/* based on gmp's mpz_import.
* see http://gmplib.org/manual/Integer-Import-and-Export.html
*/
int mp_import(mp_int* rop, size_t count, int order, size_t size,
int endian, size_t nails, const void* op) {
int result;
size_t odd_nails, nail_bytes, i, j;
unsigned char odd_nail_mask;
mp_zero(rop);
if (endian == 0) {
union {
unsigned int i;
char c[4];
} lint;
lint.i = 0x01020304;
endian = (lint.c[0] == 4) ? -1 : 1;
}
odd_nails = (nails % 8);
odd_nail_mask = 0xff;
for (i = 0; i < odd_nails; ++i) {
odd_nail_mask ^= (1 << (7 - i));
}
nail_bytes = nails / 8;
for (i = 0; i < count; ++i) {
for (j = 0; j < (size - nail_bytes); ++j) {
unsigned char byte = *(
(unsigned char*)op +
(((order == 1) ? i : ((count - 1) - i)) * size) +
((endian == 1) ? (j + nail_bytes) : (((size - 1) - j) - nail_bytes))
);
if (
(result = mp_mul_2d(rop, ((j == 0) ? (8 - odd_nails) : 8), rop)) != MP_OKAY) {
return result;
}
rop->dp[0] |= (j == 0) ? (byte & odd_nail_mask) : byte;
rop->used += 1;
}
}
mp_clamp(rop);
return MP_OKAY;
}
/* single digit division (based on routine from MPI) */
int
mp_div_d (mp_int * a, mp_digit b, mp_int * c, mp_digit * d)
{
mp_int q;
mp_word w;
mp_digit t;
int res, ix;
if (b == 0) {
return MP_VAL;
}
if (b == 3) {
return mp_div_3(a, c, d);
}
if ((res = mp_init_size(&q, a->used)) != MP_OKAY) {
return res;
}
q.used = a->used;
q.sign = a->sign;
w = 0;
for (ix = a->used - 1; ix >= 0; ix--) {
w = (w << ((mp_word)DIGIT_BIT)) | ((mp_word)a->dp[ix]);
if (w >= b) {
t = (mp_digit)(w / b);
w = w % b;
} else {
t = 0;
}
q.dp[ix] = (mp_digit)t;
}
if (d != NULL) {
*d = (mp_digit)w;
}
if (c != NULL) {
mp_clamp(&q);
mp_exch(&q, c);
}
mp_clear(&q);
return res;
}
/* b = a/2 */
int
mp_div_2 (mp_int * a, mp_int * b)
{
int x, res, oldused;
/* copy */
if (b->alloc < a->used) {
if ((res = mp_grow (b, a->used)) != MP_OKAY) {
return res;
}
}
oldused = b->used;
b->used = a->used;
{
register mp_digit r, rr, *tmpa, *tmpb;
/* source alias */
tmpa = a->dp + b->used - 1;
/* dest alias */
tmpb = b->dp + b->used - 1;
/* carry */
r = 0;
for (x = b->used - 1; x >= 0; x--) {
/* get the carry for the next iteration */
rr = *tmpa & 1;
/* shift the current digit, add in carry and store */
*tmpb-- = (*tmpa-- >> 1) | (r << (DIGIT_BIT - 1));
/* forward carry to next iteration */
r = rr;
}
/* zero excess digits */
tmpb = b->dp + b->used;
for (x = b->used; x < oldused; x++) {
*tmpb++ = 0;
}
}
b->sign = a->sign;
mp_clamp (b);
return MP_OKAY;
}
/* b = a/2 */
int mp_div_2(mp_int * a, mp_int * b)
{
int x, res, oldused;
/* copy */
if (ALLOC(b) < USED(a)) {
if ((res = mp_grow (b, USED(a))) != MP_OKAY) {
return res;
}
}
oldused = USED(b);
SET_USED(b,USED(a));
{
register mp_digit r, rr, *tmpa, *tmpb;
/* source alias */
tmpa = DIGITS(a) + USED(b) - 1;
/* dest alias */
tmpb = DIGITS(b) + USED(b) - 1;
/* carry */
r = 0;
for (x = USED(b) - 1; x >= 0; x--) {
/* get the carry for the next iteration */
rr = *tmpa & 1;
/* shift the current digit, add in carry and store */
*tmpb-- = (*tmpa-- >> 1) | (r << (DIGIT_BIT - 1));
/* forward carry to next iteration */
r = rr;
}
/* zero excess digits */
tmpb = DIGITS(b) + USED(b);
