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#ifndef BN_DEBUG
60# undef NDEBUG /* avoid conflicting definitions */
61# define NDEBUG
62#endif
63
64#include <stdio.h>
65#include <assert.h>
66#include "cryptlib.h"
67#include "bn_lcl.h"
68
69#if defined(BN_LLONG) || defined(BN_UMULT_HIGH)
70
71BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
72	{
73	BN_ULONG c1=0;
74
75	assert(num >= 0);
76	if (num <= 0) return(c1);
77
78	while (num&~3)
79		{
80		mul_add(rp[0],ap[0],w,c1);
81		mul_add(rp[1],ap[1],w,c1);
82		mul_add(rp[2],ap[2],w,c1);
83		mul_add(rp[3],ap[3],w,c1);
84		ap+=4; rp+=4; num-=4;
85		}
86	if (num)
87		{
88		mul_add(rp[0],ap[0],w,c1); if (--num==0) return c1;
89		mul_add(rp[1],ap[1],w,c1); if (--num==0) return c1;
90		mul_add(rp[2],ap[2],w,c1); return c1;
91		}
92
93	return(c1);
94	}
95
96BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
97	{
98	BN_ULONG c1=0;
99
100	assert(num >= 0);
101	if (num <= 0) return(c1);
102
103	while (num&~3)
104		{
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);
109		ap+=4; rp+=4; num-=4;
110		}
111	if (num)
112		{
113		mul(rp[0],ap[0],w,c1); if (--num == 0) return c1;
114		mul(rp[1],ap[1],w,c1); if (--num == 0) return c1;
115		mul(rp[2],ap[2],w,c1);
116		}
117	return(c1);
118	}
119
120void bn_sqr_words(BN_ULONG *r, const BN_ULONG *a, int n)
121        {
122	assert(n >= 0);
123	if (n <= 0) return;
124	while (n&~3)
125		{
126		sqr(r[0],r[1],a[0]);
127		sqr(r[2],r[3],a[1]);
128		sqr(r[4],r[5],a[2]);
129		sqr(r[6],r[7],a[3]);
130		a+=4; r+=8; n-=4;
131		}
132	if (n)
133		{
134		sqr(r[0],r[1],a[0]); if (--n == 0) return;
135		sqr(r[2],r[3],a[1]); if (--n == 0) return;
136		sqr(r[4],r[5],a[2]);
137		}
138	}
139
140#else /* !(defined(BN_LLONG) || defined(BN_UMULT_HIGH)) */
141
142BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
143	{
144	BN_ULONG c=0;
145	BN_ULONG bl,bh;
146
147	assert(num >= 0);
148	if (num <= 0) return((BN_ULONG)0);
149
150	bl=LBITS(w);
151	bh=HBITS(w);
152
153	for (;;)
154		{
155		mul_add(rp[0],ap[0],bl,bh,c);
156		if (--num == 0) break;
157		mul_add(rp[1],ap[1],bl,bh,c);
158		if (--num == 0) break;
159		mul_add(rp[2],ap[2],bl,bh,c);
160		if (--num == 0) break;
161		mul_add(rp[3],ap[3],bl,bh,c);
162		if (--num == 0) break;
163		ap+=4;
164		rp+=4;
165		}
166	return(c);
167	}
168
169BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
170	{
171	BN_ULONG carry=0;
172	BN_ULONG bl,bh;
173
174	assert(num >= 0);
175	if (num <= 0) return((BN_ULONG)0);
176
177	bl=LBITS(w);
178	bh=HBITS(w);
179
