This commit was generated by cvs2svn to track changes on a CVS vendor
[openssl.git] / crypto / bn / bn_asm.c
1 /* crypto/bn/bn_asm.c */
2 /* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
3  * All rights reserved.
4  *
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.
8  * 
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).
15  * 
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.
22  * 
23  * Redistribution and use in source and binary forms, with or without
24  * modification, are permitted provided that the following conditions
25  * are met:
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)"
40  * 
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
51  * SUCH DAMAGE.
52  * 
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.]
57  */
58
59 #include <stdio.h>
60 #include "cryptlib.h"
61 #include "bn_lcl.h"
62
63 #ifdef BN_LLONG 
64
65 BN_ULONG bn_mul_add_words(rp,ap,num,w)
66 BN_ULONG *rp,*ap;
67 int num;
68 BN_ULONG w;
69         {
70         BN_ULONG c1=0;
71
72         bn_check_num(num);
73         if (num <= 0) return(c1);
74
75         for (;;)
76                 {
77                 mul_add(rp[0],ap[0],w,c1);
78                 if (--num == 0) break;
79                 mul_add(rp[1],ap[1],w,c1);
80                 if (--num == 0) break;
81                 mul_add(rp[2],ap[2],w,c1);
82                 if (--num == 0) break;
83                 mul_add(rp[3],ap[3],w,c1);
84                 if (--num == 0) break;
85                 ap+=4;
86                 rp+=4;
87                 }
88         
89         return(c1);
90         } 
91
92 BN_ULONG bn_mul_words(rp,ap,num,w)
93 BN_ULONG *rp,*ap;
94 int num;
95 BN_ULONG w;
96         {
97         BN_ULONG c1=0;
98
99         bn_check_num(num);
100         if (num <= 0) return(c1);
101
102         for (;;)
103                 {
104                 mul(rp[0],ap[0],w,c1);
105                 if (--num == 0) break;
106                 mul(rp[1],ap[1],w,c1);
107                 if (--num == 0) break;
108                 mul(rp[2],ap[2],w,c1);
109                 if (--num == 0) break;
110                 mul(rp[3],ap[3],w,c1);
111                 if (--num == 0) break;
112                 ap+=4;
113                 rp+=4;
114                 }
115         return(c1);
116         } 
117
118 void bn_sqr_words(r,a,n)
119 BN_ULONG *r,*a;
120 int n;
121         {
122         bn_check_num(n);
123         if (n <= 0) return;
124         for (;;)
125                 {
126                 BN_ULLONG t;
127
128                 t=(BN_ULLONG)(a[0])*(a[0]);
129                 r[0]=Lw(t); r[1]=Hw(t);
130                 if (--n == 0) break;
131
132                 t=(BN_ULLONG)(a[1])*(a[1]);
133                 r[2]=Lw(t); r[3]=Hw(t);
134                 if (--n == 0) break;
135
136                 t=(BN_ULLONG)(a[2])*(a[2]);
137                 r[4]=Lw(t); r[5]=Hw(t);
138                 if (--n == 0) break;
139
140                 t=(BN_ULLONG)(a[3])*(a[3]);
141                 r[6]=Lw(t); r[7]=Hw(t);
142                 if (--n == 0) break;
143
144                 a+=4;
145                 r+=8;
146                 }
147         }
148
149 #else
150
151 BN_ULONG bn_mul_add_words(rp,ap,num,w)
