ULONG
bn_mul_words(BN_ULONG
*rp
, const BN_ULONG
*ap
, int num
, BN_ULONG w
)
191 return ((BN_ULONG
)0);
196 # ifndef OPENSSL_SMALL_FOOTPRINT
198 mul(rp
[0], ap
[0], bl
, bh
, carry
);
199 mul(rp
[1], ap
[1], bl
, bh
, carry
);
200 mul(rp
[2], ap
[2], bl
, bh
, carry
);
201 mul(rp
[3], ap
[3], bl
, bh
, carry
);
208 mul(rp
[0], ap
[0], bl
, bh
, carry
);
216 void bn_sqr_words(BN_ULONG
*r
, const BN_ULONG
*a
, int n
)
222 # ifndef OPENSSL_SMALL_FOOTPRINT
224 sqr64(r
[0], r
[1], a
[0]);
225 sqr64(r
[2], r
[3], a
[1]);
226 sqr64(r
[4], r
[5], a
[2]);
227 sqr64(r
[6], r
[7], a
[3]);
234 sqr64(r
[0], r
[1], a
[0]);
241 #endif /* !(defined(BN_LLONG) ||
242 * defined(BN_UMULT_HIGH)) */
244 #if defined(BN_LLONG) && defined(BN_DIV2W)
246 BN_ULONG
bn_div_words(BN_ULONG h
, BN_ULONG l
, BN_ULONG d
)
248 return ((BN_ULONG
)(((((BN_ULLONG
) h
) << BN_BITS2
) | l
) / (BN_ULLONG
) d
));
253 /* Divide h,l by d and return the result. */
254 /* I need to test this some more :-( */
255 BN_ULONG
bn_div_words(BN_ULONG h
, BN_ULONG l
, BN_ULONG d
)
257 BN_ULONG dh
, dl
, q
, ret
= 0, th
, tl
, t
;
263 i
= BN_num_bits_word(d
);
264 assert((i
== BN_BITS2
) || (h
<= (BN_ULONG
)1 << i
));
272 h
= (h
<< i
) | (l
>> (BN_BITS2
- i
));
275 dh
= (d
& BN_MASK2h
) >> BN_BITS4
;
276 dl
= (d
& BN_MASK2l
);
278 if ((h
>> BN_BITS4
) == dh
)
287 if ((t
& BN_MASK2h
) ||
288 ((tl
) <= ((t
<< BN_BITS4
) | ((l
& BN_MASK2h
) >> BN_BITS4
))))
294 t
= (tl
>> BN_BITS4
);
295 tl
= (tl
<< BN_BITS4
) & BN_MASK2h
;
311 h
= ((h
<< BN_BITS4
) | (l
>> BN_BITS4
)) & BN_MASK2
;
312 l
= (l
& BN_MASK2l
) << BN_BITS4
;
317 #endif /* !defined(BN_LLONG) && defined(BN_DIV2W) */
320 BN_ULONG
bn_add_words(BN_ULONG
*r
, const BN_ULONG
*a
, const BN_ULONG
*b
,
327 return ((BN_ULONG
)0);
329 # ifndef OPENSSL_SMALL_FOOTPRINT
331 ll
+= (BN_ULLONG
) a
[0] + b
[0];
332 r
[0] = (BN_ULONG
)ll
& BN_MASK2
;
334 ll
+= (BN_ULLONG
) a
[1] + b
[1];
335 r
[1] = (BN_ULONG
)ll
& BN_MASK2
;
337 ll
+= (BN_ULLONG
) a
[2] + b
[2];
338 r
[2] = (BN_ULONG
)ll
& BN_MASK2
;
340 ll
+= (BN_ULLONG
) a
[3] + b
[3];
341 r
[3] = (BN_ULONG
)ll
& BN_MASK2
;
350 ll
+= (BN_ULLONG
) a
[0] + b
[0];
351 r
[0] = (BN_ULONG
)ll
& BN_MASK2
;
358 return ((BN_ULONG
)ll
);
360 #else /* !BN_LLONG */
361 BN_ULONG
bn_add_words(BN_ULONG
*r
, const BN_ULONG
*a
, const BN_ULONG
*b
,
368 return ((BN_ULONG
)0);
371 # ifndef OPENSSL_SMALL_FOOTPRINT
374 t
= (t
+ c
) & BN_MASK2