180	for (;;)
181		{
182		mul(rp[0],ap[0],bl,bh,carry);
183		if (--num == 0) break;
184		mul(rp[1],ap[1],bl,bh,carry);
185		if (--num == 0) break;
186		mul(rp[2],ap[2],bl,bh,carry);
187		if (--num == 0) break;
188		mul(rp[3],ap[3],bl,bh,carry);
189		if (--num == 0) break;
190		ap+=4;
191		rp+=4;
192		}
193	return(carry);
194	}
195
196void bn_sqr_words(BN_ULONG *r, const BN_ULONG *a, int n)
197        {
198	assert(n >= 0);
199	if (n <= 0) return;
200	for (;;)
201		{
202		sqr64(r[0],r[1],a[0]);
203		if (--n == 0) break;
204
205		sqr64(r[2],r[3],a[1]);
206		if (--n == 0) break;
207
208		sqr64(r[4],r[5],a[2]);
209		if (--n == 0) break;
210
211		sqr64(r[6],r[7],a[3]);
212		if (--n == 0) break;
213
214		a+=4;
215		r+=8;
216		}
217	}
218
219#endif /* !(defined(BN_LLONG) || defined(BN_UMULT_HIGH)) */
220
221#if defined(BN_LLONG) && defined(BN_DIV2W)
222
223BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
224	{
225	return((BN_ULONG)(((((BN_ULLONG)h)<<BN_BITS2)|l)/(BN_ULLONG)d));
226	}
227
228#else
229
230/* Divide h,l by d and return the result. */
231/* I need to test this some more :-( */
232BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
233	{
234	BN_ULONG dh,dl,q,ret=0,th,tl,t;
235	int i,count=2;
236
237	if (d == 0) return(BN_MASK2);
238
239	i=BN_num_bits_word(d);
240	assert((i == BN_BITS2) || (h > (BN_ULONG)1<<i));
241
242	i=BN_BITS2-i;
243	if (h >= d) h-=d;
244
245	if (i)
246		{
247		d<<=i;
248		h=(h<<i)|(l>>(BN_BITS2-i));
249		l<<=i;
250		}
251	dh=(d&BN_MASK2h)>>BN_BITS4;
252	dl=(d&BN_MASK2l);
253	for (;;)
254		{
255		if ((h>>BN_BITS4) == dh)
256			q=BN_MASK2l;
257		else
258			q=h/dh;
259
260		th=q*dh;
261		tl=dl*q;
262		for (;;)
263			{
264			t=h-th;
265			if ((t&BN_MASK2h) ||
266				((tl) <= (
267					(t<<BN_BITS4)|
268					((l&BN_MASK2h)>>BN_BITS4))))
269				break;
270			q--;
271			th-=dh;
272			tl-=dl;
273			}
274		t=(tl>>BN_BITS4);
275		tl=(tl<<BN_BITS4)&BN_MASK2h;
276		th+=t;
277
278		if (l < tl) th++;
279		l-=tl;
280		if (h < th)
281			{
282			h+=d;
283			q--;
284			}
285		h-=th;
286
287		if (--count == 0) break;
288
289		ret=q<<BN_BITS4;
290		h=((h<<BN_BITS4)|(l>>BN_BITS4))&BN_MASK2;
291		l=(l&BN_MASK2l)<<BN_BITS4;
292		}
293	ret|=q;
294	return(ret);
295	}
296#endif /* !defined(BN_LLONG) && defined(BN_DIV2W) */
297
298#ifdef BN_LLONG
299BN_ULONG bn_add_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b, int n)
300        {
301	BN_ULLONG ll=0;
302
303	assert(n >= 0);
304	if (n <= 0) return((BN_ULONG)0);
305
306	for (;;)
307		{
308		ll+=(BN_ULLONG)a[0]+b[0];
