bn_sqr.c revision 280304
11556Srgrimes/* crypto/bn/bn_sqr.c */
21556Srgrimes/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
31556Srgrimes * All rights reserved.
41556Srgrimes *
51556Srgrimes * This package is an SSL implementation written
61556Srgrimes * by Eric Young (eay@cryptsoft.com).
71556Srgrimes * The implementation was written so as to conform with Netscapes SSL.
81556Srgrimes *
91556Srgrimes * This library is free for commercial and non-commercial use as long as
101556Srgrimes * the following conditions are aheared to.  The following conditions
111556Srgrimes * apply to all code found in this distribution, be it the RC4, RSA,
121556Srgrimes * lhash, DES, etc., code; not just the SSL code.  The SSL documentation
131556Srgrimes * included with this distribution is covered by the same copyright terms
141556Srgrimes * except that the holder is Tim Hudson (tjh@cryptsoft.com).
151556Srgrimes *
161556Srgrimes * Copyright remains Eric Young's, and as such any Copyright notices in
171556Srgrimes * the code are not to be removed.
181556Srgrimes * If this package is used in a product, Eric Young should be given attribution
191556Srgrimes * as the author of the parts of the library used.
201556Srgrimes * This can be in the form of a textual message at program startup or
211556Srgrimes * in documentation (online or textual) provided with the package.
221556Srgrimes *
231556Srgrimes * Redistribution and use in source and binary forms, with or without
241556Srgrimes * modification, are permitted provided that the following conditions
251556Srgrimes * are met:
261556Srgrimes * 1. Redistributions of source code must retain the copyright
271556Srgrimes *    notice, this list of conditions and the following disclaimer.
281556Srgrimes * 2. Redistributions in binary form must reproduce the above copyright
291556Srgrimes *    notice, this list of conditions and the following disclaimer in the
301556Srgrimes *    documentation and/or other materials provided with the distribution.
311556Srgrimes * 3. All advertising materials mentioning features or use of this software
321556Srgrimes *    must display the following acknowledgement:
331556Srgrimes *    "This product includes cryptographic software written by
341556Srgrimes *     Eric Young (eay@cryptsoft.com)"
351556Srgrimes *    The word 'cryptographic' can be left out if the rouines from the library
361556Srgrimes *    being used are not cryptographic related :-).
371556Srgrimes * 4. If you include any Windows specific code (or a derivative thereof) from
3836150Scharnier *    the apps directory (application code) you must include an acknowledgement:
3936150Scharnier *    "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
4036150Scharnier *
4136150Scharnier * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
4250471Speter * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
431556Srgrimes * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
441556Srgrimes * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
451556Srgrimes * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
461556Srgrimes * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
471556Srgrimes * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
481556Srgrimes * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
491556Srgrimes * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
501556Srgrimes * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
511556Srgrimes * SUCH DAMAGE.
521556Srgrimes *
531556Srgrimes * The licence and distribution terms for any publically available version or
541556Srgrimes * derivative of this code cannot be changed.  i.e. this code cannot simply be
551556Srgrimes * copied and put under another distribution licence
561556Srgrimes * [including the GNU Public Licence.]
