1 /* crypto/bn/bn_asm.c */
2 /* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
5 * This package is an SSL implementation written
6 * by Eric Young (eay@cryptsoft.com).
7 * The implementation was written so as to conform with Netscapes SSL.
9 * This library is free for commercial and non-commercial use as long as
10 * the following conditions are aheared to. The following conditions
11 * apply to all code found in this distribution, be it the RC4, RSA,
12 * lhash, DES, etc., code; not just the SSL code. The SSL documentation
13 * included with this distribution is covered by the same copyright terms
14 * except that the holder is Tim Hudson (tjh@cryptsoft.com).
16 * Copyright remains Eric Young's, and as such any Copyright notices in
17 * the code are not to be removed.
18 * If this package is used in a product, Eric Young should be given attribution
19 * as the author of the parts of the library used.
20 * This can be in the form of a textual message at program startup or
21 * in documentation (online or textual) provided with the package.
23 * Redistribution and use in source and binary forms, with or without
24 * modification, are permitted provided that the following conditions
26 * 1. Redistributions of source code must retain the copyright
27 * notice, this list of conditions and the following disclaimer.
28 * 2. Redistributions in binary form must reproduce the above copyright
29 * notice, this list of conditions and the following disclaimer in the
30 * documentation and/or other materials provided with the distribution.
31 * 3. All advertising materials mentioning features or use of this software
32 * must display the following acknowledgement:
33 * "This product includes cryptographic software written by
34 * Eric Young (eay@cryptsoft.com)"
35 * The word 'cryptographic' can be left out if the rouines from the library
36 * being used are not cryptographic related :-).
37 * 4. If you include any Windows specific code (or a derivative thereof) from
38 * the apps directory (application code) you must include an acknowledgement:
39 * "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
41 * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
42 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
43 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
44 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
45 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
46 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
47 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
48 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
49 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
50 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
53 * The licence and distribution terms for any publically available version or
54 * derivative of this code cannot be changed. i.e. this code cannot simply be