309		r[0]=(BN_ULONG)ll&BN_MASK2;
310		ll>>=BN_BITS2;
311		if (--n <= 0) break;
312
313		ll+=(BN_ULLONG)a[1]+b[1];
314		r[1]=(BN_ULONG)ll&BN_MASK2;
315		ll>>=BN_BITS2;
316		if (--n <= 0) break;
317
318		ll+=(BN_ULLONG)a[2]+b[2];
319		r[2]=(BN_ULONG)ll&BN_MASK2;
320		ll>>=BN_BITS2;
321		if (--n <= 0) break;
322
323		ll+=(BN_ULLONG)a[3]+b[3];
324		r[3]=(BN_ULONG)ll&BN_MASK2;
325		ll>>=BN_BITS2;
326		if (--n <= 0) break;
327
328		a+=4;
329		b+=4;
330		r+=4;
331		}
332	return((BN_ULONG)ll);
333	}
334#else /* !BN_LLONG */
335BN_ULONG bn_add_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b, int n)
336        {
337	BN_ULONG c,l,t;
338
339	assert(n >= 0);
340	if (n <= 0) return((BN_ULONG)0);
341
342	c=0;
343	for (;;)
344		{
345		t=a[0];
346		t=(t+c)&BN_MASK2;
347		c=(t < c);
348		l=(t+b[0])&BN_MASK2;
349		c+=(l < t);
350		r[0]=l;
351		if (--n <= 0) break;
352
353		t=a[1];
354		t=(t+c)&BN_MASK2;
355		c=(t < c);
356		l=(t+b[1])&BN_MASK2;
357		c+=(l < t);
358		r[1]=l;
359		if (--n <= 0) break;
360
361		t=a[2];
362		t=(t+c)&BN_MASK2;
363		c=(t < c);
364		l=(t+b[2])&BN_MASK2;
365		c+=(l < t);
366		r[2]=l;
367		if (--n <= 0) break;
368
369		t=a[3];
370		t=(t+c)&BN_MASK2;
371		c=(t < c);
372		l=(t+b[3])&BN_MASK2;
373		c+=(l < t);
374		r[3]=l;
375		if (--n <= 0) break;
376
377		a+=4;
378		b+=4;
379		r+=4;
380		}
381	return((BN_ULONG)c);
382	}
383#endif /* !BN_LLONG */
384
385BN_ULONG bn_sub_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b, int n)
386        {
387	BN_ULONG t1,t2;
388	int c=0;
389
390	assert(n >= 0);
391	if (n <= 0) return((BN_ULONG)0);
392
393	for (;;)
394		{
395		t1=a[0]; t2=b[0];
396		r[0]=(t1-t2-c)&BN_MASK2;
397		if (t1 != t2) c=(t1 < t2);
398		if (--n <= 0) break;
399
400		t1=a[1]; t2=b[1];
401		r[1]=(t1-t2-c)&BN_MASK2;
402		if (t1 != t2) c=(t1 < t2);
403		if (--n <= 0) break;
404
405		t1=a[2]; t2=b[2];
406		r[2]=(t1-t2-c)&BN_MASK2;
407		if (t1 != t2) c=(t1 < t2);
408		if (--n <= 0) break;
409
410		t1=a[3]; t2=b[3];
411		r[3]=(t1-t2-c)&BN_MASK2;
412		if (t1 != t2) c=(t1 < t2);
413		if (--n <= 0) break;
414
415		a+=4;
416		b+=4;
417		r+=4;
418		}
419	return(c);
420	}
421
422#ifdef BN_MUL_COMBA
423
424#undef bn_mul_comba8
425#undef bn_mul_comba4
426#undef bn_sqr_comba8
427#undef bn_sqr_comba4
428
429/* mul_add_c(a,b,c0,c1,c2)  -- c+=a*b for three word number c=(c2,c1,c0) */
430/* mul_add_c2(a,b,c0,c1,c2) -- c+=2*a*b for three word number c=(c2,c1,c0) */
431/* sqr_add_c(a,i,c0,c1,c2)  -- c+=a[i]^2 for three word number c=(c2,c1,c0) */