152 BN_ULONG *rp,*ap;
153 int num;
154 BN_ULONG w;
155         {
156         BN_ULONG c=0;
157         BN_ULONG bl,bh;
158
159         bn_check_num(num);
160         if (num <= 0) return((BN_ULONG)0);
161
162         bl=LBITS(w);
163         bh=HBITS(w);
164
165         for (;;)
166                 {
167                 mul_add(rp[0],ap[0],bl,bh,c);
168                 if (--num == 0) break;
169                 mul_add(rp[1],ap[1],bl,bh,c);
170                 if (--num == 0) break;
171                 mul_add(rp[2],ap[2],bl,bh,c);
172                 if (--num == 0) break;
173                 mul_add(rp[3],ap[3],bl,bh,c);
174                 if (--num == 0) break;
175                 ap+=4;
176                 rp+=4;
177                 }
178         return(c);
179         } 
180
181 BN_ULONG bn_mul_words(rp,ap,num,w)
182 BN_ULONG *rp,*ap;
183 int num;
184 BN_ULONG w;
185         {
186         BN_ULONG carry=0;
187         BN_ULONG bl,bh;
188
189         bn_check_num(num);
190         if (num <= 0) return((BN_ULONG)0);
191
192         bl=LBITS(w);
193         bh=HBITS(w);
194
195         for (;;)
196                 {
197                 mul(rp[0],ap[0],bl,bh,carry);
198                 if (--num == 0) break;
199                 mul(rp[1],ap[1],bl,bh,carry);
200                 if (--num == 0) break;
201                 mul(rp[2],ap[2],bl,bh,carry);
202                 if (--num == 0) break;
203                 mul(rp[3],ap[3],bl,bh,carry);
204                 if (--num == 0) break;
205                 ap+=4;
206                 rp+=4;
207                 }
208         return(carry);
209         } 
210
211 void bn_sqr_words(r,a,n)
212 BN_ULONG *r,*a;
213 int n;
214         {
215         bn_check_num(n);
216         if (n <= 0) return;
217         for (;;)
218                 {
219                 sqr64(r[0],r[1],a[0]);
220                 if (--n == 0) break;
221
222                 sqr64(r[2],r[3],a[1]);
223                 if (--n == 0) break;
224
225                 sqr64(r[4],r[5],a[2]);
226                 if (--n == 0) break;
227
228                 sqr64(r[6],r[7],a[3]);
229                 if (--n == 0) break;
230
231                 a+=4;
232                 r+=8;
233                 }
234         }
235
236 #endif
237
238 #if defined(BN_LLONG) && defined(BN_DIV2W)
239
240 BN_ULONG bn_div_words(h,l,d)
241 BN_ULONG h,l,d;
242         {
243         return((BN_ULONG)(((((BN_ULLONG)h)<<BN_BITS2)|l)/(BN_ULLONG)d));
244         }
245
246 #else
247
248 /* Divide h-l by d and return the result. */
249 /* I need to test this some more :-( */
250 BN_ULONG bn_div_words(h,l,d)
251 BN_ULONG h,l,d;
252         {
253         BN_ULONG dh,dl,q,ret=0,th,tl,t;
254         int i,count=2;
255
256         if (d == 0) return(BN_MASK2);
257
258         i=BN_num_bits_word(d);
259         if ((i != BN_BITS2) && (h > (BN_ULONG)1<<i))
260                 {
261 #if !defined(NO_STDIO) && !defined(WIN16)
262                 fprintf(stderr,"Division would overflow (%d)\n",i);
263 #endif
264                 abort();
265                 }
266         i=BN_BITS2-i;
267         if (h >= d) h-=d;
268
269         if (i)
270                 {
271                 d<<=i;