;
376 l
= (t
+ b
[0]) & BN_MASK2
;
380 t
= (t
+ c
) & BN_MASK2
;
382 l
= (t
+ b
[1]) & BN_MASK2
;
386 t
= (t
+ c
) & BN_MASK2
;
388 l
= (t
+ b
[2]) & BN_MASK2
;
392 t
= (t
+ c
) & BN_MASK2
;
394 l
= (t
+ b
[3]) & BN_MASK2
;
405 t
= (t
+ c
) & BN_MASK2
;
407 l
= (t
+ b
[0]) & BN_MASK2
;
415 return ((BN_ULONG
)c
);
417 #endif /* !BN_LLONG */
419 BN_ULONG
bn_sub_words(BN_ULONG
*r
, const BN_ULONG
*a
, const BN_ULONG
*b
,
427 return ((BN_ULONG
)0);
429 #ifndef OPENSSL_SMALL_FOOTPRINT
433 r
[0] = (t1
- t2
- c
) & BN_MASK2
;
438 r
[1] = (t1
- t2
- c
) & BN_MASK2
;
443 r
[2] = (t1
- t2
- c
) & BN_MASK2
;
448 r
[3] = (t1
- t2
- c
) & BN_MASK2
;
460 r
[0] = (t1
- t2
- c
) & BN_MASK2
;
471 #if defined(BN_MUL_COMBA) && !defined(OPENSSL_SMALL_FOOTPRINT)
473 # undef bn_mul_comba8
474 # undef bn_mul_comba4
475 # undef bn_sqr_comba8
476 # undef bn_sqr_comba4
478 /* mul_add_c(a,b,c0,c1,c2) -- c+=a*b for three word number c=(c2,c1,c0) */
479 /* mul_add_c2(a,b,c0,c1,c2) -- c+=2*a*b for three word number c=(c2,c1,c0) */
480 /* sqr_add_c(a,i,c0,c1,c2) -- c+=a[i]^2 for three word number c=(c2,c1,c0) */
482 * sqr_add_c2(a,i,c0,c1,c2) -- c+=2*a[i]*a[j] for three word number
488 * Keep in mind that additions to multiplication result can not
489 * overflow, because its high half cannot be all-ones.
491 # define mul_add_c(a,b,c0,c1,c2) do { \
493 BN_ULLONG t = (BN_ULLONG)(a)*(b); \
494 t += c0; /* no carry */ \
495 c0 = (BN_ULONG)Lw(t); \
496 hi = (BN_ULONG)Hw(t); \
497 c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
500 # define mul_add_c2(a,b,c0,c1,c2) do { \
502 BN_ULLONG t = (BN_ULLONG)(a)*(b); \
503 BN_ULLONG tt = t+c0; /* no carry */ \
504 c0 = (BN_ULONG)Lw(tt); \
505 hi = (BN_ULONG)Hw(tt); \
506 c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
507 t += c0; /* no carry */ \
508 c0 = (BN_ULONG)Lw(t); \
509 hi = (BN_ULONG)Hw(t); \
510 c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
513 # define sqr_add_c(a,i,c0,c1,c2) do { \
515 BN_ULLONG t = (BN_ULLONG)a[i]*a[i]; \
516 t += c0; /* no carry */ \
517 c0 = (BN_ULONG)Lw(t); \
518 hi = (BN_ULONG)Hw(t); \
519 c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
522 # define sqr_add_c2(a,i,j,c0,c1,c2) \
523 mul_add_c2((a)[i],(a)[j],c0,c1,c2)
525 # elif defined(BN_UMULT_LOHI)
527 * Keep in mind that additions to hi can not overflow, because
528 * the high word of a multiplication result cannot be all-ones.