432/* sqr_add_c2(a,i,c0,c1,c2) -- c+=2*a[i]*a[j] for three word number c=(c2,c1,c0) */
433
434#ifdef BN_LLONG
435#define mul_add_c(a,b,c0,c1,c2) \
436	t=(BN_ULLONG)a*b; \
437	t1=(BN_ULONG)Lw(t); \
438	t2=(BN_ULONG)Hw(t); \
439	c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
440	c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
441
442#define mul_add_c2(a,b,c0,c1,c2) \
443	t=(BN_ULLONG)a*b; \
444	tt=(t+t)&BN_MASK; \
445	if (tt < t) c2++; \
446	t1=(BN_ULONG)Lw(tt); \
447	t2=(BN_ULONG)Hw(tt); \
448	c0=(c0+t1)&BN_MASK2;  \
449	if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
450	c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
451
452#define sqr_add_c(a,i,c0,c1,c2) \
453	t=(BN_ULLONG)a[i]*a[i]; \
454	t1=(BN_ULONG)Lw(t); \
455	t2=(BN_ULONG)Hw(t); \
456	c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
457	c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
458
459#define sqr_add_c2(a,i,j,c0,c1,c2) \
460	mul_add_c2((a)[i],(a)[j],c0,c1,c2)
461
462#elif defined(BN_UMULT_HIGH)
463
464#define mul_add_c(a,b,c0,c1,c2)	{	\
465	BN_ULONG ta=(a),tb=(b);		\
466	t1 = ta * tb;			\
467	t2 = BN_UMULT_HIGH(ta,tb);	\
468	c0 += t1; t2 += (c0<t1)?1:0;	\
469	c1 += t2; c2 += (c1<t2)?1:0;	\
470	}
471
472#define mul_add_c2(a,b,c0,c1,c2) {	\
473	BN_ULONG ta=(a),tb=(b),t0;	\
474	t1 = BN_UMULT_HIGH(ta,tb);	\
475	t0 = ta * tb;			\
476	t2 = t1+t1; c2 += (t2<t1)?1:0;	\
477	t1 = t0+t0; t2 += (t1<t0)?1:0;	\
478	c0 += t1; t2 += (c0<t1)?1:0;	\
479	c1 += t2; c2 += (c1<t2)?1:0;	\
480	}
481
482#define sqr_add_c(a,i,c0,c1,c2)	{	\
483	BN_ULONG ta=(a)[i];		\
484	t1 = ta * ta;			\
485	t2 = BN_UMULT_HIGH(ta,ta);	\
486	c0 += t1; t2 += (c0<t1)?1:0;	\
487	c1 += t2; c2 += (c1<t2)?1:0;	\
488	}
489
490#define sqr_add_c2(a,i,j,c0,c1,c2)	\
491	mul_add_c2((a)[i],(a)[j],c0,c1,c2)
492
493#else /* !BN_LLONG */
494#define mul_add_c(a,b,c0,c1,c2) \
495	t1=LBITS(a); t2=HBITS(a); \
496	bl=LBITS(b); bh=HBITS(b); \
497	mul64(t1,t2,bl,bh); \
498	c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
499	c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
500
501#define mul_add_c2(a,b,c0,c1,c2) \
502	t1=LBITS(a); t2=HBITS(a); \
503	bl=LBITS(b); bh=HBITS(b); \
504	mul64(t1,t2,bl,bh); \
505	if (t2 & BN_TBIT) c2++; \
506	t2=(t2+t2)&BN_MASK2; \
507	if (t1 & BN_TBIT) t2++; \
508	t1=(t1+t1)&BN_MASK2; \
509	c0=(c0+t1)&BN_MASK2;  \
510	if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
511	c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
512
513#define sqr_add_c(a,i,c0,c1,c2) \
514	sqr64(t1,t2,(a)[i]); \