for (x = USED(b); x < oldused; x++) {
*tmpb++ = 0;
}
}
SET_SIGN(b,SIGN(a));
mp_clamp (b);
return MP_OKAY;
}
static void from_num(MVMnum64 d, mp_int *a) {
MVMnum64 d_digit = pow(2, DIGIT_BIT);
MVMnum64 da = fabs(d);
MVMnum64 upper;
MVMnum64 lower;
MVMnum64 lowest;
MVMnum64 rest;
int digits = 0;
mp_zero(a);
while (da > d_digit * d_digit * d_digit) {;
da /= d_digit;
digits++;
}
mp_grow(a, digits + 3);
/* populate the top 3 digits */
upper = da / (d_digit*d_digit);
rest = fmod(da, d_digit*d_digit);
lower = rest / d_digit;
lowest = fmod(rest,d_digit );
if (upper >= 1) {
mp_set_long(a, (unsigned long) upper);
mp_mul_2d(a, DIGIT_BIT , a);
DIGIT(a, 0) = (mp_digit) lower;
mp_mul_2d(a, DIGIT_BIT , a);
} else {
if (lower >= 1) {
mp_set_long(a, (unsigned long) lower);
mp_mul_2d(a, DIGIT_BIT , a);
a->used = 2;
} else {
a->used = 1;
}
}
DIGIT(a, 0) = (mp_digit) lowest;
/* shift the rest */
mp_mul_2d(a, DIGIT_BIT * digits, a);
if (d < 0)
mp_neg(a, a);
mp_clamp(a);
mp_shrink(a);
}
/* reads a unsigned char array, assumes the msb is stored first [big endian] */
int
mp_read_unsigned_bin (mp_int * a, unsigned char *b, int c)
{
int res;
mp_zero (a);
while (c-- > 0) {
if ((res = mp_mul_2d (a, 8, a)) != MP_OKAY) {
return res;
}
if (DIGIT_BIT != 7) {
a->dp[0] |= *b++;
a->used += 1;
} else {
a->dp[0] = (*b & MP_MASK);
a->dp[1] |= ((*b++ >> 7U) & 1);
a->used += 2;
}
}
mp_clamp (a);
return MP_OKAY;
}
/* Taken from mp_set_long, but portably accepts a 64-bit number. */
int MVM_bigint_mp_set_uint64(mp_int * a, MVMuint64 b) {
int x, res;
mp_zero (a);
/* set four bits at a time */
for (x = 0; x < sizeof(MVMuint64) * 2; x++) {
/* shift the number up four bits */
if ((res = mp_mul_2d (a, 4, a)) != MP_OKAY) {
return res;
}
/* OR in the top four bits of the source */
a->dp[0] |= (b >> ((sizeof(MVMuint64)) * 8 - 4)) & 15;
/* shift the source up to the next four bits */
b <<= 4;
/* ensure that digits are not clamped off */
a->used += 1;
}
mp_clamp(a);
return MP_OKAY;
}
/* c = |a| * |b| using Karatsuba Multiplication using
* three half size multiplications
*
* Let B represent the radix [e.g. 2**DIGIT_BIT] and
* let n represent half of the number of digits in
* the min(a,b)
*
* a = a1 * B**n + a0
* b = b1 * B**n + b0
*
* Then, a * b =>
a1b1 * B**2n + ((a1 + a0)(b1 + b0) - (a0b0 + a1b1)) * B + a0b0
*
* Note that a1b1 and a0b0 are used twice and only need to be
* computed once. So in total three half size (half # of
* digit) multiplications are performed, a0b0, a1b1 and
* (a1+b1)(a0+b0)
*
* Note that a multiplication of half the digits requires
* 1/4th the number of single precision multiplications so in
* total after one call 25% of the single precision multiplications
* are saved. Note also that the call to mp_mul can end up back
* in this function if the a0, a1, b0, or b1 are above the threshold.
* This is known as divide-and-conquer and leads to the famous
* O(N**lg(3)) or O(N**1.584) work which is asymptopically lower than
* the standard O(N**2) that the baseline/comba methods use.
* Generally though the overhead of this method doesn't pay off
* until a certain size (N ~ 80) is reached.