530 # define mul_add_c(a,b,c0,c1,c2) do { \
531 BN_ULONG ta = (a), tb = (b); \
533 BN_UMULT_LOHI(lo,hi,ta,tb); \
534 c0 += lo; hi += (c0<lo)?1:0; \
535 c1 += hi; c2 += (c1<hi)?1:0; \
538 # define mul_add_c2(a,b,c0,c1,c2) do { \
539 BN_ULONG ta = (a), tb = (b); \
540 BN_ULONG lo, hi, tt; \
541 BN_UMULT_LOHI(lo,hi,ta,tb); \
542 c0 += lo; tt = hi+((c0<lo)?1:0); \
543 c1 += tt; c2 += (c1<tt)?1:0; \
544 c0 += lo; hi += (c0<lo)?1:0; \
545 c1 += hi; c2 += (c1<hi)?1:0; \
548 # define sqr_add_c(a,i,c0,c1,c2) do { \
549 BN_ULONG ta = (a)[i]; \
551 BN_UMULT_LOHI(lo,hi,ta,ta); \
552 c0 += lo; hi += (c0<lo)?1:0; \
553 c1 += hi; c2 += (c1<hi)?1:0; \
556 # define sqr_add_c2(a,i,j,c0,c1,c2) \
557 mul_add_c2((a)[i],(a)[j],c0,c1,c2)
559 # elif defined(BN_UMULT_HIGH)
561 * Keep in mind that additions to hi can not overflow, because
562 * the high word of a multiplication result cannot be all-ones.
564 # define mul_add_c(a,b,c0,c1,c2) do { \
565 BN_ULONG ta = (a), tb = (b); \
566 BN_ULONG lo = ta * tb; \
567 BN_ULONG hi = BN_UMULT_HIGH(ta,tb); \
568 c0 += lo; hi += (c0<lo)?1:0; \
569 c1 += hi; c2 += (c1<hi)?1:0; \
572 # define mul_add_c2(a,b,c0,c1,c2) do { \
573 BN_ULONG ta = (a), tb = (b), tt; \
574 BN_ULONG lo = ta * tb; \
575 BN_ULONG hi = BN_UMULT_HIGH(ta,tb); \
576 c0 += lo; tt = hi + ((c0<lo)?1:0); \
577 c1 += tt; c2 += (c1<tt)?1:0; \
578 c0 += lo; hi += (c0<lo)?1:0; \
579 c1 += hi; c2 += (c1<hi)?1:0; \
582 # define sqr_add_c(a,i,c0,c1,c2) do { \
583 BN_ULONG ta = (a)[i]; \
584 BN_ULONG lo = ta * ta; \
585 BN_ULONG hi = BN_UMULT_HIGH(ta,ta); \
586 c0 += lo; hi += (c0<lo)?1:0; \
587 c1 += hi; c2 += (c1<hi)?1:0; \
590 # define sqr_add_c2(a,i,j,c0,c1,c2) \
591 mul_add_c2((a)[i],(a)[j],c0,c1,c2)
593 # else /* !BN_LLONG */
595 * Keep in mind that additions to hi can not overflow, because
596 * the high word of a multiplication result cannot be all-ones.
598 # define mul_add_c(a,b,c0,c1,c2) do { \
599 BN_ULONG lo = LBITS(a), hi = HBITS(a); \
600 BN_ULONG bl = LBITS(b), bh = HBITS(b); \
601 mul64(lo,hi,bl,bh); \
602 c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++; \
603 c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
606 # define mul_add_c2(a,b,c0,c1,c2) do { \
608 BN_ULONG lo = LBITS(a), hi = HBITS(a); \
609 BN_ULONG bl = LBITS(b), bh = HBITS(b); \
610 mul64(lo,hi,bl,bh); \
612 c0 = (c0+lo)&BN_MASK2; if (c0<lo) tt++; \
613 c1 = (c1+tt)&BN_MASK2; if (c1<tt) c2++; \
614 c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++; \
615 c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
618 # define sqr_add_c(a,i,c0,c1,c2) do { \
620 sqr64(lo,hi,(a)[i]); \
621 c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++; \