272                 h=(h<<i)|(l>>(BN_BITS2-i));
273                 l<<=i;
274                 }
275         dh=(d&BN_MASK2h)>>BN_BITS4;
276         dl=(d&BN_MASK2l);
277         for (;;)
278                 {
279                 if ((h>>BN_BITS4) == dh)
280                         q=BN_MASK2l;
281                 else
282                         q=h/dh;
283
284                 for (;;)
285                         {
286                         t=(h-q*dh);
287                         if ((t&BN_MASK2h) ||
288                                 ((dl*q) <= (
289                                         (t<<BN_BITS4)+
290                                         ((l&BN_MASK2h)>>BN_BITS4))))
291                                 break;
292                         q--;
293                         }
294                 th=q*dh;
295                 tl=q*dl;
296                 t=(tl>>BN_BITS4);
297                 tl=(tl<<BN_BITS4)&BN_MASK2h;
298                 th+=t;
299
300                 if (l < tl) th++;
301                 l-=tl;
302                 if (h < th)
303                         {
304                         h+=d;
305                         q--;
306                         }
307                 h-=th;
308
309                 if (--count == 0) break;
310
311                 ret=q<<BN_BITS4;
312                 h=((h<<BN_BITS4)|(l>>BN_BITS4))&BN_MASK2;
313                 l=(l&BN_MASK2l)<<BN_BITS4;
314                 }
315         ret|=q;
316         return(ret);
317         }
318 #endif
319
320 #ifdef BN_LLONG
321 BN_ULONG bn_add_words(r,a,b,n)
322 BN_ULONG *r,*a,*b;
323 int n;
324         {
325         BN_ULLONG ll=0;
326
327         bn_check_num(n);
328         if (n <= 0) return((BN_ULONG)0);
329
330         for (;;)
331                 {
332                 ll+=(BN_ULLONG)a[0]+b[0];
333                 r[0]=(BN_ULONG)ll&BN_MASK2;
334                 ll>>=BN_BITS2;
335                 if (--n <= 0) break;
336
337                 ll+=(BN_ULLONG)a[1]+b[1];
338                 r[1]=(BN_ULONG)ll&BN_MASK2;
339                 ll>>=BN_BITS2;
340                 if (--n <= 0) break;
341
342                 ll+=(BN_ULLONG)a[2]+b[2];
343                 r[2]=(BN_ULONG)ll&BN_MASK2;
344                 ll>>=BN_BITS2;
345                 if (--n <= 0) break;
346
347                 ll+=(BN_ULLONG)a[3]+b[3];
348                 r[3]=(BN_ULONG)ll&BN_MASK2;
349                 ll>>=BN_BITS2;
350                 if (--n <= 0) break;
351
352                 a+=4;
353                 b+=4;
354                 r+=4;
355                 }
356         return((BN_ULONG)ll);
357         }
358 #else
359 BN_ULONG bn_add_words(r,a,b,n)
360 BN_ULONG *r,*a,*b;
361 int n;
362         {
363         BN_ULONG c,l,t;
364
365         bn_check_num(n);
366         if (n <= 0) return((BN_ULONG)0);
367
368         c=0;
369         for (;;)
370                 {
371                 t=a[0];
372                 t=(t+c)&BN_MASK2;
373                 c=(t < c);
374                 l=(t+b[0])&BN_MASK2;
375                 c+=(l < t);
376                 r[0]=l;
377                 if (--n <= 0) break;
378
379                 t=a[1];
380                 t=(t+c)&BN_MASK2;
381                 c=(t < c);
