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.]
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 if (--num == 0) break;
76 mul_add(rp[1],ap[1],w,c1);
77 if (--num == 0) break;
78 mul_add(rp[2],ap[2],w,c1);
79 if (--num == 0) break;
80 mul_add(rp[3],ap[3],w,c1);
81 if (--num == 0) break;
89 BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
94 if (num <= 0) return(c1);
98 mul(rp[0],ap[0],w,c1);
99 if (--num == 0) break;
100 mul(rp[1],ap[1],w,c1);
101 if (--num == 0) break;
102 mul(rp[2],ap[2],w,c1);
103 if (--num == 0) break;
104 mul(rp[3],ap[3],w,c1);
105 if (--num == 0) break;
112 void bn_sqr_words(BN_ULONG *r, BN_ULONG *a, int n)
120 t=(BN_ULLONG)(a[0])*(a[0]);
121 r[0]=Lw(t); r[1]=Hw(t);
124 t=(BN_ULLONG)(a[1])*(a[1]);
125 r[2]=Lw(t); r[3]=Hw(t);
128 t=(BN_ULLONG)(a[2])*(a[2]);
129 r[4]=Lw(t); r[5]=Hw(t);
132 t=(BN_ULLONG)(a[3])*(a[3]);
133 r[6]=Lw(t); r[7]=Hw(t);
143 BN_ULONG bn_mul_add_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
149 if (num <= 0) return((BN_ULONG)0);
156 mul_add(rp[0],ap[0],bl,bh,c);
157 if (--num == 0) break;
158 mul_add(rp[1],ap[1],bl,bh,c);
159 if (--num == 0) break;
160 mul_add(rp[2],ap[2],bl,bh,c);
161 if (--num == 0) break;
162 mul_add(rp[3],ap[3],bl,bh,c);
163 if (--num == 0) break;
170 BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
176 if (num <= 0) return((BN_ULONG)0);
183 mul(rp[0],ap[0],bl,bh,carry);
184 if (--num == 0) break;
185 mul(rp[1],ap[1],bl,bh,carry);
186 if (--num == 0) break;
187 mul(rp[2],ap[2],bl,bh,carry);
188 if (--num == 0) break;
189 mul(rp[3],ap[3],bl,bh,carry);
190 if (--num == 0) break;
197 void bn_sqr_words(BN_ULONG *r, BN_ULONG *a, int n)
203 sqr64(r[0],r[1],a[0]);
206 sqr64(r[2],r[3],a[1]);
209 sqr64(r[4],r[5],a[2]);
212 sqr64(r[6],r[7],a[3]);
222 #if defined(BN_LLONG) && defined(BN_DIV2W)
224 BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
226 return((BN_ULONG)(((((BN_ULLONG)h)<<BN_BITS2)|l)/(BN_ULLONG)d));
231 /* Divide h-l by d and return the result. */
232 /* I need to test this some more :-( */
233 BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
235 BN_ULONG dh,dl,q,ret=0,th,tl,t;
238 if (d == 0) return(BN_MASK2);
240 i=BN_num_bits_word(d);
241 if ((i != BN_BITS2) && (h > (BN_ULONG)1<<i))
243 #if !defined(NO_STDIO) && !defined(WIN16)
244 fprintf(stderr,"Division would overflow (%d)\n",i);
254 h=(h<<i)|(l>>(BN_BITS2-i));
257 dh=(d&BN_MASK2h)>>BN_BITS4;
261 if ((h>>BN_BITS4) == dh)
272 ((l&BN_MASK2h)>>BN_BITS4))))
279 tl=(tl<<BN_BITS4)&BN_MASK2h;
291 if (--count == 0) break;
294 h=((h<<BN_BITS4)|(l>>BN_BITS4))&BN_MASK2;
295 l=(l&BN_MASK2l)<<BN_BITS4;