622 c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
625 # define sqr_add_c2(a,i,j,c0,c1,c2) \
626 mul_add_c2((a)[i],(a)[j],c0,c1,c2)
627 # endif /* !BN_LLONG */
629 void bn_mul_comba8(BN_ULONG
*r
, BN_ULONG
*a
, BN_ULONG
*b
)
636 mul_add_c(a
[0], b
[0], c1
, c2
, c3
);
639 mul_add_c(a
[0], b
[1], c2
, c3
, c1
);
640 mul_add_c(a
[1], b
[0], c2
, c3
, c1
);
643 mul_add_c(a
[2], b
[0], c3
, c1
, c2
);
644 mul_add_c(a
[1], b
[1], c3
, c1
, c2
);
645 mul_add_c(a
[0], b
[2], c3
, c1
, c2
);
648 mul_add_c(a
[0], b
[3], c1
, c2
, c3
);
649 mul_add_c(a
[1], b
[2], c1
, c2
, c3
);
650 mul_add_c(a
[2], b
[1], c1
, c2
, c3
);
651 mul_add_c(a
[3], b
[0], c1
, c2
, c3
);
654 mul_add_c(a
[4], b
[0], c2
, c3
, c1
);
655 mul_add_c(a
[3], b
[1], c2
, c3
, c1
);
656 mul_add_c(a
[2], b
[2], c2
, c3
, c1
);
657 mul_add_c(a
[1], b
[3], c2
, c3
, c1
);
658 mul_add_c(a
[0], b
[4], c2
, c3
, c1
);
661 mul_add_c(a
[0], b
[5], c3
, c1
, c2
);
662 mul_add_c(a
[1], b
[4], c3
, c1
, c2
);
663 mul_add_c(a
[2], b
[3], c3
, c1
, c2
);
664 mul_add_c(a
[3], b
[2], c3
, c1
, c2
);
665 mul_add_c(a
[4], b
[1], c3
, c1
, c2
);
666 mul_add_c(a
[5], b
[0], c3
, c1
, c2
);
669 mul_add_c(a
[6], b
[0], c1
, c2
, c3
);
670 mul_add_c(a
[5], b
[1], c1
, c2
, c3
);
671 mul_add_c(a
[4], b
[2], c1
, c2
, c3
);
672 mul_add_c(a
[3], b
[3], c1
, c2
, c3
);
673 mul_add_c(a
[2], b
[4], c1
, c2
, c3
);
674 mul_add_c(a
[1], b
[5], c1
, c2
, c3
);
675 mul_add_c(a
[0], b
[6], c1
, c2
, c3
);
678 mul_add_c(a
[0], b
[7], c2
, c3
, c1
);
679 mul_add_c(a
[1], b
[6], c2
, c3
, c1
);
680 mul_add_c(a
[2], b
[5], c2
, c3
, c1
);
681 mul_add_c(a
[3], b
[4], c2
, c3
, c1
);
682 mul_add_c(a
[4], b
[3], c2
, c3
, c1
);
683 mul_add_c(a
[5], b
[2], c2
, c3
, c1
);
684 mul_add_c(a
[6], b
[1], c2
, c3
, c1
);
685 mul_add_c(a
[7], b
[0], c2
, c3
, c1
);
688 mul_add_c(a
[7], b
[1], c3
, c1
, c2
);
689 mul_add_c(a
[6], b
[2], c3
, c1
, c2
);
690 mul_add_c(a
[5], b
[3], c3
, c1
, c2
);
691 mul_add_c(a
[4], b
[4], c3
, c1
, c2
);
692 mul_add_c(a
[3], b
[5], c3
, c1
, c2
);
693 mul_add_c(a
[2], b
[6], c3
, c1
, c2
);
694 mul_add_c(a
[1], b
[7], c3
, c1
, c2
);
697 mul_add_c(a
[2], b
[7], c1
, c2
, c3
);
698 mul_add_c(a
[3], b
[6], c1
, c2
, c3
);
699 mul_add_c(a
[4], b
[5], c1
, c2
, c3
);
700 mul_add_c(a
[5], b
[4], c1
, c2
, c3
);
701 mul_add_c(a
[6], b
[3], c1
, c2
, c3
);
702 mul_add_c(a
[7], b
[2], c1
, c2
, c3
);
705 mul_add_c(a
[7], b
[3], c2
, c3
, c1
);
706 mul_add_c(a
[6], b
[4], c2
, c3
, c1
);
707 mul_add_c(a
[5], b
[5], c2
, c3
, c1
);
708 mul_add_c(a
[4], b
[6], c2
, c3
, c1
);
709 mul_add_c(a
[3], b
[7], c2
, c3
, c1
);
712 mul_add_c(a
