sl@0: /* crypto/bn/bn_sqr.c */
sl@0: /* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
sl@0:  * All rights reserved.
sl@0:  *
sl@0:  * This package is an SSL implementation written
sl@0:  * by Eric Young (eay@cryptsoft.com).
sl@0:  * The implementation was written so as to conform with Netscapes SSL.
sl@0:  * 
sl@0:  * This library is free for commercial and non-commercial use as long as
sl@0:  * the following conditions are aheared to.  The following conditions
sl@0:  * apply to all code found in this distribution, be it the RC4, RSA,
sl@0:  * lhash, DES, etc., code; not just the SSL code.  The SSL documentation
sl@0:  * included with this distribution is covered by the same copyright terms
sl@0:  * except that the holder is Tim Hudson (tjh@cryptsoft.com).
sl@0:  * 
sl@0:  * Copyright remains Eric Young's, and as such any Copyright notices in
sl@0:  * the code are not to be removed.
sl@0:  * If this package is used in a product, Eric Young should be given attribution
sl@0:  * as the author of the parts of the library used.
sl@0:  * This can be in the form of a textual message at program startup or
sl@0:  * in documentation (online or textual) provided with the package.
sl@0:  * 
sl@0:  * Redistribution and use in source and binary forms, with or without
sl@0:  * modification, are permitted provided that the following conditions
sl@0:  * are met:
sl@0:  * 1. Redistributions of source code must retain the copyright
sl@0:  *    notice, this list of conditions and the following disclaimer.
sl@0:  * 2. Redistributions in binary form must reproduce the above copyright
sl@0:  *    notice, this list of conditions and the following disclaimer in the
sl@0:  *    documentation and/or other materials provided with the distribution.
sl@0:  * 3. All advertising materials mentioning features or use of this software
sl@0:  *    must display the following acknowledgement:
sl@0:  *    "This product includes cryptographic software written by
sl@0:  *     Eric Young (eay@cryptsoft.com)"
sl@0:  *    The word 'cryptographic' can be left out if the rouines from the library
sl@0:  *    being used are not cryptographic related :-).
sl@0:  * 4. If you include any Windows specific code (or a derivative thereof) from 
sl@0:  *    the apps directory (application code) you must include an acknowledgement:
sl@0:  *    "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
sl@0:  * 
sl@0:  * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
sl@0:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
sl@0:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
sl@0:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
sl@0:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
sl@0:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
sl@0:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
sl@0:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
sl@0:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
sl@0:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
sl@0:  * SUCH DAMAGE.
sl@0:  * 
sl@0:  * The licence and distribution terms for any publically available version or
sl@0:  * derivative of this code cannot be changed.  i.e. this code cannot simply be
sl@0:  * copied and put under another distribution licence
sl@0:  * [including the GNU Public Licence.]
sl@0:  */
sl@0: 
sl@0: #include <stdio.h>