*/
int mp_karatsuba_mul (mp_int * a, mp_int * b, mp_int * c)
{
mp_int x0, x1, y0, y1, t1, x0y0, x1y1;
int B, err;
/* default the return code to an error */
err = MP_MEM;
/* min # of digits */
B = MIN (a->used, b->used);
/* now divide in two */
B = B >> 1;
/* init copy all the temps */
if (mp_init_size (&x0, B) != MP_OKAY)
goto ERR;
if (mp_init_size (&x1, a->used - B) != MP_OKAY)
goto X0;
if (mp_init_size (&y0, B) != MP_OKAY)
goto X1;
if (mp_init_size (&y1, b->used - B) != MP_OKAY)
goto Y0;
/* init temps */
if (mp_init_size (&t1, B * 2) != MP_OKAY)
goto Y1;
if (mp_init_size (&x0y0, B * 2) != MP_OKAY)
goto T1;
if (mp_init_size (&x1y1, B * 2) != MP_OKAY)
goto X0Y0;
/* now shift the digits */
x0.used = y0.used = B;
x1.used = a->used - B;
y1.used = b->used - B;
{
register int x;
register mp_digit *tmpa, *tmpb, *tmpx, *tmpy;
/* we copy the digits directly instead of using higher level functions
* since we also need to shift the digits
*/
tmpa = a->dp;
tmpb = b->dp;
tmpx = x0.dp;
tmpy = y0.dp;
for (x = 0; x < B; x++) {
*tmpx++ = *tmpa++;
*tmpy++ = *tmpb++;
}
tmpx = x1.dp;
for (x = B; x < a->used; x++) {
*tmpx++ = *tmpa++;
}
tmpy = y1.dp;
for (x = B; x < b->used; x++) {
*tmpy++ = *tmpb++;
}
}
/* only need to clamp the lower words since by definition the
* upper words x1/y1 must have a known number of digits
*/
mp_clamp (&x0);
mp_clamp (&y0);
/* now calc the products x0y0 and x1y1 */
/* after this x0 is no longer required, free temp [x0==t2]! */
if (mp_mul (&x0, &y0, &x0y0) != MP_OKAY)
goto X1Y1; /* x0y0 = x0*y0 */
if (mp_mul (&x1, &y1, &x1y1) != MP_OKAY)
goto X1Y1; /* x1y1 = x1*y1 */
/* now calc x1+x0 and y1+y0 */
if (s_mp_add (&x1, &x0, &t1) != MP_OKAY)
goto X1Y1; /* t1 = x1 - x0 */
if (s_mp_add (&y1, &y0, &x0) != MP_OKAY)
goto X1Y1; /* t2 = y1 - y0 */
if (mp_mul (&t1, &x0, &t1) != MP_OKAY)
goto X1Y1; /* t1 = (x1 + x0) * (y1 + y0) */
/* add x0y0 */
if (mp_add (&x0y0, &x1y1, &x0) != MP_OKAY)
goto X1Y1; /* t2 = x0y0 + x1y1 */
if (s_mp_sub (&t1, &x0, &t1) != MP_OKAY)
goto X1Y1; /* t1 = (x1+x0)*(y1+y0) - (x1y1 + x0y0) */
/* shift by B */
if (mp_lshd (&t1, B) != MP_OKAY)
goto X1Y1; /* t1 = (x0y0 + x1y1 - (x1-x0)*(y1-y0))<<B */
if (mp_lshd (&x1y1, B * 2) != MP_OKAY)
goto X1Y1; /* x1y1 = x1y1 << 2*B */
if (mp_add (&x0y0, &t1, &t1) != MP_OKAY)
goto X1Y1; /* t1 = x0y0 + t1 */
if (mp_add (&t1, &x1y1, c) != MP_OKAY)
goto X1Y1; /* t1 = x0y0 + t1 + x1y1 */
/* Algorithm succeeded set the return code to MP_OKAY */
err = MP_OKAY;
X1Y1:
mp_clear (&x1y1);
X0Y0:
mp_clear (&x0y0);
T1:
mp_clear (&t1);
Y1:
mp_clear (&y1);
Y0:
mp_clear (&y0);
X1:
mp_clear (&x1);
X0:
mp_clear (&x0);
ERR:
return err;
}
/* single digit division (based on routine from MPI) */