[4], b
[7], c3
, c1
, c2
);
713 mul_add_c(a
[5], b
[6], c3
, c1
, c2
);
714 mul_add_c(a
[6], b
[5], c3
, c1
, c2
);
715 mul_add_c(a
[7], b
[4], c3
, c1
, c2
);
718 mul_add_c(a
[7], b
[5], c1
, c2
, c3
);
719 mul_add_c(a
[6], b
[6], c1
, c2
, c3
);
720 mul_add_c(a
[5], b
[7], c1
, c2
, c3
);
723 mul_add_c(a
[6], b
[7], c2
, c3
, c1
);
724 mul_add_c(a
[7], b
[6], c2
, c3
, c1
);
727 mul_add_c(a
[7], b
[7], c3
, c1
, c2
);
732 void bn_mul_comba4(BN_ULONG
*r
, BN_ULONG
*a
, BN_ULONG
*b
)
739 mul_add_c(a
[0], b
[0], c1
, c2
, c3
);
742 mul_add_c(a
[0], b
[1], c2
, c3
, c1
);
743 mul_add_c(a
[1], b
[0], c2
, c3
, c1
);
746 mul_add_c(a
[2], b
[0], c3
, c1
, c2
);
747 mul_add_c(a
[1], b
[1], c3
, c1
, c2
);
748 mul_add_c(a
[0], b
[2], c3
, c1
, c2
);
751 mul_add_c(a
[0], b
[3], c1
, c2
, c3
);
752 mul_add_c(a
[1], b
[2], c1
, c2
, c3
);
753 mul_add_c(a
[2], b
[1], c1
, c2
, c3
);
754 mul_add_c(a
[3], b
[0], c1
, c2
, c3
);
757 mul_add_c(a
[3], b
[1], c2
, c3
, c1
);
758 mul_add_c(a
[2], b
[2], c2
, c3
, c1
);
759 mul_add_c(a
[1], b
[3], c2
, c3
, c1
);
762 mul_add_c(a
[2], b
[3], c3
, c1
, c2
);
763 mul_add_c(a
[3], b
[2], c3
, c1
, c2
);
766 mul_add_c(a
[3], b
[3], c1
, c2
, c3
);
771 void bn_sqr_comba8(BN_ULONG
*r
, const BN_ULONG
*a
)
778 sqr_add_c(a
, 0, c1
, c2
, c3
);
781 sqr_add_c2(a
, 1, 0, c2
, c3
, c1
);
784 sqr_add_c(a
, 1, c3
, c1
, c2
);
785 sqr_add_c2(a
, 2, 0, c3
, c1
, c2
);
788 sqr_add_c2(a
, 3, 0, c1
, c2
, c3
);
789 sqr_add_c2(a
, 2, 1, c1
, c2
, c3
);
792 sqr_add_c(a
, 2, c2
, c3
, c1
);
793 sqr_add_c2(a
, 3, 1, c2
, c3
, c1
);
794 sqr_add_c2(a
, 4, 0, c2
, c3
, c1
);
797 sqr_add_c2(a
, 5, 0, c3
, c1
, c2
);
798 sqr_add_c2(a
, 4, 1, c3
, c1
, c2
);
799 sqr_add_c2(a
, 3, 2, c3
, c1
, c2
);
802 sqr_add_c(a
, 3, c1
, c2
, c3
);
803 sqr_add_c2(a
, 4, 2, c1
, c2
, c3
);
804 sqr_add_c2(a
, 5, 1, c1
, c2
, c3
);
805 sqr_add_c2(a
, 6, 0, c1
, c2
, c3
);
808 sqr_add_c2(a
, 7, 0, c2
, c3
, c1
);
809 sqr_add_c2(a
, 6, 1, c2
, c3
, c1
);
810 sqr_add_c2(a
, 5, 2, c2
, c3
, c1
);
811 sqr_add_c2(a
, 4, 3, c2
, c3
, c1
);
814 sqr_add_c(a
, 4, c3
, c1
, c2
);
815 sqr_add_c2(a
, 5, 3, c3
, c1
, c2
);
816 sqr_add_c2(a
, 6, 2, c3
, c1
, c2
);
817 sqr_add_c2(a
, 7, 1, c3
, c1
, c2
);
820 sqr_add_c2(a
, 7, 2, c1
, c2
, c3
);
821 sqr_add_c2(a
, 6, 3, c1
, c2
, c3
);
822 sqr_add_c2(a
, 5, 4, c1
, c2
, c3
);
825 sqr_add_c(a
, 5, c2
, c3
, c1
);
826 sqr_add_c2(a
, 6, 4, c2
, c3
, c1
);
827 sqr_add_c2(a
, 7, 3, c2
, c3
, c1
);
830 sqr_add_c2(a
, 7, 4, c3
, c1
, c2
);
831 sqr_add_c2(a
, 6, 5, c3
, c1
, c2
);
834 sqr_add_c(a
, 6, c1
, c2
, c3
);
835 sqr_add_c2(a
, 7, 5, c1
, c2
, c3
);
838 sqr_add_c2(a
, 7, 6, c2
, c3