303 BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
308 if (n <= 0) return((BN_ULONG)0);
312 ll+=(BN_ULLONG)a[0]+b[0];
313 r[0]=(BN_ULONG)ll&BN_MASK2;
317 ll+=(BN_ULLONG)a[1]+b[1];
318 r[1]=(BN_ULONG)ll&BN_MASK2;
322 ll+=(BN_ULLONG)a[2]+b[2];
323 r[2]=(BN_ULONG)ll&BN_MASK2;
327 ll+=(BN_ULLONG)a[3]+b[3];
328 r[3]=(BN_ULONG)ll&BN_MASK2;
336 return((BN_ULONG)ll);
339 BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
344 if (n <= 0) return((BN_ULONG)0);
389 BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
395 if (n <= 0) return((BN_ULONG)0);
400 r[0]=(t1-t2-c)&BN_MASK2;
401 if (t1 != t2) c=(t1 < t2);
405 r[1]=(t1-t2-c)&BN_MASK2;
406 if (t1 != t2) c=(t1 < t2);
410 r[2]=(t1-t2-c)&BN_MASK2;
411 if (t1 != t2) c=(t1 < t2);
415 r[3]=(t1-t2-c)&BN_MASK2;
416 if (t1 != t2) c=(t1 < t2);
434 #define mul_add_c(a,b,c0,c1,c2) \
436 t1=(BN_ULONG)Lw(t); \
437 t2=(BN_ULONG)Hw(t); \
438 c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
439 c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
441 #define mul_add_c2(a,b,c0,c1,c2) \
445 t1=(BN_ULONG)Lw(tt); \
446 t2=(BN_ULONG)Hw(tt); \
447 c0=(c0+t1)&BN_MASK2; \
448 if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
449 c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
451 #define sqr_add_c(a,i,c0,c1,c2) \
452 t=(BN_ULLONG)a[i]*a[i]; \
453 t1=(BN_ULONG)Lw(t); \
454 t2=(BN_ULONG)Hw(t); \
455 c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
456 c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
458 #define sqr_add_c2(a,i,j,c0,c1,c2) \
459 mul_add_c2((a)[i],(a)[j],c0,c1,c2)
461 #define mul_add_c(a,b,c0,c1,c2) \
462 t1=LBITS(a); t2=HBITS(a); \
463 bl=LBITS(b); bh=HBITS(b); \
464 mul64(t1,t2,bl,bh); \
465 c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
466 c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
468 #define mul_add_c2(a,b,c0,c1,c2) \
469 t1=LBITS(a); t2=HBITS(a); \
470 bl=LBITS(b); bh=HBITS(b); \
471 mul64(t1,t2,bl,bh); \
472 if (t2 & BN_TBIT) c2++; \
473 t2=(t2+t2)&BN_MASK2; \
474 if (t1 & BN_TBIT) t2++; \
475 t1=(t1+t1)&BN_MASK2; \
476 c0=(c0+t1)&BN_MASK2; \
477 if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
478 c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
480 #define sqr_add_c(a,i,c0,c1,c2) \
481 sqr64(t1,t2,(a)[i]); \
482 c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
483 c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
485 #define sqr_add_c2(a,i,j,c0,c1,c2) \
486 mul_add_c2((a)[i],(a)[j],c0,c1,c2)
489 void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
502 mul_add_c(a[0],b[0],c1,c2,c3);
505 mul_add_c(a[0],b[1],c2,c3,c1);
506 mul_add_c(a[1],b[0],c2,c3,c1);