382                 l=(t+b[1])&BN_MASK2;
383                 c+=(l < t);
384                 r[1]=l;
385                 if (--n <= 0) break;
386
387                 t=a[2];
388                 t=(t+c)&BN_MASK2;
389                 c=(t < c);
390                 l=(t+b[2])&BN_MASK2;
391                 c+=(l < t);
392                 r[2]=l;
393                 if (--n <= 0) break;
394
395                 t=a[3];
396                 t=(t+c)&BN_MASK2;
397                 c=(t < c);
398                 l=(t+b[3])&BN_MASK2;
399                 c+=(l < t);
400                 r[3]=l;
401                 if (--n <= 0) break;
402
403                 a+=4;
404                 b+=4;
405                 r+=4;
406                 }
407         return((BN_ULONG)c);
408         }
409 #endif
410
411 BN_ULONG bn_sub_words(r,a,b,n)
412 BN_ULONG *r,*a,*b;
413 int n;
414         {
415         BN_ULONG t1,t2;
416         int c=0;
417
418         bn_check_num(n);
419         if (n <= 0) return((BN_ULONG)0);
420
421         for (;;)
422                 {
423                 t1=a[0]; t2=b[0];
424                 r[0]=(t1-t2-c)&BN_MASK2;
425                 if (t1 != t2) c=(t1 < t2);
426                 if (--n <= 0) break;
427
428                 t1=a[1]; t2=b[1];
429                 r[1]=(t1-t2-c)&BN_MASK2;
430                 if (t1 != t2) c=(t1 < t2);
431                 if (--n <= 0) break;
432
433                 t1=a[2]; t2=b[2];
434                 r[2]=(t1-t2-c)&BN_MASK2;
435                 if (t1 != t2) c=(t1 < t2);
436                 if (--n <= 0) break;
437
438                 t1=a[3]; t2=b[3];
439                 r[3]=(t1-t2-c)&BN_MASK2;
440                 if (t1 != t2) c=(t1 < t2);
441                 if (--n <= 0) break;
442
443                 a+=4;
444                 b+=4;
445                 r+=4;
446                 }
447         return(c);
448         }
449
450 #ifdef BN_COMBA
451
452 #undef bn_mul_comba8
453 #undef bn_mul_comba4
454 #undef bn_sqr_comba8
455 #undef bn_sqr_comba4
456
457 #ifdef BN_LLONG
458 #define mul_add_c(a,b,c0,c1,c2) \
459         t=(BN_ULLONG)a*b; \
460         t1=(BN_ULONG)Lw(t); \
461         t2=(BN_ULONG)Hw(t); \
462         c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
463         c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
464
465 #define mul_add_c2(a,b,c0,c1,c2) \
466         t=(BN_ULLONG)a*b; \
467         tt=(t+t)&BN_MASK; \
468         if (tt < t) c2++; \
469         t1=(BN_ULONG)Lw(tt); \
470         t2=(BN_ULONG)Hw(tt); \
471         c0=(c0+t1)&BN_MASK2;  \
472         if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
473         c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
474
475 #define sqr_add_c(a,i,c0,c1,c2) \
476         t=(BN_ULLONG)a[i]*a[i]; \
477         t1=(BN_ULONG)Lw(t); \
478         t2=(BN_ULONG)Hw(t); \
479         c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
480         c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
481
482 #define sqr_add_c2(a,i,j,c0,c1,c2) \
483         mul_add_c2((a)[i],(a)[j],c0,c1,c2)
484 #else
485 #define mul_add_c(a,b,c0,c1,c2) \
486         t1=LBITS(a); t2=HBITS(a); \
487         bl=LBITS(b); bh=HBITS(b); \
488         mul64(t1,t2,bl,bh); \