int mp_div_d (mp_int * a, mp_digit b, mp_int * c, mp_digit * d)
{
mp_int q;
mp_word w;
mp_digit t;
int res, ix;
/* cannot divide by zero */
if (b == 0) {
return MP_VAL;
}
/* quick outs */
if (b == 1 || mp_iszero(a) == 1) {
if (d != NULL) {
*d = 0;
}
if (c != NULL) {
return mp_copy(a, c);
}
return MP_OKAY;
}
/* power of two ? */
if (s_is_power_of_two(b, &ix) == 1) {
if (d != NULL) {
*d = a->dp[0] & ((((mp_digit)1)<<ix) - 1);
}
if (c != NULL) {
return mp_div_2d(a, ix, c, NULL);
}
return MP_OKAY;
}
#ifdef BN_MP_DIV_3_C
/* three? */
if (b == 3) {
return mp_div_3(a, c, d);
}
#endif
/* no easy answer [c'est la vie]. Just division */
if ((res = mp_init_size(&q, a->used)) != MP_OKAY) {
return res;
}
q.used = a->used;
q.sign = a->sign;
w = 0;
for (ix = a->used - 1; ix >= 0; ix--) {
w = (w << ((mp_word)DIGIT_BIT)) | ((mp_word)a->dp[ix]);
if (w >= b) {
t = (mp_digit)(w / b);
w -= ((mp_word)t) * ((mp_word)b);
} else {
t = 0;
}
q.dp[ix] = (mp_digit)t;
}
if (d != NULL) {
*d = (mp_digit)w;
}
if (c != NULL) {
mp_clamp(&q);
mp_exch(&q, c);
}
mp_clear(&q);
return res;
}
/* single digit addition */
int
mp_add_d (mp_int * a, mp_digit b, mp_int * c)
{
int res, ix, oldused;
mp_digit *tmpa, *tmpc, mu;
/* grow c as required */
if (c->alloc < a->used + 1) {
if ((res = mp_grow(c, a->used + 1)) != MP_OKAY) {
return res;
}
}
/* if a is negative and |a| >= b, call c = |a| - b */
if (a->sign == MP_NEG && (a->used > 1 || a->dp[0] >= b)) {
/* temporarily fix sign of a */
a->sign = MP_ZPOS;
/* c = |a| - b */
res = mp_sub_d(a, b, c);
/* fix sign */
a->sign = c->sign = MP_NEG;
/* clamp */
mp_clamp(c);
return res;
}
/* old number of used digits in c */
oldused = c->used;
/* sign always positive */
c->sign = MP_ZPOS;
/* source alias */
tmpa = a->dp;
/* destination alias */
tmpc = c->dp;
/* if a is positive */
if (a->sign == MP_ZPOS) {
/* add digit, after this we're propagating
* the carry.
*/
*tmpc = *tmpa++ + b;
mu = *tmpc >> DIGIT_BIT;
*tmpc++ &= MP_MASK;
/* now handle rest of the digits */
for (ix = 1; ix < a->used; ix++) {
*tmpc = *tmpa++ + mu;
mu = *tmpc >> DIGIT_BIT;
*tmpc++ &= MP_MASK;
}
/* set final carry */
ix++;
*tmpc++ = mu;
/* setup size */
c->used = a->used + 1;
} else {
/* integer signed division.
* c*b + d == a [e.g. a/b, c=quotient, d=remainder]
* HAC pp.598 Algorithm 14.20
*
* Note that the description in HAC is horribly
* incomplete. For example, it doesn't consider
* the case where digits are removed from 'x' in
* the inner loop. It also doesn't consider the
* case that y has fewer than three digits, etc..
*
* The overall algorithm is as described as
* 14.20 from HAC but fixed to treat these cases.
