2 * Copyright 1995-2017 The OpenSSL Project Authors. All Rights Reserved.
4 * Licensed under the OpenSSL license (the "License"). You may not use
5 * this file except in compliance with the License. You can obtain a copy
6 * in the file LICENSE in the source distribution or at
7 * https://www.openssl.org/source/license.html
10 #include "internal/cryptlib.h"
15 * I've just gone over this and it is now %20 faster on x86 - eay - 27 Jun 96
17 int BN_sqr(BIGNUM *r, const BIGNUM *a, BN_CTX *ctx)
33 rr = (a != r) ? r : BN_CTX_get(ctx);
34 tmp = BN_CTX_get(ctx);
35 if (rr == NULL || tmp == NULL)
38 max = 2 * al; /* Non-zero (from above) */
39 if (bn_wexpand(rr, max) == NULL)
45 bn_sqr_normal(rr->d, a->d, 4, t);
47 bn_sqr_comba4(rr->d, a->d);
52 bn_sqr_normal(rr->d, a->d, 8, t);
54 bn_sqr_comba8(rr->d, a->d);
57 #if defined(BN_RECURSION)
58 if (al < BN_SQR_RECURSIVE_SIZE_NORMAL) {
59 BN_ULONG t[BN_SQR_RECURSIVE_SIZE_NORMAL * 2];
60 bn_sqr_normal(rr->d, a->d, al, t);
64 j = BN_num_bits_word((BN_ULONG)al);
68 if (bn_wexpand(tmp, k * 2) == NULL)
70 bn_sqr_recursive(rr->d, a->d, al, tmp->d);
72 if (bn_wexpand(tmp, max) == NULL)
74 bn_sqr_normal(rr->d, a->d, al, tmp->d);
78 if (bn_wexpand(tmp, max) == NULL)
80 bn_sqr_normal(rr->d, a->d, al, tmp->d);
87 if (r != rr && BN_copy(r, rr) == NULL)
98 /* tmp must have 2*n words */
99 void bn_sqr_normal(BN_ULONG *r, const BN_ULONG *a, int n, BN_ULONG *tmp)
108 rp[0] = rp[max - 1] = 0;
114 rp[j] = bn_mul_words(rp, ap, j, ap[-1]);
118 for (i = n - 2; i > 0; i--) {
121 rp[j] = bn_mul_add_words(rp, ap, j, ap[-1]);
125 bn_add_words(r, r, r, max);
127 /* There will not be a carry */
129 bn_sqr_words(tmp, a, n);
131 bn_add_words(r, r, tmp, max);
136 * r is 2*n words in size,
137 * a and b are both n words in size. (There's not actually a 'b' here ...)
138 * n must be a power of 2.
139 * We multiply and return the result.
140 * t must be 2*n words in size
143 * a[0]*b[0]+a[1]*b[1]+(a[0]-a[1])*(b[1]-b[0])
146 void bn_sqr_recursive(BN_ULONG *r, const BN_ULONG *a, int n2, BN_ULONG *t)
153 # ifndef BN_SQR_COMBA
154 bn_sqr_normal(r, a, 4, t);
159 } else if (n2 == 8) {
160 # ifndef BN_SQR_COMBA
161 bn_sqr_normal(r, a, 8, t);
167 if (n2 < BN_SQR_RECURSIVE_SIZE_NORMAL) {
168 bn_sqr_normal(r, a, n2, t);
171 /* r=(a[0]-a[1])*(a[1]-a[0]) */
172 c1 = bn_cmp_words(a, &(a[n]), n);
175 bn_sub_words(t, a, &(a[n]), n);
177 bn_sub_words(t, &(a[n]), a, n);
181 /* The result will always be negative unless it is zero */
185 bn_sqr_recursive(&(t[n2]), t, n, p);
187 memset(&t[n2], 0, sizeof(*t) * n2);
188 bn_sqr_recursive(r, a, n, p);
189 bn_sqr_recursive(&(r[n2]), &(a[n]), n, p);
192 * t[32] holds (a[0]-a[1])*(a[1]-a[0]), it is negative or zero
193 * r[10] holds (a[0]*b[0])
194 * r[32] holds (b[1]*b[1])
197 c1 = (int)(bn_add_words(t, r, &(r[n2]), n2));
199 /* t[32] is negative */
200 c1 -= (int)(bn_sub_words(&(t[n2]), t, &(t[n2]), n2));
203 * t[32] holds (a[0]-a[1])*(a[1]-a[0])+(a[0]*a[0])+(a[1]*a[1])
204 * r[10] holds (a[0]*a[0])
205 * r[32] holds (a[1]*a[1])
206 * c1 holds the carry bits
208 c1 += (int)(bn_add_words(&(r[n]), &(r[n]), &(t[n2]), n2));
212 ln = (lo + c1) & BN_MASK2;
216 * The overflow will stop before we over write words we should not
219 if (ln < (BN_ULONG)c1) {
223 ln = (lo + 1) & BN_MASK2;