515	c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
516	c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
517
518#define sqr_add_c2(a,i,j,c0,c1,c2) \
519	mul_add_c2((a)[i],(a)[j],c0,c1,c2)
520#endif /* !BN_LLONG */
521
522void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
523	{
524#ifdef BN_LLONG
525	BN_ULLONG t;
526#else
527	BN_ULONG bl,bh;
528#endif
529	BN_ULONG t1,t2;
530	BN_ULONG c1,c2,c3;
531
532	c1=0;
533	c2=0;
534	c3=0;
535	mul_add_c(a[0],b[0],c1,c2,c3);
536	r[0]=c1;
537	c1=0;
538	mul_add_c(a[0],b[1],c2,c3,c1);
539	mul_add_c(a[1],b[0],c2,c3,c1);
540	r[1]=c2;
541	c2=0;
542	mul_add_c(a[2],b[0],c3,c1,c2);
543	mul_add_c(a[1],b[1],c3,c1,c2);
544	mul_add_c(a[0],b[2],c3,c1,c2);
545	r[2]=c3;
546	c3=0;
547	mul_add_c(a[0],b[3],c1,c2,c3);
548	mul_add_c(a[1],b[2],c1,c2,c3);
549	mul_add_c(a[2],b[1],c1,c2,c3);
550	mul_add_c(a[3],b[0],c1,c2,c3);
551	r[3]=c1;
552	c1=0;
553	mul_add_c(a[4],b[0],c2,c3,c1);
554	mul_add_c(a[3],b[1],c2,c3,c1);
555	mul_add_c(a[2],b[2],c2,c3,c1);
556	mul_add_c(a[1],b[3],c2,c3,c1);
557	mul_add_c(a[0],b[4],c2,c3,c1);
558	r[4]=c2;
559	c2=0;
560	mul_add_c(a[0],b[5],c3,c1,c2);
561	mul_add_c(a[1],b[4],c3,c1,c2);
562	mul_add_c(a[2],b[3],c3,c1,c2);
563	mul_add_c(a[3],b[2],c3,c1,c2);
564	mul_add_c(a[4],b[1],c3,c1,c2);
565	mul_add_c(a[5],b[0],c3,c1,c2);
566	r[5]=c3;
567	c3=0;
568	mul_add_c(a[6],b[0],c1,c2,c3);
569	mul_add_c(a[5],b[1],c1,c2,c3);
570	mul_add_c(a[4],b[2],c1,c2,c3);
571	mul_add_c(a[3],b[3],c1,c2,c3);
572	mul_add_c(a[2],b[4],c1,c2,c3);
573	mul_add_c(a[1],b[5],c1,c2,c3);
574	mul_add_c(a[0],b[6],c1,c2,c3);
575	r[6]=c1;
576	c1=0;
577	mul_add_c(a[0],b[7],c2,c3,c1);
578	mul_add_c(a[1],b[6],c2,c3,c1);
579	mul_add_c(a[2],b[5],c2,c3,c1);
580	mul_add_c(a[3],b[4],c2,c3,c1);
581	mul_add_c(a[4],b[3],c2,c3,c1);
582	mul_add_c(a[5],b[2],c2,c3,c1);
583	mul_add_c(a[6],b[1],c2,c3,c1);
584	mul_add_c(a[7],b[0],c2,c3,c1);
585	r[7]=c2;
586	c2=0;
587	mul_add_c(a[7],b[1],c3,c1,c2);
588	mul_add_c(a[6],b[2],c3,c1,c2);
589	mul_add_c(a[5],b[3],c3,c1,c2);
590	mul_add_c(a[4],b[4],c3,c1,c2);
591	mul_add_c(a[3],b[5],c3,c1,c2);
592	mul_add_c(a[2],b[6],c3,c1,c2);
593	mul_add_c(a[1],b[7],c3,c1,c2);
594	r[8]=c3;
595	c3=0;
596	mul_add_c(a[2],b[7],c1,c2,c3);
597	mul_add_c(a[3],b[6],c1,c2,c3);
598	mul_add_c(a[4],b[5],c1,c2,c3);
599	mul_add_c(a[5],b[4],c1,c2,c3);
600	mul_add_c(a[6],b[3],c1,c2,c3);
601	mul_add_c(a[7],b[2],c1,c2,c3);
602	r[9]=c1;
603	c1=0;
604	mul_add_c(a[7],b[3],c2,c3,c1);
605	mul_add_c(a[6],b[4],c2,c3,c1);
606	mul_add_c(a[5],b[5],c2,c3,c1);