sl@0: #include "cryptlib.h"
sl@0: #include "bn_lcl.h"
sl@0: 
sl@0: /* r must not be a */
sl@0: /* I've just gone over this and it is now %20 faster on x86 - eay - 27 Jun 96 */
sl@0: EXPORT_C int BN_sqr(BIGNUM *r, const BIGNUM *a, BN_CTX *ctx)
sl@0: 	{
sl@0: 	int max,al;
sl@0: 	int ret = 0;
sl@0: 	BIGNUM *tmp,*rr;
sl@0: 
sl@0: #ifdef BN_COUNT
sl@0: 	fprintf(stderr,"BN_sqr %d * %d\n",a->top,a->top);
sl@0: #endif
sl@0: 	bn_check_top(a);
sl@0: 
sl@0: 	al=a->top;
sl@0: 	if (al <= 0)
sl@0: 		{
sl@0: 		r->top=0;
sl@0: 		return 1;
sl@0: 		}
sl@0: 
sl@0: 	BN_CTX_start(ctx);
sl@0: 	rr=(a != r) ? r : BN_CTX_get(ctx);
sl@0: 	tmp=BN_CTX_get(ctx);
sl@0: 	if (!rr || !tmp) goto err;
sl@0: 
sl@0: 	max = 2 * al; /* Non-zero (from above) */
sl@0: 	if (bn_wexpand(rr,max) == NULL) goto err;
sl@0: 
sl@0: 	if (al == 4)
sl@0: 		{
sl@0: #ifndef BN_SQR_COMBA
sl@0: 		BN_ULONG t[8];
sl@0: 		bn_sqr_normal(rr->d,a->d,4,t);
sl@0: #else
sl@0: 		bn_sqr_comba4(rr->d,a->d);
sl@0: #endif
sl@0: 		}
sl@0: 	else if (al == 8)
sl@0: 		{
sl@0: #ifndef BN_SQR_COMBA
sl@0: 		BN_ULONG t[16];
sl@0: 		bn_sqr_normal(rr->d,a->d,8,t);
sl@0: #else
sl@0: 		bn_sqr_comba8(rr->d,a->d);
sl@0: #endif
sl@0: 		}
sl@0: 	else 
sl@0: 		{
sl@0: #if defined(BN_RECURSION)
sl@0: 		if (al < BN_SQR_RECURSIVE_SIZE_NORMAL)
sl@0: 			{
sl@0: 			BN_ULONG t[BN_SQR_RECURSIVE_SIZE_NORMAL*2];
sl@0: 			bn_sqr_normal(rr->d,a->d,al,t);
sl@0: 			}
sl@0: 		else
sl@0: 			{
sl@0: 			int j,k;
sl@0: 
sl@0: 			j=BN_num_bits_word((BN_ULONG)al);
sl@0: 			j=1<<(j-1);
sl@0: 			k=j+j;
sl@0: 			if (al == j)
sl@0: 				{
sl@0: 				if (bn_wexpand(tmp,k*2) == NULL) goto err;
sl@0: 				bn_sqr_recursive(rr->d,a->d,al,tmp->d);
sl@0: 				}
sl@0: 			else
sl@0: 				{
sl@0: 				if (bn_wexpand(tmp,max) == NULL) goto err;
sl@0: 				bn_sqr_normal(rr->d,a->d,al,tmp->d);
sl@0: 				}
sl@0: 			}
sl@0: #else
sl@0: 		if (bn_wexpand(tmp,max) == NULL) goto err;
sl@0: 		bn_sqr_normal(rr->d,a->d,al,tmp->d);
sl@0: #endif
sl@0: 		}
sl@0: 
sl@0: 	rr->neg=0;
sl@0: 	/* If the most-significant half of the top word of 'a' is zero, then
sl@0: 	 * the square of 'a' will max-1 words. */
sl@0: 	if(a->d[al - 1] == (a->d[al - 1] & BN_MASK2l))
sl@0: 		rr->top = max - 1;
sl@0: 	else
sl@0: 		rr->top = max;
sl@0: 	if (rr != r) BN_copy(r,rr);
sl@0: 	ret = 1;
sl@0:  err:
sl@0: 	bn_check_top(rr);
sl@0: 	bn_check_top(tmp);
sl@0: 	BN_CTX_end(ctx);
sl@0: 	return(ret);
sl@0: 	}
sl@0: 
sl@0: /* tmp must have 2*n words */
sl@0: EXPORT_C void bn_sqr_normal(BN_ULONG *r, const BN_ULONG *a, int n, BN_ULONG *tmp)
sl@0: 	{
sl@0: 	int i,j,max;
sl@0: 	const BN_ULONG *ap;
sl@0: 	BN_ULONG *rp;
sl@0: 
sl@0: 	max=n*2;
sl@0: 	ap=a;
sl@0: 	rp=r;
sl@0: 	rp[0]=rp[max-1]=0;
sl@0: 	rp++;
sl@0: 	j=n;
sl@0: 
sl@0: 	if (--j > 0)
sl@0: 		{
sl@0: 		ap++;
sl@0: 		rp[j]=bn_mul_words(rp,ap,j,ap[-1]);
sl@0: 		rp+=2;
sl@0: 		}
sl@0: 
sl@0: 	for (i=n-2; i>0; i--)
sl@0: 		{
sl@0: 		j--;
sl@0: 		ap++;
sl@0: 		rp[j]=bn_mul_add_words(rp,ap,j,ap[-1]);
sl@0: 		rp+=2;
sl@0: 		}
sl@0: 
sl@0: 	bn_add_words(r,r,r,max);
sl@0: 
sl@0: 	/* There will not be a carry */
sl@0: 
sl@0: 	bn_sqr_words(tmp,a,n);
sl@0: 
sl@0: 	bn_add_words(r,r,tmp,max);
sl@0: 	}
sl@0: 
sl@0: #ifdef BN_RECURSION
sl@0: /* r is 2*n words in size,
sl@0:  * a and b are both n words in size.    (There's not actually a 'b' here ...)
sl@0:  * n must be a power of 2.
sl@0:  * We multiply and return the result.