489         c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
490         c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
491
492 #define mul_add_c2(a,b,c0,c1,c2) \
493         t1=LBITS(a); t2=HBITS(a); \
494         bl=LBITS(b); bh=HBITS(b); \
495         mul64(t1,t2,bl,bh); \
496         if (t2 & BN_TBIT) c2++; \
497         t2=(t2+t2)&BN_MASK2; \
498         if (t1 & BN_TBIT) t2++; \
499         t1=(t1+t1)&BN_MASK2; \
500         c0=(c0+t1)&BN_MASK2;  \
501         if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
502         c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
503
504 #define sqr_add_c(a,i,c0,c1,c2) \
505         sqr64(t1,t2,(a)[i]); \
506         c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
507         c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
508
509 #define sqr_add_c2(a,i,j,c0,c1,c2) \
510         mul_add_c2((a)[i],(a)[j],c0,c1,c2)
511 #endif
512
513 void bn_mul_comba8(r,a,b)
514 BN_ULONG *r,*a,*b;
515         {
516 #ifdef BN_LLONG
517         BN_ULLONG t;
518 #else
519         BN_ULONG bl,bh;
520 #endif
521         BN_ULONG t1,t2;
522         BN_ULONG c1,c2,c3;
523
524         c1=0;
525         c2=0;
526         c3=0;
527         mul_add_c(a[0],b[0],c1,c2,c3);
528         r[0]=c1;
529         c1=0;
530         mul_add_c(a[0],b[1],c2,c3,c1);
531         mul_add_c(a[1],b[0],c2,c3,c1);
532         r[1]=c2;
533         c2=0;
534         mul_add_c(a[2],b[0],c3,c1,c2);
535         mul_add_c(a[1],b[1],c3,c1,c2);
536         mul_add_c(a[0],b[2],c3,c1,c2);
537         r[2]=c3;
538         c3=0;
539         mul_add_c(a[0],b[3],c1,c2,c3);
540         mul_add_c(a[1],b[2],c1,c2,c3);
541         mul_add_c(a[2],b[1],c1,c2,c3);
542         mul_add_c(a[3],b[0],c1,c2,c3);
543         r[3]=c1;
544         c1=0;
545         mul_add_c(a[4],b[0],c2,c3,c1);
546         mul_add_c(a[3],b[1],c2,c3,c1);
547         mul_add_c(a[2],b[2],c2,c3,c1);
548         mul_add_c(a[1],b[3],c2,c3,c1);
549         mul_add_c(a[0],b[4],c2,c3,c1);
550         r[4]=c2;
551         c2=0;
552         mul_add_c(a[0],b[5],c3,c1,c2);
553         mul_add_c(a[1],b[4],c3,c1,c2);
554         mul_add_c(a[2],b[3],c3,c1,c2);
555         mul_add_c(a[3],b[2],c3,c1,c2);
556         mul_add_c(a[4],b[1],c3,c1,c2);
557         mul_add_c(a[5],b[0],c3,c1,c2);
558         r[5]=c3;
559         c3=0;
560         mul_add_c(a[6],b[0],c1,c2,c3);
561         mul_add_c(a[5],b[1],c1,c2,c3);
562         mul_add_c(a[4],b[2],c1,c2,c3);
563         mul_add_c(a[3],b[3],c1,c2,c3);
564         mul_add_c(a[2],b[4],c1,c2,c3);
565         mul_add_c(a[1],b[5],c1,c2,c3);
566         mul_add_c(a[0],b[6],c1,c2,c3);
567         r[6]=c1;
568         c1=0;
569         mul_add_c(a[0],b[7],c2,c3,c1);
570         mul_add_c(a[1],b[6],c2,c3,c1);
571         mul_add_c(a[2],b[5],c2,c3,c1);
572         mul_add_c(a[3],b[4],c2,c3,c1);
573         mul_add_c(a[4],b[3],c2,c3,c1);
574         mul_add_c(a[5],b[2],c2,c3,c1);
575         mul_add_c(a[6],b[1],c2,c3,c1);
576         mul_add_c(a[7],b[0],c2,c3,c1);
577         r[7]=c2;
578         c2=0;
579         mul_add_c(a[7],b[1],c3,c1,c2);
580         mul_add_c(a[6],b[2],c3,c1,c2);
581         mul_add_c(a[5],b[3],c3,c1,c2);