55 * copied and put under another distribution licence
56 * [including the GNU Public Licence.]
63 #if defined(BN_LLONG) || defined(BN_UMULT_HIGH)
65 BN_ULONG bn_mul_add_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
70 if (num <= 0) return(c1);
74 mul_add(rp[0],ap[0],w,c1);
75 mul_add(rp[1],ap[1],w,c1);
76 mul_add(rp[2],ap[2],w,c1);
77 mul_add(rp[3],ap[3],w,c1);
82 mul_add(rp[0],ap[0],w,c1); if (--num==0) return c1;
83 mul_add(rp[1],ap[1],w,c1); if (--num==0) return c1;
84 mul_add(rp[2],ap[2],w,c1); return c1;
90 BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
95 if (num <= 0) return(c1);
99 mul(rp[0],ap[0],w,c1);
100 mul(rp[1],ap[1],w,c1);
101 mul(rp[2],ap[2],w,c1);
102 mul(rp[3],ap[3],w,c1);
103 ap+=4; rp+=4; num-=4;
107 mul(rp[0],ap[0],w,c1); if (--num == 0) return c1;
108 mul(rp[1],ap[1],w,c1); if (--num == 0) return c1;
109 mul(rp[2],ap[2],w,c1);
114 void bn_sqr_words(BN_ULONG *r, BN_ULONG *a, int n)
128 sqr(r[0],r[1],a[0]); if (--n == 0) return;
129 sqr(r[2],r[3],a[1]); if (--n == 0) return;
136 BN_ULONG bn_mul_add_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
142 if (num <= 0) return((BN_ULONG)0);
149 mul_add(rp[0],ap[0],bl,bh,c);
150 if (--num == 0) break;
151 mul_add(rp[1],ap[1],bl,bh,c);
152 if (--num == 0) break;
153 mul_add(rp[2],ap[2],bl,bh,c);
154 if (--num == 0) break;
155 mul_add(rp[3],ap[3],bl,bh,c);
156 if (--num == 0) break;
163 BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
169 if (num <= 0) return((BN_ULONG)0);
176 mul(rp[0],ap[0],bl,bh,carry);
177 if (--num == 0) break;
178 mul(rp[1],ap[1],bl,bh,carry);
179 if (--num == 0) break;
180 mul(rp[2],ap[2],bl,bh,carry);
181 if (--num == 0) break;
182 mul(rp[3],ap[3],bl,bh,carry);
183 if (--num == 0) break;
190 void bn_sqr_words(BN_ULONG *r, BN_ULONG *a, int n)
196 sqr64(r[0],r[1],a[0]);
199 sqr64(r[2],r[3],a[1]);
202 sqr64(r[4],r[5],a[2]);
205 sqr64(r[6],r[7],a[3]);
215 #if defined(BN_LLONG) && defined(BN_DIV2W)
217 BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
219 return((BN_ULONG)(((((BN_ULLONG)h)<<BN_BITS2)|l)/(BN_ULLONG)d));
224 /* Divide h-l by d and return the result. */
225 /* I need to test this some more :-( */
226 BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
228 BN_ULONG dh,dl,q,ret=0,th,tl,t;
231 if (d == 0) return(BN_MASK2);
233 i=BN_num_bits_word(d);
234 if ((i != BN_BITS2) && (h > (BN_ULONG)1<<i))
236 #if !defined(NO_STDIO) && !defined(WIN16)
237 fprintf(stderr,"Division would overflow (%d)\n",i);
247 h=(h<<i)|(l>>(BN_BITS2-i));
250 dh=(d&BN_MASK2h)>>BN_BITS4;
254 if ((h>>BN_BITS4) == dh)
267 ((l&BN_MASK2h)>>BN_BITS4))))
274 tl=(tl<<BN_BITS4)&BN_MASK2h;
286 if (--count == 0) break;
289 h=((h<<BN_BITS4)|(l>>BN_BITS4))&BN_MASK2;
290 l=(l&BN_MASK2l)<<BN_BITS4;
298 BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