sl@0:  * t must be 2*n words in size
sl@0:  * We calculate
sl@0:  * a[0]*b[0]
sl@0:  * a[0]*b[0]+a[1]*b[1]+(a[0]-a[1])*(b[1]-b[0])
sl@0:  * a[1]*b[1]
sl@0:  */
sl@0: EXPORT_C void bn_sqr_recursive(BN_ULONG *r, const BN_ULONG *a, int n2, BN_ULONG *t)
sl@0: 	{
sl@0: 	int n=n2/2;
sl@0: 	int zero,c1;
sl@0: 	BN_ULONG ln,lo,*p;
sl@0: 
sl@0: #ifdef BN_COUNT
sl@0: 	fprintf(stderr," bn_sqr_recursive %d * %d\n",n2,n2);
sl@0: #endif
sl@0: 	if (n2 == 4)
sl@0: 		{
sl@0: #ifndef BN_SQR_COMBA
sl@0: 		bn_sqr_normal(r,a,4,t);
sl@0: #else
sl@0: 		bn_sqr_comba4(r,a);
sl@0: #endif
sl@0: 		return;
sl@0: 		}
sl@0: 	else if (n2 == 8)
sl@0: 		{
sl@0: #ifndef BN_SQR_COMBA
sl@0: 		bn_sqr_normal(r,a,8,t);
sl@0: #else
sl@0: 		bn_sqr_comba8(r,a);
sl@0: #endif
sl@0: 		return;
sl@0: 		}
sl@0: 	if (n2 < BN_SQR_RECURSIVE_SIZE_NORMAL)
sl@0: 		{
sl@0: 		bn_sqr_normal(r,a,n2,t);
sl@0: 		return;
sl@0: 		}
sl@0: 	/* r=(a[0]-a[1])*(a[1]-a[0]) */
sl@0: 	c1=bn_cmp_words(a,&(a[n]),n);
sl@0: 	zero=0;
sl@0: 	if (c1 > 0)
sl@0: 		bn_sub_words(t,a,&(a[n]),n);
sl@0: 	else if (c1 < 0)
sl@0: 		bn_sub_words(t,&(a[n]),a,n);
sl@0: 	else
sl@0: 		zero=1;
sl@0: 
sl@0: 	/* The result will always be negative unless it is zero */
sl@0: 	p= &(t[n2*2]);
sl@0: 
sl@0: 	if (!zero)
sl@0: 		bn_sqr_recursive(&(t[n2]),t,n,p);
sl@0: 	else
sl@0: 		memset(&(t[n2]),0,n2*sizeof(BN_ULONG));
sl@0: 	bn_sqr_recursive(r,a,n,p);
sl@0: 	bn_sqr_recursive(&(r[n2]),&(a[n]),n,p);
sl@0: 
sl@0: 	/* t[32] holds (a[0]-a[1])*(a[1]-a[0]), it is negative or zero
sl@0: 	 * r[10] holds (a[0]*b[0])
sl@0: 	 * r[32] holds (b[1]*b[1])
sl@0: 	 */
sl@0: 
sl@0: 	c1=(int)(bn_add_words(t,r,&(r[n2]),n2));
sl@0: 
sl@0: 	/* t[32] is negative */
sl@0: 	c1-=(int)(bn_sub_words(&(t[n2]),t,&(t[n2]),n2));
sl@0: 
sl@0: 	/* t[32] holds (a[0]-a[1])*(a[1]-a[0])+(a[0]*a[0])+(a[1]*a[1])
sl@0: 	 * r[10] holds (a[0]*a[0])
sl@0: 	 * r[32] holds (a[1]*a[1])
sl@0: 	 * c1 holds the carry bits
sl@0: 	 */
sl@0: 	c1+=(int)(bn_add_words(&(r[n]),&(r[n]),&(t[n2]),n2));
sl@0: 	if (c1)
sl@0: 		{
sl@0: 		p= &(r[n+n2]);
sl@0: 		lo= *p;
sl@0: 		ln=(lo+c1)&BN_MASK2;
sl@0: 		*p=ln;
sl@0: 
sl@0: 		/* The overflow will stop before we over write
sl@0: 		 * words we should not overwrite */
sl@0: 		if (ln < (BN_ULONG)c1)
sl@0: 			{
sl@0: 			do	{
sl@0: 				p++;
sl@0: 				lo= *p;
sl@0: 				ln=(lo+1)&BN_MASK2;
sl@0: 				*p=ln;
sl@0: 				} while (ln == 0);
sl@0: 			}
sl@0: 		}
sl@0: 	}
sl@0: #endif