, c1
);
841 sqr_add_c(a
, 7, c3
, c1
, c2
);
846 void bn_sqr_comba4(BN_ULONG
*r
, const BN_ULONG
*a
)
853 sqr_add_c(a
, 0, c1
, c2
, c3
);
856 sqr_add_c2(a
, 1, 0, c2
, c3
, c1
);
859 sqr_add_c(a
, 1, c3
, c1
, c2
);
860 sqr_add_c2(a
, 2, 0, c3
, c1
, c2
);
863 sqr_add_c2(a
, 3, 0, c1
, c2
, c3
);
864 sqr_add_c2(a
, 2, 1, c1
, c2
, c3
);
867 sqr_add_c(a
, 2, c2
, c3
, c1
);
868 sqr_add_c2(a
, 3, 1, c2
, c3
, c1
);
871 sqr_add_c2(a
, 3, 2, c3
, c1
, c2
);
874 sqr_add_c(a
, 3, c1
, c2
, c3
);
879 # ifdef OPENSSL_NO_ASM
880 # ifdef OPENSSL_BN_ASM_MONT
883 * This is essentially reference implementation, which may or may not
884 * result in performance improvement. E.g. on IA-32 this routine was
885 * observed to give 40% faster rsa1024 private key operations and 10%
886 * faster rsa4096 ones, while on AMD64 it improves rsa1024 sign only
887 * by 10% and *worsens* rsa4096 sign by 15%. Once again, it's a
888 * reference implementation, one to be used as starting point for
889 * platform-specific assembler. Mentioned numbers apply to compiler
890 * generated code compiled with and without -DOPENSSL_BN_ASM_MONT and
891 * can vary not only from platform to platform, but even for compiler
892 * versions. Assembler vs. assembler improvement coefficients can
893 * [and are known to] differ and are to be documented elsewhere.
895 int bn_mul_mont(BN_ULONG
*rp
, const BN_ULONG
*ap
, const BN_ULONG
*bp
,
896 const BN_ULONG
*np
, const BN_ULONG
*n0p
, int num
)
898 BN_ULONG c0
, c1
, ml
, *tp
, n0
;
902 volatile BN_ULONG
*vp
;
905 # if 0 /* template for platform-specific
908 return bn_sqr_mont(rp
, ap
, np
, n0p
, num
);
910 vp
= tp
= alloca((num
+ 2) * sizeof(BN_ULONG
));
919 for (j
= 0; j
< num
; ++j
)
920 mul(tp
[j
], ap
[j
], ml
, mh
, c0
);
922 for (j
= 0; j
< num
; ++j
)
923 mul(tp
[j
], ap
[j
], ml
, c0
);
930 for (i
= 0; i
< num
; i
++) {
936 for (j
= 0; j
< num
; ++j
)
937 mul_add(tp
[j
], ap
[j
], ml
, mh
, c0
);
939 for (j
= 0; j
< num
; ++j
)
940 mul_add(tp
[j
], ap
[j
], ml
, c0
);
942 c1
= (tp
[num
] + c0
) & BN_MASK2
;
944 tp
[num
+ 1] = (c1
< c0
? 1 : 0);
947 ml
= (c1
* n0
) & BN_MASK2
;
952 mul_add(c1
, np
[0], ml
, mh
, c0
);
954 mul_add(c1
, ml
, np
[0], c0
);
956 for (j
= 1; j
< num
; j
++) {
959 mul_add(c1
, np
[j
], ml
, mh
, c0
);
961 mul_add(c1
, ml
, np
[j
], c0
);
963 tp
[j
- 1] = c1
& BN_MASK2
;
965 c1
= (tp
[num
] + c0
) & BN_MASK2
;
967 tp
[num
] = tp
[num
+ 1] + (c1
< c0
? 1 : 0);
970 if (tp
[num
] != 0 || tp
[num
- 1] >= np
[num
- 1]) {
971 c0
= bn_sub_words(rp
, tp
, np
, num
);
972 if (tp
[num
] != 0 || c0
== 0) {
973 for (i
= 0; i
< num
+ 2; i
++)
978 for (i
= 0; i
< num
; i
++)