5778469Sdes */
581556Srgrimes
591556Srgrimes#include <stdio.h>
601556Srgrimes#include "cryptlib.h"
611556Srgrimes#include "bn_lcl.h"
621556Srgrimes
631556Srgrimes/* r must not be a */
641556Srgrimes/*
651556Srgrimes * I've just gone over this and it is now %20 faster on x86 - eay - 27 Jun 96
661556Srgrimes */
671556Srgrimesint BN_sqr(BIGNUM *r, const BIGNUM *a, BN_CTX *ctx)
681556Srgrimes{
691556Srgrimes    int max, al;
701556Srgrimes    int ret = 0;
711556Srgrimes    BIGNUM *tmp, *rr;
721556Srgrimes
731556Srgrimes#ifdef BN_COUNT
741556Srgrimes    fprintf(stderr, "BN_sqr %d * %d\n", a->top, a->top);
7517987Speter#endif
7617987Speter    bn_check_top(a);
7717987Speter
7825222Ssteve    al = a->top;
791556Srgrimes    if (al <= 0) {
801556Srgrimes        r->top = 0;
8125222Ssteve        r->neg = 0;
821556Srgrimes        return 1;
831556Srgrimes    }
841556Srgrimes
851556Srgrimes    BN_CTX_start(ctx);
861556Srgrimes    rr = (a != r) ? r : BN_CTX_get(ctx);
871556Srgrimes    tmp = BN_CTX_get(ctx);
881556Srgrimes    if (!rr || !tmp)
891556Srgrimes        goto err;
901556Srgrimes
911556Srgrimes    max = 2 * al;               /* Non-zero (from above) */
921556Srgrimes    if (bn_wexpand(rr, max) == NULL)
931556Srgrimes        goto err;
941556Srgrimes
951556Srgrimes    if (al == 4) {
961556Srgrimes#ifndef BN_SQR_COMBA
971556Srgrimes        BN_ULONG t[8];
981556Srgrimes        bn_sqr_normal(rr->d, a->d, 4, t);
991556Srgrimes#else
1001556Srgrimes        bn_sqr_comba4(rr->d, a->d);
1011556Srgrimes#endif
1021556Srgrimes    } else if (al == 8) {
1031556Srgrimes#ifndef BN_SQR_COMBA
1041556Srgrimes        BN_ULONG t[16];
1051556Srgrimes        bn_sqr_normal(rr->d, a->d, 8, t);
1061556Srgrimes#else
1071556Srgrimes        bn_sqr_comba8(rr->d, a->d);
1081556Srgrimes#endif
1091556Srgrimes    } else {
1101556Srgrimes#if defined(BN_RECURSION)
1111556Srgrimes        if (al < BN_SQR_RECURSIVE_SIZE_NORMAL) {
1121556Srgrimes            BN_ULONG t[BN_SQR_RECURSIVE_SIZE_NORMAL * 2];
1131556Srgrimes            bn_sqr_normal(rr->d, a->d, al, t);
1141556Srgrimes        } else {
1151556Srgrimes            int j, k;
1161556Srgrimes
1171556Srgrimes            j = BN_num_bits_word((BN_ULONG)al);
1181556Srgrimes            j = 1 << (j - 1);
1191556Srgrimes            k = j + j;
1201556Srgrimes            if (al == j) {
1211556Srgrimes                if (bn_wexpand(tmp, k * 2) == NULL)
122                    goto err;
123                bn_sqr_recursive(rr->d, a->d, al, tmp->d);
124            } else {
125                if (bn_wexpand(tmp, max) == NULL)
126                    goto err;
127                bn_sqr_normal(rr->d, a->d, al, tmp->d);
128            }
129        }
130#else
131        if (bn_wexpand(tmp, max) == NULL)
132            goto err;
133        bn_sqr_normal(rr->d, a->d, al, tmp->d);
134#endif
135    }
136
137    rr->neg = 0;
138    /*
139     * If the most-significant half of the top word of 'a' is zero, then the
140     * square of 'a' will max-1 words.
141     */
142    if (a->d[al - 1] == (a->d[al - 1] & BN_MASK2l))
143        rr->top = max - 1;
144    else
145        rr->top = max;
146    if (rr != r)
147        BN_copy(r, rr);
148    ret = 1;
149 err:
150    bn_check_top(rr);
151    bn_check_top(tmp);
152    BN_CTX_end(ctx);
153    return (ret);
154}
155
156/* tmp must have 2*n words */
157void bn_sqr_normal(BN_ULONG *r, const BN_ULONG *a, int n, BN_ULONG *tmp)
158{
159    int i, j, max;
160    const BN_ULONG *ap;
161    BN_ULONG *rp;
162
163    max = n * 2;
164    ap = a;
165    rp = r;
166    rp[0] = rp[max - 1] = 0;
167    rp++;
168    j = n;
169
170    if (--j > 0) {
171        ap++;
172        rp[j] = bn_mul_words(rp, ap, j, ap[-1]);
173        rp += 2;
174    }
175
176    for (i = n - 2; i > 0; i--) {
177        j--;
178        ap++;
179        rp[j] = bn_mul_add_words(rp, ap, j, ap[-1]);
180        rp += 2;
181    }
182
183    bn_add_words(r, r, r, max);
184
185    /* There will not be a carry */
186
187    bn_sqr_words(tmp, a, n);
188
189    bn_add_words(r, r, tmp, max);
190}
191
192#ifdef BN_RECURSION
193/*-
194 * r is 2*n words in size,
195 * a and b are both n words in size.    (There's not actually a 'b' here ...)