582         mul_add_c(a[4],b[4],c3,c1,c2);
583         mul_add_c(a[3],b[5],c3,c1,c2);
584         mul_add_c(a[2],b[6],c3,c1,c2);
585         mul_add_c(a[1],b[7],c3,c1,c2);
586         r[8]=c3;
587         c3=0;
588         mul_add_c(a[2],b[7],c1,c2,c3);
589         mul_add_c(a[3],b[6],c1,c2,c3);
590         mul_add_c(a[4],b[5],c1,c2,c3);
591         mul_add_c(a[5],b[4],c1,c2,c3);
592         mul_add_c(a[6],b[3],c1,c2,c3);
593         mul_add_c(a[7],b[2],c1,c2,c3);
594         r[9]=c1;
595         c1=0;
596         mul_add_c(a[7],b[3],c2,c3,c1);
597         mul_add_c(a[6],b[4],c2,c3,c1);
598         mul_add_c(a[5],b[5],c2,c3,c1);
599         mul_add_c(a[4],b[6],c2,c3,c1);
600         mul_add_c(a[3],b[7],c2,c3,c1);
601         r[10]=c2;
602         c2=0;
603         mul_add_c(a[4],b[7],c3,c1,c2);
604         mul_add_c(a[5],b[6],c3,c1,c2);
605         mul_add_c(a[6],b[5],c3,c1,c2);
606         mul_add_c(a[7],b[4],c3,c1,c2);
607         r[11]=c3;
608         c3=0;
609         mul_add_c(a[7],b[5],c1,c2,c3);
610         mul_add_c(a[6],b[6],c1,c2,c3);
611         mul_add_c(a[5],b[7],c1,c2,c3);
612         r[12]=c1;
613         c1=0;
614         mul_add_c(a[6],b[7],c2,c3,c1);
615         mul_add_c(a[7],b[6],c2,c3,c1);
616         r[13]=c2;
617         c2=0;
618         mul_add_c(a[7],b[7],c3,c1,c2);
619         r[14]=c3;
620         r[15]=c1;
621         }
622
623 void bn_mul_comba4(r,a,b)
624 BN_ULONG *r,*a,*b;
625         {
626 #ifdef BN_LLONG
627         BN_ULLONG t;
628 #else
629         BN_ULONG bl,bh;
630 #endif
631         BN_ULONG t1,t2;
632         BN_ULONG c1,c2,c3;
633
634         c1=0;
635         c2=0;
636         c3=0;
637         mul_add_c(a[0],b[0],c1,c2,c3);
638         r[0]=c1;
639         c1=0;
640         mul_add_c(a[0],b[1],c2,c3,c1);
641         mul_add_c(a[1],b[0],c2,c3,c1);
642         r[1]=c2;
643         c2=0;
644         mul_add_c(a[2],b[0],c3,c1,c2);
645         mul_add_c(a[1],b[1],c3,c1,c2);
646         mul_add_c(a[0],b[2],c3,c1,c2);
647         r[2]=c3;
648         c3=0;
649         mul_add_c(a[0],b[3],c1,c2,c3);
650         mul_add_c(a[1],b[2],c1,c2,c3);
651         mul_add_c(a[2],b[1],c1,c2,c3);
652         mul_add_c(a[3],b[0],c1,c2,c3);
653         r[3]=c1;
654         c1=0;
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         r[4]=c2;
659         c2=0;
660         mul_add_c(a[2],b[3],c3,c1,c2);
661         mul_add_c(a[3],b[2],c3,c1,c2);
662         r[5]=c3;
663         c3=0;
664         mul_add_c(a[3],b[3],c1,c2,c3);
665         r[6]=c1;
666         r[7]=c2;
667         }
668
669 void bn_sqr_comba8(r,a)
670 BN_ULONG *r,*a;
671         {
672 #ifdef BN_LLONG
673         BN_ULLONG t,tt;
674 #else
675         BN_ULONG bl,bh;
676 #endif
677         BN_ULONG t1,t2;
678         BN_ULONG c1,c2,c3;
679
680         c1=0;
681         c2=0;
682         c3=0;
683         sqr_add_c(a,0,c1,c2,c3);
684         r[0]=c1;
685         c1=0;
686         sqr_add_c2(a,1,0,c2,c3,c1);
687         r[1]=c2;
688         c2=0;
689         sqr_add_c(a,1,c3,c1,c2);
690         sqr_add_c2(a,2,0,c3,c1,c2);
691         r[2]=c3;
692         c3=0;
693         sqr_add_c2(a,3,0,c1,c2,c3);
694         sqr_add_c2(a,2,1,c1,c2,c3);
695         r[3]=c1;
696         c1=0;