509 mul_add_c(a[2],b[0],c3,c1,c2);
510 mul_add_c(a[1],b[1],c3,c1,c2);
511 mul_add_c(a[0],b[2],c3,c1,c2);
514 mul_add_c(a[0],b[3],c1,c2,c3);
515 mul_add_c(a[1],b[2],c1,c2,c3);
516 mul_add_c(a[2],b[1],c1,c2,c3);
517 mul_add_c(a[3],b[0],c1,c2,c3);
520 mul_add_c(a[4],b[0],c2,c3,c1);
521 mul_add_c(a[3],b[1],c2,c3,c1);
522 mul_add_c(a[2],b[2],c2,c3,c1);
523 mul_add_c(a[1],b[3],c2,c3,c1);
524 mul_add_c(a[0],b[4],c2,c3,c1);
527 mul_add_c(a[0],b[5],c3,c1,c2);
528 mul_add_c(a[1],b[4],c3,c1,c2);
529 mul_add_c(a[2],b[3],c3,c1,c2);
530 mul_add_c(a[3],b[2],c3,c1,c2);
531 mul_add_c(a[4],b[1],c3,c1,c2);
532 mul_add_c(a[5],b[0],c3,c1,c2);
535 mul_add_c(a[6],b[0],c1,c2,c3);
536 mul_add_c(a[5],b[1],c1,c2,c3);
537 mul_add_c(a[4],b[2],c1,c2,c3);
538 mul_add_c(a[3],b[3],c1,c2,c3);
539 mul_add_c(a[2],b[4],c1,c2,c3);
540 mul_add_c(a[1],b[5],c1,c2,c3);
541 mul_add_c(a[0],b[6],c1,c2,c3);
544 mul_add_c(a[0],b[7],c2,c3,c1);
545 mul_add_c(a[1],b[6],c2,c3,c1);
546 mul_add_c(a[2],b[5],c2,c3,c1);
547 mul_add_c(a[3],b[4],c2,c3,c1);
548 mul_add_c(a[4],b[3],c2,c3,c1);
549 mul_add_c(a[5],b[2],c2,c3,c1);
550 mul_add_c(a[6],b[1],c2,c3,c1);
551 mul_add_c(a[7],b[0],c2,c3,c1);
554 mul_add_c(a[7],b[1],c3,c1,c2);
555 mul_add_c(a[6],b[2],c3,c1,c2);
556 mul_add_c(a[5],b[3],c3,c1,c2);
557 mul_add_c(a[4],b[4],c3,c1,c2);
558 mul_add_c(a[3],b[5],c3,c1,c2);
559 mul_add_c(a[2],b[6],c3,c1,c2);
560 mul_add_c(a[1],b[7],c3,c1,c2);
563 mul_add_c(a[2],b[7],c1,c2,c3);
564 mul_add_c(a[3],b[6],c1,c2,c3);
565 mul_add_c(a[4],b[5],c1,c2,c3);
566 mul_add_c(a[5],b[4],c1,c2,c3);
567 mul_add_c(a[6],b[3],c1,c2,c3);
568 mul_add_c(a[7],b[2],c1,c2,c3);
571 mul_add_c(a[7],b[3],c2,c3,c1);
572 mul_add_c(a[6],b[4],c2,c3,c1);
573 mul_add_c(a[5],b[5],c2,c3,c1);
574 mul_add_c(a[4],b[6],c2,c3,c1);
575 mul_add_c(a[3],b[7],c2,c3,c1);
578 mul_add_c(a[4],b[7],c3,c1,c2);
579 mul_add_c(a[5],b[6],c3,c1,c2);
580 mul_add_c(a[6],b[5],c3,c1,c2);
581 mul_add_c(a[7],b[4],c3,c1,c2);
584 mul_add_c(a[7],b[5],c1,c2,c3);
585 mul_add_c(a[6],b[6],c1,c2,c3);
586 mul_add_c(a[5],b[7],c1,c2,c3);
589 mul_add_c(a[6],b[7],c2,c3,c1);
590 mul_add_c(a[7],b[6],c2,c3,c1);
593 mul_add_c(a[7],b[7],c3,c1,c2);
598 void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
611 mul_add_c(a[0],b[0],c1,c2,c3);
614 mul_add_c(a[0],b[1],c2,c3,c1);
615 mul_add_c(a[1],b[0],c2,c3,c1);
618 mul_add_c(a[2],b[0],c3,c1,c2);
619 mul_add_c(a[1],b[1],c3,c1,c2);
620 mul_add_c(a[0],b[2],c3,c1,c2);
623 mul_add_c(a[0],b[3],c1,c2,c3);
624 mul_add_c(a[1],b[2],c1,c2,c3);
625 mul_add_c(a[2],b[1],c1,c2,c3);
626 mul_add_c(a[3],b[0],c1,c2,c3);
629 mul_add_c(a[3],b[1],c2,c3,c1);
630 mul_add_c(a[2],b[2],c2,c3,c1);