*/
int mp_div MPA(mp_int * a, mp_int * b, mp_int * c, mp_int * d)
{
mp_int q, x, y, t1, t2;
int res, n, t, i, norm, neg;
/* is divisor zero ? */
if (mp_iszero (b) == 1) {
return MP_VAL;
}
/* if a < b then q=0, r = a */
if (mp_cmp_mag (a, b) == MP_LT) {
if (d != NULL) {
res = mp_copy (MPST, a, d);
} else {
res = MP_OKAY;
}
if (c != NULL) {
mp_zero (c);
}
return res;
}
if ((res = mp_init_size (&q, a->used + 2)) != MP_OKAY) {
return res;
}
q.used = a->used + 2;
if ((res = mp_init (&t1)) != MP_OKAY) {
goto LBL_Q;
}
if ((res = mp_init (&t2)) != MP_OKAY) {
goto LBL_T1;
}
if ((res = mp_init_copy (MPST, &x, a)) != MP_OKAY) {
goto LBL_T2;
}
if ((res = mp_init_copy (MPST, &y, b)) != MP_OKAY) {
goto LBL_X;
}
/* fix the sign */
neg = (a->sign == b->sign) ? MP_ZPOS : MP_NEG;
x.sign = y.sign = MP_ZPOS;
/* normalize both x and y, ensure that y >= b/2, [b == 2**DIGIT_BIT] */
norm = mp_count_bits(&y) % DIGIT_BIT;
if (norm < (int)(DIGIT_BIT-1)) {
norm = (DIGIT_BIT-1) - norm;
if ((res = mp_mul_2d (MPST, &x, norm, &x)) != MP_OKAY) {
goto LBL_Y;
}
if ((res = mp_mul_2d (MPST, &y, norm, &y)) != MP_OKAY) {
goto LBL_Y;
}
} else {
norm = 0;
}
/* note hac does 0 based, so if used==5 then its 0,1,2,3,4, e.g. use 4 */
n = x.used - 1;
t = y.used - 1;
/* while (x >= y*b**n-t) do { q[n-t] += 1; x -= y*b**{n-t} } */
if ((res = mp_lshd (MPST, &y, n - t)) != MP_OKAY) { /* y = y*b**{n-t} */
goto LBL_Y;
}
while (mp_cmp (&x, &y) != MP_LT) {
++(q.dp[n - t]);
if ((res = mp_sub (MPST, &x, &y, &x)) != MP_OKAY) {
goto LBL_Y;
}
}
/* reset y by shifting it back down */
mp_rshd (&y, n - t);
/* step 3. for i from n down to (t + 1) */
for (i = n; i >= (t + 1); i--) {
if (i > x.used) {
continue;
}
/* step 3.1 if xi == yt then set q{i-t-1} to b-1,
* otherwise set q{i-t-1} to (xi*b + x{i-1})/yt */
if (x.dp[i] == y.dp[t]) {
q.dp[i - t - 1] = ((((mp_digit)1) << DIGIT_BIT) - 1);
} else {
mp_word tmp;
tmp = ((mp_word) x.dp[i]) << ((mp_word) DIGIT_BIT);
tmp |= ((mp_word) x.dp[i - 1]);
tmp /= ((mp_word) y.dp[t]);
if (tmp > (mp_word) MP_MASK)
tmp = MP_MASK;
q.dp[i - t - 1] = (mp_digit) (tmp & (mp_word) (MP_MASK));
}
/* while (q{i-t-1} * (yt * b + y{t-1})) >
xi * b**2 + xi-1 * b + xi-2
do q{i-t-1} -= 1;
*/
q.dp[i - t - 1] = (q.dp[i - t - 1] + 1) & MP_MASK;
do {
q.dp[i - t - 1] = (q.dp[i - t - 1] - 1) & MP_MASK;
/* find left hand */
mp_zero (&t1);
t1.dp[0] = (t - 1 < 0) ? 0 : y.dp[t - 1];
t1.dp[1] = y.dp[t];
t1.used = 2;
if ((res = mp_mul_d (MPST, &t1, q.dp[i - t - 1], &t1)) != MP_OKAY) {
goto LBL_Y;
}
/* find right hand */
t2.dp[0] = (i - 2 < 0) ? 0 : x.dp[i - 2];
t2.dp[1] = (i - 1 < 0) ? 0 : x.dp[i - 1];
t2.dp[2] = x.dp[i];
t2.used = 3;
} while (mp_cmp_mag(&t1, &t2) == MP_GT);
/* step 3.3 x = x - q{i-t-1} * y * b**{i-t-1} */
if ((res = mp_mul_d (MPST, &y, q.dp[i - t - 1], &t1)) != MP_OKAY) {
goto LBL_Y;
}
if ((res = mp_lshd (MPST, &t1, i - t - 1)) != MP_OKAY) {
goto LBL_Y;
}
if ((res = mp_sub (MPST, &x, &t1, &x)) != MP_OKAY) {
goto LBL_Y;
}
/* if x < 0 then { x = x + y*b**{i-t-1}; q{i-t-1} -= 1; } */
if (x.sign == MP_NEG) {
if ((res = mp_copy (MPST, &y, &t1)) != MP_OKAY) {
goto LBL_Y;
}
if ((res = mp_lshd (MPST, &t1, i - t - 1)) != MP_OKAY) {
goto LBL_Y;
}
if ((res = mp_add (MPST, &x, &t1, &x)) != MP_OKAY) {
goto LBL_Y;
}
q.dp[i - t - 1] = (q.dp[i - t - 1] - 1UL) & MP_MASK;
}
}
/* now q is the quotient and x is the remainder
* [which we have to normalize]
*/
/* get sign before writing to c */
x.sign = x.used == 0 ? MP_ZPOS : a->sign;
if (c != NULL) {
mp_clamp (&q);
mp_managed_copy (MPST, &q, c);
c->sign = neg;
}
if (d != NULL) {
mp_div_2d (MPST, &x, norm, &x, NULL);
mp_managed_copy (MPST, &x, d);
}
res = MP_OKAY;
LBL_Y:mp_clear (&y);
LBL_X:mp_clear (&x);
LBL_T2:mp_clear (&t2);
LBL_T1:mp_clear (&t1);
LBL_Q:mp_clear (&q);
return res;
}
/* computes xR**-1 == x (mod N) via Montgomery Reduction
*
* This is an optimized implementation of montgomery_reduce
* which uses the comba method to quickly calculate the columns of the
* reduction.