303 if (n <= 0) return((BN_ULONG)0);
307 ll+=(BN_ULLONG)a[0]+b[0];
308 r[0]=(BN_ULONG)ll&BN_MASK2;
312 ll+=(BN_ULLONG)a[1]+b[1];
313 r[1]=(BN_ULONG)ll&BN_MASK2;
317 ll+=(BN_ULLONG)a[2]+b[2];
318 r[2]=(BN_ULONG)ll&BN_MASK2;
322 ll+=(BN_ULLONG)a[3]+b[3];
323 r[3]=(BN_ULONG)ll&BN_MASK2;
331 return((BN_ULONG)ll);
334 BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
339 if (n <= 0) return((BN_ULONG)0);
384 BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
390 if (n <= 0) return((BN_ULONG)0);
395 r[0]=(t1-t2-c)&BN_MASK2;
396 if (t1 != t2) c=(t1 < t2);
400 r[1]=(t1-t2-c)&BN_MASK2;
401 if (t1 != t2) c=(t1 < t2);
405 r[2]=(t1-t2-c)&BN_MASK2;
406 if (t1 != t2) c=(t1 < t2);
410 r[3]=(t1-t2-c)&BN_MASK2;
411 if (t1 != t2) c=(t1 < t2);
429 #define mul_add_c(a,b,c0,c1,c2) \
431 t1=(BN_ULONG)Lw(t); \
432 t2=(BN_ULONG)Hw(t); \
433 c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
434 c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
436 #define mul_add_c2(a,b,c0,c1,c2) \
440 t1=(BN_ULONG)Lw(tt); \
441 t2=(BN_ULONG)Hw(tt); \
442 c0=(c0+t1)&BN_MASK2; \
443 if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
444 c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
446 #define sqr_add_c(a,i,c0,c1,c2) \
447 t=(BN_ULLONG)a[i]*a[i]; \
448 t1=(BN_ULONG)Lw(t); \
449 t2=(BN_ULONG)Hw(t); \
450 c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
451 c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
453 #define sqr_add_c2(a,i,j,c0,c1,c2) \
454 mul_add_c2((a)[i],(a)[j],c0,c1,c2)
456 #elif defined(BN_UMULT_HIGH)
458 #define mul_add_c(a,b,c0,c1,c2) { \
459 BN_ULONG ta=(a),tb=(b); \
461 t2 = BN_UMULT_HIGH(ta,tb); \
462 c0 += t1; t2 += (c0<t1)?1:0; \
463 c1 += t2; c2 += (c1<t2)?1:0; \
466 #define mul_add_c2(a,b,c0,c1,c2) { \
467 BN_ULONG ta=(a),tb=(b),t0; \
468 t1 = BN_UMULT_HIGH(ta,tb); \
470 t2 = t1+t1; c2 += (t2<t1)?1:0; \
471 t1 = t0+t0; t2 += (t1<t0)?1:0; \
472 c0 += t1; t2 += (c0<t1)?1:0; \
473 c1 += t2; c2 += (c1<t2)?1:0; \
476 #define sqr_add_c(a,i,c0,c1,c2) { \
477 BN_ULONG ta=(a)[i]; \
479 t2 = BN_UMULT_HIGH(ta,ta); \
480 c0 += t1; t2 += (c0<t1)?1:0; \
481 c1 += t2; c2 += (c1<t2)?1:0; \
484 #define sqr_add_c2(a,i,j,c0,c1,c2) \
485 mul_add_c2((a)[i],(a)[j],c0,c1,c2)
488 #define mul_add_c(a,b,c0,c1,c2) \
489 t1=LBITS(a); t2=HBITS(a); \
490 bl=LBITS(b); bh=HBITS(b); \
491 mul64(t1,t2,bl,bh); \
492 c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
493 c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
495 #define mul_add_c2(a,b,c0,c1,c2) \
496 t1=LBITS(a); t2=HBITS(a); \
497 bl=LBITS(b); bh=HBITS(b); \
498 mul64(t1,t2,bl,bh); \
499 if (t2 & BN_TBIT) c2++; \
500 t2=(t2+t2)&BN_MASK2; \
501 if (t1 & BN_TBIT) t2++; \
502 t1=(t1+t1)&BN_MASK2; \
503 c0=(c0+t1)&BN_MASK2; \
504 if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
505 c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