196 * n must be a power of 2.
197 * We multiply and return the result.
198 * t must be 2*n words in size
199 * We calculate
200 * a[0]*b[0]
201 * a[0]*b[0]+a[1]*b[1]+(a[0]-a[1])*(b[1]-b[0])
202 * a[1]*b[1]
203 */
204void bn_sqr_recursive(BN_ULONG *r, const BN_ULONG *a, int n2, BN_ULONG *t)
205{
206    int n = n2 / 2;
207    int zero, c1;
208    BN_ULONG ln, lo, *p;
209
210# ifdef BN_COUNT
211    fprintf(stderr, " bn_sqr_recursive %d * %d\n", n2, n2);
212# endif
213    if (n2 == 4) {
214# ifndef BN_SQR_COMBA
215        bn_sqr_normal(r, a, 4, t);
216# else
217        bn_sqr_comba4(r, a);
218# endif
219        return;
220    } else if (n2 == 8) {
221# ifndef BN_SQR_COMBA
222        bn_sqr_normal(r, a, 8, t);
223# else
224        bn_sqr_comba8(r, a);
225# endif
226        return;
227    }
228    if (n2 < BN_SQR_RECURSIVE_SIZE_NORMAL) {
229        bn_sqr_normal(r, a, n2, t);
230        return;
231    }
232    /* r=(a[0]-a[1])*(a[1]-a[0]) */
233    c1 = bn_cmp_words(a, &(a[n]), n);
234    zero = 0;
235    if (c1 > 0)
236        bn_sub_words(t, a, &(a[n]), n);
237    else if (c1 < 0)
238        bn_sub_words(t, &(a[n]), a, n);
239    else
240        zero = 1;
241
242    /* The result will always be negative unless it is zero */
243    p = &(t[n2 * 2]);
244
245    if (!zero)
246        bn_sqr_recursive(&(t[n2]), t, n, p);
247    else
248        memset(&(t[n2]), 0, n2 * sizeof(BN_ULONG));
249    bn_sqr_recursive(r, a, n, p);
250    bn_sqr_recursive(&(r[n2]), &(a[n]), n, p);
251
252    /*-
253     * t[32] holds (a[0]-a[1])*(a[1]-a[0]), it is negative or zero
254     * r[10] holds (a[0]*b[0])
255     * r[32] holds (b[1]*b[1])
256     */
257
258    c1 = (int)(bn_add_words(t, r, &(r[n2]), n2));
259
260    /* t[32] is negative */
261    c1 -= (int)(bn_sub_words(&(t[n2]), t, &(t[n2]), n2));
262
263    /*-
264     * t[32] holds (a[0]-a[1])*(a[1]-a[0])+(a[0]*a[0])+(a[1]*a[1])
265     * r[10] holds (a[0]*a[0])
266     * r[32] holds (a[1]*a[1])
267     * c1 holds the carry bits
268     */
269    c1 += (int)(bn_add_words(&(r[n]), &(r[n]), &(t[n2]), n2));
270    if (c1) {
271        p = &(r[n + n2]);
272        lo = *p;
273        ln = (lo + c1) & BN_MASK2;
274        *p = ln;
275
276        /*
277         * The overflow will stop before we over write words we should not
278         * overwrite
279         */
280        if (ln < (BN_ULONG)c1) {
281            do {
282                p++;
283                lo = *p;
284                ln = (lo + 1) & BN_MASK2;
285                *p = ln;
286            } while (ln == 0);
287        }
288    }
289}
290#endif
291