979 rp
[i
] = tp
[i
], vp
[i
] = 0;
986 * Return value of 0 indicates that multiplication/convolution was not
987 * performed to signal the caller to fall down to alternative/original
990 int bn_mul_mont(BN_ULONG
*rp
, const BN_ULONG
*ap
, const BN_ULONG
*bp
,
991 const BN_ULONG
*np
, const BN_ULONG
*n0
, int num
)
995 # endif /* OPENSSL_BN_ASM_MONT */
998 #else /* !BN_MUL_COMBA */
1000 /* hmm... is it faster just to do a multiply? */
1001 # undef bn_sqr_comba4
1002 # undef bn_sqr_comba8
1003 void bn_sqr_comba4(BN_ULONG
*r
, const BN_ULONG
*a
)
1006 bn_sqr_normal(r
, a
, 4, t
);
1009 void bn_sqr_comba8(BN_ULONG
*r
, const BN_ULONG
*a
)
1012 bn_sqr_normal(r
, a
, 8, t
);
1015 void bn_mul_comba4(BN_ULONG
*r
, BN_ULONG
*a
, BN_ULONG
*b
)
1017 r
[4] = bn_mul_words(&(r
[0]), a
, 4, b
[0]);
1018 r
[5] = bn_mul_add_words(&(r
[1]), a
, 4, b
[1]);
1019 r
[6] = bn_mul_add_words(&(r
[2]), a
, 4, b
[2]);
1020 r
[7] = bn_mul_add_words(&(r
[3]), a
, 4, b
[3]);
1023 void bn_mul_comba8(BN_ULONG
*r
, BN_ULONG
*a
, BN_ULONG
*b
)
1025 r
[8] = bn_mul_words(&(r
[0]), a
, 8, b
[0]);
1026 r
[9] = bn_mul_add_words(&(r
[1]), a
, 8, b
[1]);
1027 r
[10] = bn_mul_add_words(&(r
[2]), a
, 8, b
[2]);
1028 r
[11] = bn_mul_add_words(&(r
[3]), a
, 8, b
[3]);
1029 r
[12] = bn_mul_add_words(&(r
[4]), a
, 8, b
[4]);
1030 r
[13] = bn_mul_add_words(&(r
[5]), a
, 8, b
[5]);
1031 r
[14] = bn_mul_add_words(&(r
[6]), a
, 8, b
[6]);
1032 r
[15] = bn_mul_add_words(&(r
[7]), a
, 8, b
[7]);
1035 # ifdef OPENSSL_NO_ASM
1036 # ifdef OPENSSL_BN_ASM_MONT
1037 # include <alloca.h>
1038 int bn_mul_mont(BN_ULONG
*rp
, const BN_ULONG
*ap
, const BN_ULONG
*bp
,
1039 const BN_ULONG
*np
, const BN_ULONG
*n0p
, int num
)
1041 BN_ULONG c0
, c1
, *tp
, n0
= *n0p
;
1042 volatile BN_ULONG
*vp
;
1045 vp
= tp
= alloca((num
+ 2) * sizeof(BN_ULONG
));
1047 for (i
= 0; i
<= num
; i
++)
1050 for (i
= 0; i
< num
; i
++) {
1051 c0
= bn_mul_add_words(tp
, ap
, num
, bp
[i
]);
1052 c1
= (tp
[num
] + c0
) & BN_MASK2
;
1054 tp
[num
+ 1] = (c1
< c0
? 1 : 0);
1056 c0
= bn_mul_add_words(tp
, np
, num
, tp
[0] * n0
);
1057 c1
= (tp
[num
] + c0
) & BN_MASK2
;
1059 tp
[num
+ 1] += (c1
< c0
? 1 : 0);
1060 for (j
= 0; j
<= num
; j
++)
1064 if (tp
[num
] != 0 || tp
[num
- 1] >= np
[num
- 1]) {
1065 c0
= bn_sub_words(rp
, tp
, np
, num
);
1066 if (tp
[num
] != 0 || c0
== 0) {
1067 for (i
= 0; i
< num
+ 2; i
++)
1072 for (i
= 0; i
< num
; i
++)
1073 rp
[i
] = tp
[i
], vp
[i
] = 0;
1079 int bn_mul_mont(BN_ULONG
*rp
, const BN_ULONG
*ap
, const BN_ULONG
*bp
,
1080 const BN_ULONG
*np
, const BN_ULONG
*n0
, int num
)
1084 # endif /* OPENSSL_BN_ASM_MONT */
1087 #endif /* !BN_MUL_COMBA */