631 mul_add_c(a[1],b[3],c2,c3,c1);
634 mul_add_c(a[2],b[3],c3,c1,c2);
635 mul_add_c(a[3],b[2],c3,c1,c2);
638 mul_add_c(a[3],b[3],c1,c2,c3);
643 void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
656 sqr_add_c(a,0,c1,c2,c3);
659 sqr_add_c2(a,1,0,c2,c3,c1);
662 sqr_add_c(a,1,c3,c1,c2);
663 sqr_add_c2(a,2,0,c3,c1,c2);
666 sqr_add_c2(a,3,0,c1,c2,c3);
667 sqr_add_c2(a,2,1,c1,c2,c3);
670 sqr_add_c(a,2,c2,c3,c1);
671 sqr_add_c2(a,3,1,c2,c3,c1);
672 sqr_add_c2(a,4,0,c2,c3,c1);
675 sqr_add_c2(a,5,0,c3,c1,c2);
676 sqr_add_c2(a,4,1,c3,c1,c2);
677 sqr_add_c2(a,3,2,c3,c1,c2);
680 sqr_add_c(a,3,c1,c2,c3);
681 sqr_add_c2(a,4,2,c1,c2,c3);
682 sqr_add_c2(a,5,1,c1,c2,c3);
683 sqr_add_c2(a,6,0,c1,c2,c3);
686 sqr_add_c2(a,7,0,c2,c3,c1);
687 sqr_add_c2(a,6,1,c2,c3,c1);
688 sqr_add_c2(a,5,2,c2,c3,c1);
689 sqr_add_c2(a,4,3,c2,c3,c1);
692 sqr_add_c(a,4,c3,c1,c2);
693 sqr_add_c2(a,5,3,c3,c1,c2);
694 sqr_add_c2(a,6,2,c3,c1,c2);
695 sqr_add_c2(a,7,1,c3,c1,c2);
698 sqr_add_c2(a,7,2,c1,c2,c3);
699 sqr_add_c2(a,6,3,c1,c2,c3);
700 sqr_add_c2(a,5,4,c1,c2,c3);
703 sqr_add_c(a,5,c2,c3,c1);
704 sqr_add_c2(a,6,4,c2,c3,c1);
705 sqr_add_c2(a,7,3,c2,c3,c1);
708 sqr_add_c2(a,7,4,c3,c1,c2);
709 sqr_add_c2(a,6,5,c3,c1,c2);
712 sqr_add_c(a,6,c1,c2,c3);
713 sqr_add_c2(a,7,5,c1,c2,c3);
716 sqr_add_c2(a,7,6,c2,c3,c1);
719 sqr_add_c(a,7,c3,c1,c2);
724 void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
737 sqr_add_c(a,0,c1,c2,c3);
740 sqr_add_c2(a,1,0,c2,c3,c1);
743 sqr_add_c(a,1,c3,c1,c2);
744 sqr_add_c2(a,2,0,c3,c1,c2);
747 sqr_add_c2(a,3,0,c1,c2,c3);
748 sqr_add_c2(a,2,1,c1,c2,c3);
751 sqr_add_c(a,2,c2,c3,c1);
752 sqr_add_c2(a,3,1,c2,c3,c1);
755 sqr_add_c2(a,3,2,c3,c1,c2);
758 sqr_add_c(a,3,c1,c2,c3);
764 /* hmm... is it faster just to do a multiply? */
766 void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
769 bn_sqr_normal(r,a,4,t);
773 void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
776 bn_sqr_normal(r,a,8,t);
779 void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
781 r[4]=bn_mul_words( &(r[0]),a,4,b[0]);
782 r[5]=bn_mul_add_words(&(r[1]),a,4,b[1]);
783 r[6]=bn_mul_add_words(&(r[2]),a,4,b[2]);
784 r[7]=bn_mul_add_words(&(r[3]),a,4,b[3]);
787 void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
789 r[ 8]=bn_mul_words( &(r[0]),a,8,b[0]);
790 r[ 9]=bn_mul_add_words(&(r[1]),a,8,b[1]);
791 r[10]=bn_mul_add_words(&(r[2]),a,8,b[2]);
792 r[11]=bn_mul_add_words(&(r[3]),a,8,b[3]);
793 r[12]=bn_mul_add_words(&(r[4]),a,8,b[4]);
794 r[13]=bn_mul_add_words(&(r[5]),a,8,b[5]);
795 r[14]=bn_mul_add_words(&(r[6]),a,8,b[6]);
796 r[15]=bn_mul_add_words(&(r[7]),a,8,b[7]);
799 #endif /* BN_COMBA */