507 #define sqr_add_c(a,i,c0,c1,c2) \
508 sqr64(t1,t2,(a)[i]); \
509 c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
510 c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
512 #define sqr_add_c2(a,i,j,c0,c1,c2) \
513 mul_add_c2((a)[i],(a)[j],c0,c1,c2)
516 void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
529 mul_add_c(a[0],b[0],c1,c2,c3);
532 mul_add_c(a[0],b[1],c2,c3,c1);
533 mul_add_c(a[1],b[0],c2,c3,c1);
536 mul_add_c(a[2],b[0],c3,c1,c2);
537 mul_add_c(a[1],b[1],c3,c1,c2);
538 mul_add_c(a[0],b[2],c3,c1,c2);
541 mul_add_c(a[0],b[3],c1,c2,c3);
542 mul_add_c(a[1],b[2],c1,c2,c3);
543 mul_add_c(a[2],b[1],c1,c2,c3);
544 mul_add_c(a[3],b[0],c1,c2,c3);
547 mul_add_c(a[4],b[0],c2,c3,c1);
548 mul_add_c(a[3],b[1],c2,c3,c1);
549 mul_add_c(a[2],b[2],c2,c3,c1);
550 mul_add_c(a[1],b[3],c2,c3,c1);
551 mul_add_c(a[0],b[4],c2,c3,c1);
554 mul_add_c(a[0],b[5],c3,c1,c2);
555 mul_add_c(a[1],b[4],c3,c1,c2);
556 mul_add_c(a[2],b[3],c3,c1,c2);
557 mul_add_c(a[3],b[2],c3,c1,c2);
558 mul_add_c(a[4],b[1],c3,c1,c2);
559 mul_add_c(a[5],b[0],c3,c1,c2);
562 mul_add_c(a[6],b[0],c1,c2,c3);
563 mul_add_c(a[5],b[1],c1,c2,c3);
564 mul_add_c(a[4],b[2],c1,c2,c3);
565 mul_add_c(a[3],b[3],c1,c2,c3);
566 mul_add_c(a[2],b[4],c1,c2,c3);
567 mul_add_c(a[1],b[5],c1,c2,c3);
568 mul_add_c(a[0],b[6],c1,c2,c3);
571 mul_add_c(a[0],b[7],c2,c3,c1);
572 mul_add_c(a[1],b[6],c2,c3,c1);
573 mul_add_c(a[2],b[5],c2,c3,c1);
574 mul_add_c(a[3],b[4],c2,c3,c1);
575 mul_add_c(a[4],b[3],c2,c3,c1);
576 mul_add_c(a[5],b[2],c2,c3,c1);
577 mul_add_c(a[6],b[1],c2,c3,c1);
578 mul_add_c(a[7],b[0],c2,c3,c1);
581 mul_add_c(a[7],b[1],c3,c1,c2);
582 mul_add_c(a[6],b[2],c3,c1,c2);
583 mul_add_c(a[5],b[3],c3,c1,c2);
584 mul_add_c(a[4],b[4],c3,c1,c2);
585 mul_add_c(a[3],b[5],c3,c1,c2);
586 mul_add_c(a[2],b[6],c3,c1,c2);
587 mul_add_c(a[1],b[7],c3,c1,c2);
590 mul_add_c(a[2],b[7],c1,c2,c3);
591 mul_add_c(a[3],b[6],c1,c2,c3);
592 mul_add_c(a[4],b[5],c1,c2,c3);
593 mul_add_c(a[5],b[4],c1,c2,c3);
594 mul_add_c(a[6],b[3],c1,c2,c3);
595 mul_add_c(a[7],b[2],c1,c2,c3);
598 mul_add_c(a[7],b[3],c2,c3,c1);
599 mul_add_c(a[6],b[4],c2,c3,c1);
600 mul_add_c(a[5],b[5],c2,c3,c1);
601 mul_add_c(a[4],b[6],c2,c3,c1);
602 mul_add_c(a[3],b[7],c2,c3,c1);
605 mul_add_c(a[4],b[7],c3,c1,c2);
606 mul_add_c(a[5],b[6],c3,c1,c2);
607 mul_add_c(a[6],b[5],c3,c1,c2);
608 mul_add_c(a[7],b[4],c3,c1,c2);
611 mul_add_c(a[7],b[5],c1,c2,c3);
612 mul_add_c(a[6],b[6],c1,c2,c3);
613 mul_add_c(a[5],b[7],c1,c2,c3);
616 mul_add_c(a[6],b[7],c2,c3,c1);
617 mul_add_c(a[7],b[6],c2,c3,c1);
620 mul_add_c(a[7],b[7],c3,c1,c2);
625 void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
638 mul_add_c(a[0],b[0],c1,c2,c3);
641 mul_add_c(a[0],b[1],c2,c3,c1);
642 mul_add_c(a[1],b[0],c2,c3,c1);
645 mul_add_c(a[2],b[0],c3,c1,c2);
646 mul_add_c(a[1],b[1],c3,c1,c2);
647 mul_add_c(a[0],b[2],c3,c1,c2);