607	mul_add_c(a[4],b[6],c2,c3,c1);
608	mul_add_c(a[3],b[7],c2,c3,c1);
609	r[10]=c2;
610	c2=0;
611	mul_add_c(a[4],b[7],c3,c1,c2);
612	mul_add_c(a[5],b[6],c3,c1,c2);
613	mul_add_c(a[6],b[5],c3,c1,c2);
614	mul_add_c(a[7],b[4],c3,c1,c2);
615	r[11]=c3;
616	c3=0;
617	mul_add_c(a[7],b[5],c1,c2,c3);
618	mul_add_c(a[6],b[6],c1,c2,c3);
619	mul_add_c(a[5],b[7],c1,c2,c3);
620	r[12]=c1;
621	c1=0;
622	mul_add_c(a[6],b[7],c2,c3,c1);
623	mul_add_c(a[7],b[6],c2,c3,c1);
624	r[13]=c2;
625	c2=0;
626	mul_add_c(a[7],b[7],c3,c1,c2);
627	r[14]=c3;
628	r[15]=c1;
629	}
630
631void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
632	{
633#ifdef BN_LLONG
634	BN_ULLONG t;
635#else
636	BN_ULONG bl,bh;
637#endif
638	BN_ULONG t1,t2;
639	BN_ULONG c1,c2,c3;
640
641	c1=0;
642	c2=0;
643	c3=0;
644	mul_add_c(a[0],b[0],c1,c2,c3);
645	r[0]=c1;
646	c1=0;
647	mul_add_c(a[0],b[1],c2,c3,c1);
648	mul_add_c(a[1],b[0],c2,c3,c1);
649	r[1]=c2;
650	c2=0;
651	mul_add_c(a[2],b[0],c3,c1,c2);
652	mul_add_c(a[1],b[1],c3,c1,c2);
653	mul_add_c(a[0],b[2],c3,c1,c2);
654	r[2]=c3;
655	c3=0;
656	mul_add_c(a[0],b[3],c1,c2,c3);
657	mul_add_c(a[1],b[2],c1,c2,c3);
658	mul_add_c(a[2],b[1],c1,c2,c3);
659	mul_add_c(a[3],b[0],c1,c2,c3);
660	r[3]=c1;
661	c1=0;
662	mul_add_c(a[3],b[1],c2,c3,c1);
663	mul_add_c(a[2],b[2],c2,c3,c1);
664	mul_add_c(a[1],b[3],c2,c3,c1);
665	r[4]=c2;
666	c2=0;
667	mul_add_c(a[2],b[3],c3,c1,c2);
668	mul_add_c(a[3],b[2],c3,c1,c2);
669	r[5]=c3;
670	c3=0;
671	mul_add_c(a[3],b[3],c1,c2,c3);
672	r[6]=c1;
673	r[7]=c2;
674	}
675
676void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
677	{
678#ifdef BN_LLONG
679	BN_ULLONG t,tt;
680#else
681	BN_ULONG bl,bh;
682#endif
683	BN_ULONG t1,t2;
684	BN_ULONG c1,c2,c3;
685
686	c1=0;
687	c2=0;
688	c3=0;
689	sqr_add_c(a,0,c1,c2,c3);
690	r[0]=c1;
691	c1=0;
692	sqr_add_c2(a,1,0,c2,c3,c1);
693	r[1]=c2;
694	c2=0;
695	sqr_add_c(a,1,c3,c1,c2);
696	sqr_add_c2(a,2,0,c3,c1,c2);
697	r[2]=c3;
698	c3=0;
699	sqr_add_c2(a,3,0,c1,c2,c3);
700	sqr_add_c2(a,2,1,c1,c2,c3);
701	r[3]=c1;
702	c1=0;
703	sqr_add_c(a,2,c2,c3,c1);
704	sqr_add_c2(a,3,1,c2,c3,c1);
705	sqr_add_c2(a,4,0,c2,c3,c1);
706	r[4]=c2;
707	c2=0;
708	sqr_add_c2(a,5,0,c3,c1,c2);
709	sqr_add_c2(a,4,1,c3,c1,c2);
710	sqr_add_c2(a,3,2,c3,c1,c2);
711	r[5]=c3;
712	c3=0;
713	sqr_add_c(a,3,c1,c2,c3);
714	sqr_add_c2(a,4,2,c1,c2,c3);
715	sqr_add_c2(a,5,1,c1,c2,c3);
716	sqr_add_c2(a,6,0,c1,c2,c3);
717	r[6]=c1;
718	c1=0;
719	sqr_add_c2(a,7,0,c2,c3,c1);
720	sqr_add_c2(a,6,1,c2,c3,c1);
721	sqr_add_c2(a,5,2,c2,c3,c1);
722	sqr_add_c2(a,4,3,c2,c3,c1);
723	r[7]=c2;