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);
700         r[4]=c2;
701         c2=0;
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);
705         r[5]=c3;
706         c3=0;
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);
711         r[6]=c1;
712         c1=0;
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);
717         r[7]=c2;
718         c2=0;
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);
723         r[8]=c3;
724         c3=0;
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);
728         r[9]=c1;
729         c1=0;
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);
733         r[10]=c2;
734         c2=0;
735         sqr_add_c2(a,7,4,c3,c1,c2);
736         sqr_add_c2(a,6,5,c3,c1,c2);
737         r[11]=c3;
738         c3=0;
739         sqr_add_c(a,6,c1,c2,c3);
740         sqr_add_c2(a,7,5,c1,c2,c3);
741         r[12]=c1;
742         c1=0;
743         sqr_add_c2(a,7,6,c2,c3,c1);
744         r[13]=c2;
745         c2=0;
746         sqr_add_c(a,7,c3,c1,c2);
747         r[14]=c3;
748         r[15]=c1;
749         }
750
751 void bn_sqr_comba4(r,a)
752 BN_ULONG *r,*a;
753         {
754 #ifdef BN_LLONG
755         BN_ULLONG t,tt;
756 #else
757         BN_ULONG bl,bh;
758 #endif
759         BN_ULONG t1,t2;
760         BN_ULONG c1,c2,c3;
761
762         c1=0;
763         c2=0;
764         c3=0;
765         sqr_add_c(a,0,c1,c2,c3);
766         r[0]=c1;
767         c1=0;
768         sqr_add_c2(a,1,0,c2,c3,c1);
769         r[1]=c2;
770         c2=0;
771         sqr_add_c(a,1,c3,c1,c2);
772         sqr_add_c2(a,2,0,c3,c1,c2);
773         r[2]=c3;
774         c3=0;
775         sqr_add_c2(a,3,0,c1,c2,c3);
776         sqr_add_c2(a,2,1,c1,c2,c3);
777         r[3]=c1;
778         c1=0;
779         sqr_add_c(a,2,c2,c3,c1);
780         sqr_add_c2(a,3,1,c2,c3,c1);
781         r[4]=c2;
782         c2=0;
783         sqr_add_c2(a,3,2,c3,c1,c2);
784         r[5]=c3;
785         c3=0;
786         sqr_add_c(a,3,c1,c2,c3);
787         r[6]=c1;
788         r[7]=c2;
789         }
790 #else
791
792 /* hmm... is it faster just to do a multiply? */
793 void bn_sqr_comba4(r,a)
794 BN_ULONG *r,*a;
795         {
796         BN_ULONG t[8];
797         bn_sqr_normal(r,a,4,t);
798         }
799
800 void bn_sqr_comba8(r,a)
801 BN_ULONG *r,*a;
802         {
803         BN_ULONG t[16];
804         bn_sqr_normal(r,a,8,t);
805         }
806
807 void bn_mul_comba4(r,a,b)
808 BN_ULONG *r,*a,*b;
809         {
810         r[4]=bn_mul_words(    &(r[0]),a,4,b[0]);
811         r[5]=bn_mul_add_words(&(r[1]),a,4,b[1]);
812         r[6]=bn_mul_add_words(&(r[2]),a,4,b[2]);
813         r[7]=bn_mul_add_words(&(r[3]),a,4,b[3]);
814         }
815
816 void bn_mul_comba8(r,a,b)
817 BN_ULONG *r,*a,*b;
818         {
819         r[ 8]=bn_mul_words(    &(r[0]),a,8,b[0]);
820         r[ 9]=bn_mul_add_words(&(r[1]),a,8,b[1]);
821         r[10]=bn_mul_add_words(&(r[2]),a,8,b[2]);
822         r[11]=bn_mul_add_words(&(r[3]),a,8,b[3]);
823         r[12]=bn_mul_add_words(&(r[4]),a,8,b[4]);
824         r[13]=bn_mul_add_words(&(r[5]),a,8,b[5]);
825         r[14]=bn_mul_add_words(&(r[6]),a,8,b[6]);
826         r[15]=bn_mul_add_words(&(r[7]),a,8,b[7]);
827         }
828
829 #endif /* BN_COMBA */