650 mul_add_c(a[0],b[3],c1,c2,c3);
651 mul_add_c(a[1],b[2],c1,c2,c3);
652 mul_add_c(a[2],b[1],c1,c2,c3);
653 mul_add_c(a[3],b[0],c1,c2,c3);
656 mul_add_c(a[3],b[1],c2,c3,c1);
657 mul_add_c(a[2],b[2],c2,c3,c1);
658 mul_add_c(a[1],b[3],c2,c3,c1);
661 mul_add_c(a[2],b[3],c3,c1,c2);
662 mul_add_c(a[3],b[2],c3,c1,c2);
665 mul_add_c(a[3],b[3],c1,c2,c3);
670 void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
683 sqr_add_c(a,0,c1,c2,c3);
686 sqr_add_c2(a,1,0,c2,c3,c1);
689 sqr_add_c(a,1,c3,c1,c2);
690 sqr_add_c2(a,2,0,c3,c1,c2);
693 sqr_add_c2(a,3,0,c1,c2,c3);
694 sqr_add_c2(a,2,1,c1,c2,c3);
697 sqr_add_c(a,2,c2,c3,c1);
698 sqr_add_c2(a,3,1,c2,c3,c1);
699 sqr_add_c2(a,4,0,c2,c3,c1);
702 sqr_add_c2(a,5,0,c3,c1,c2);
703 sqr_add_c2(a,4,1,c3,c1,c2);
704 sqr_add_c2(a,3,2,c3,c1,c2);
707 sqr_add_c(a,3,c1,c2,c3);
708 sqr_add_c2(a,4,2,c1,c2,c3);
709 sqr_add_c2(a,5,1,c1,c2,c3);
710 sqr_add_c2(a,6,0,c1,c2,c3);
713 sqr_add_c2(a,7,0,c2,c3,c1);
714 sqr_add_c2(a,6,1,c2,c3,c1);
715 sqr_add_c2(a,5,2,c2,c3,c1);
716 sqr_add_c2(a,4,3,c2,c3,c1);
719 sqr_add_c(a,4,c3,c1,c2);
720 sqr_add_c2(a,5,3,c3,c1,c2);
721 sqr_add_c2(a,6,2,c3,c1,c2);
722 sqr_add_c2(a,7,1,c3,c1,c2);
725 sqr_add_c2(a,7,2,c1,c2,c3);
726 sqr_add_c2(a,6,3,c1,c2,c3);
727 sqr_add_c2(a,5,4,c1,c2,c3);
730 sqr_add_c(a,5,c2,c3,c1);
731 sqr_add_c2(a,6,4,c2,c3,c1);
732 sqr_add_c2(a,7,3,c2,c3,c1);
735 sqr_add_c2(a,7,4,c3,c1,c2);
736 sqr_add_c2(a,6,5,c3,c1,c2);
739 sqr_add_c(a,6,c1,c2,c3);
740 sqr_add_c2(a,7,5,c1,c2,c3);
743 sqr_add_c2(a,7,6,c2,c3,c1);
746 sqr_add_c(a,7,c3,c1,c2);
751 void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
764 sqr_add_c(a,0,c1,c2,c3);
767 sqr_add_c2(a,1,0,c2,c3,c1);
770 sqr_add_c(a,1,c3,c1,c2);
771 sqr_add_c2(a,2,0,c3,c1,c2);
774 sqr_add_c2(a,3,0,c1,c2,c3);
775 sqr_add_c2(a,2,1,c1,c2,c3);
778 sqr_add_c(a,2,c2,c3,c1);
779 sqr_add_c2(a,3,1,c2,c3,c1);
782 sqr_add_c2(a,3,2,c3,c1,c2);
785 sqr_add_c(a,3,c1,c2,c3);
791 /* hmm... is it faster just to do a multiply? */
793 void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
796 bn_sqr_normal(r,a,4,t);
800 void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
803 bn_sqr_normal(r,a,8,t);
806 void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
808 r[4]=bn_mul_words( &(r[0]),a,4,b[0]);
809 r[5]=bn_mul_add_words(&(r[1]),a,4,b[1]);
810 r[6]=bn_mul_add_words(&(r[2]),a,4,b[2]);
811 r[7]=bn_mul_add_words(&(r[3]),a,4,b[3]);
814 void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
816 r[ 8]=bn_mul_words( &(r[0]),a,8,b[0]);
817 r[ 9]=bn_mul_add_words(&(r[1]),a,8,b[1]);
818 r[10]=bn_mul_add_words(&(r[2]),a,8,b[2]);
819 r[11]=bn_mul_add_words(&(r[3]),a,8,b[3]);
820 r[12]=bn_mul_add_words(&(r[4]),a,8,b[4]);
821 r[13]=bn_mul_add_words(&(r[5]),a,8,b[5]);
822 r[14]=bn_mul_add_words(&(r[6]),a,8,b[6]);
823 r[15]=bn_mul_add_words(&(r[7]),a,8,b[7]);
826 #endif /* BN_COMBA */