724	c2=0;
725	sqr_add_c(a,4,c3,c1,c2);
726	sqr_add_c2(a,5,3,c3,c1,c2);
727	sqr_add_c2(a,6,2,c3,c1,c2);
728	sqr_add_c2(a,7,1,c3,c1,c2);
729	r[8]=c3;
730	c3=0;
731	sqr_add_c2(a,7,2,c1,c2,c3);
732	sqr_add_c2(a,6,3,c1,c2,c3);
733	sqr_add_c2(a,5,4,c1,c2,c3);
734	r[9]=c1;
735	c1=0;
736	sqr_add_c(a,5,c2,c3,c1);
737	sqr_add_c2(a,6,4,c2,c3,c1);
738	sqr_add_c2(a,7,3,c2,c3,c1);
739	r[10]=c2;
740	c2=0;
741	sqr_add_c2(a,7,4,c3,c1,c2);
742	sqr_add_c2(a,6,5,c3,c1,c2);
743	r[11]=c3;
744	c3=0;
745	sqr_add_c(a,6,c1,c2,c3);
746	sqr_add_c2(a,7,5,c1,c2,c3);
747	r[12]=c1;
748	c1=0;
749	sqr_add_c2(a,7,6,c2,c3,c1);
750	r[13]=c2;
751	c2=0;
752	sqr_add_c(a,7,c3,c1,c2);
753	r[14]=c3;
754	r[15]=c1;
755	}
756
757void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
758	{
759#ifdef BN_LLONG
760	BN_ULLONG t,tt;
761#else
762	BN_ULONG bl,bh;
763#endif
764	BN_ULONG t1,t2;
765	BN_ULONG c1,c2,c3;
766
767	c1=0;
768	c2=0;
769	c3=0;
770	sqr_add_c(a,0,c1,c2,c3);
771	r[0]=c1;
772	c1=0;
773	sqr_add_c2(a,1,0,c2,c3,c1);
774	r[1]=c2;
775	c2=0;
776	sqr_add_c(a,1,c3,c1,c2);
777	sqr_add_c2(a,2,0,c3,c1,c2);
778	r[2]=c3;
779	c3=0;
780	sqr_add_c2(a,3,0,c1,c2,c3);
781	sqr_add_c2(a,2,1,c1,c2,c3);
782	r[3]=c1;
783	c1=0;
784	sqr_add_c(a,2,c2,c3,c1);
785	sqr_add_c2(a,3,1,c2,c3,c1);
786	r[4]=c2;
787	c2=0;
788	sqr_add_c2(a,3,2,c3,c1,c2);
789	r[5]=c3;
790	c3=0;
791	sqr_add_c(a,3,c1,c2,c3);
792	r[6]=c1;
793	r[7]=c2;
794	}
795#else /* !BN_MUL_COMBA */
796
797/* hmm... is it faster just to do a multiply? */
798#undef bn_sqr_comba4
799void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
800	{
801	BN_ULONG t[8];
802	bn_sqr_normal(r,a,4,t);
803	}
804
805#undef bn_sqr_comba8
806void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
807	{
808	BN_ULONG t[16];
809	bn_sqr_normal(r,a,8,t);
810	}
811
812void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
813	{
814	r[4]=bn_mul_words(    &(r[0]),a,4,b[0]);
815	r[5]=bn_mul_add_words(&(r[1]),a,4,b[1]);
816	r[6]=bn_mul_add_words(&(r[2]),a,4,b[2]);
817	r[7]=bn_mul_add_words(&(r[3]),a,4,b[3]);
818	}
819
820void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
821	{
822	r[ 8]=bn_mul_words(    &(r[0]),a,8,b[0]);
823	r[ 9]=bn_mul_add_words(&(r[1]),a,8,b[1]);
824	r[10]=bn_mul_add_words(&(r[2]),a,8,b[2]);
825	r[11]=bn_mul_add_words(&(r[3]),a,8,b[3]);
826	r[12]=bn_mul_add_words(&(r[4]),a,8,b[4]);
827	r[13]=bn_mul_add_words(&(r[5]),a,8,b[5]);
828	r[14]=bn_mul_add_words(&(r[6]),a,8,b[6]);
829	r[15]=bn_mul_add_words(&(r[7]),a,8,b[7]);
830	}
831
832#endif /* !BN_MUL_COMBA */
833