*
* Based on Algorithm 14.32 on pp.601 of HAC.
*/
int fast_mp_montgomery_reduce (mp_int * x, mp_int * n, mp_digit rho)
{
int ix, res, olduse;
mp_word W[MP_WARRAY] = { 0 };
/* get old used count */
olduse = x->used;
/* grow a as required */
if (x->alloc < n->used + 1) {
if ((res = mp_grow (x, n->used + 1)) != MP_OKAY) {
return res;
}
}
/* first we have to get the digits of the input into
* an array of double precision words W[...]
*/
{
register mp_word *_W;
register mp_digit *tmpx;
/* alias for the W[] array */
_W = W;
/* alias for the digits of x*/
tmpx = x->dp;
/* copy the digits of a into W[0..a->used-1] */
for (ix = 0; ix < x->used; ix++) {
*_W++ = *tmpx++;
}
/* zero the high words of W[a->used..m->used*2] */
for (; ix < n->used * 2 + 1; ix++) {
*_W++ = 0;
}
}
/* now we proceed to zero successive digits
* from the least significant upwards
*/
for (ix = 0; ix < n->used; ix++) {
/* mu = ai * m' mod b
*
* We avoid a double precision multiplication (which isn't required)
* by casting the value down to a mp_digit. Note this requires
* that W[ix-1] have the carry cleared (see after the inner loop)
*/
register mp_digit mu;
mu = (mp_digit) (((W[ix] & MP_MASK) * rho) & MP_MASK);
/* a = a + mu * m * b**i
*
* This is computed in place and on the fly. The multiplication
* by b**i is handled by offseting which columns the results
* are added to.
*
* Note the comba method normally doesn't handle carries in the
* inner loop In this case we fix the carry from the previous
* column since the Montgomery reduction requires digits of the
* result (so far) [see above] to work. This is
* handled by fixing up one carry after the inner loop. The
* carry fixups are done in order so after these loops the
* first m->used words of W[] have the carries fixed
*/
{
register int iy;
register mp_digit *tmpn;
register mp_word *_W;
/* alias for the digits of the modulus */
tmpn = n->dp;
/* Alias for the columns set by an offset of ix */
_W = W + ix;
/* inner loop */
for (iy = 0; iy < n->used; iy++) {
*_W++ += ((mp_word)mu) * ((mp_word)*tmpn++);
}
}
/* now fix carry for next digit, W[ix+1] */
W[ix + 1] += W[ix] >> ((mp_word) DIGIT_BIT);
}
/* now we have to propagate the carries and
* shift the words downward [all those least
* significant digits we zeroed].
*/
{
register mp_digit *tmpx;
register mp_word *_W, *_W1;
/* nox fix rest of carries */
/* alias for current word */
_W1 = W + ix;
/* alias for next word, where the carry goes */
_W = W + ++ix;
for (; ix <= n->used * 2 + 1; ix++) {
*_W++ += *_W1++ >> ((mp_word) DIGIT_BIT);
}
/* copy out, A = A/b**n
*
* The result is A/b**n but instead of converting from an
* array of mp_word to mp_digit than calling mp_rshd
* we just copy them in the right order
*/
/* alias for destination word */
tmpx = x->dp;
/* alias for shifted double precision result */
_W = W + n->used;
for (ix = 0; ix < n->used + 1; ix++) {
*tmpx++ = (mp_digit)(*_W++ & ((mp_word) MP_MASK));
}
/* zero oldused digits, if the input a was larger than
* m->used+1 we'll have to clear the digits
*/
for (; ix < olduse; ix++) {
*tmpx++ = 0;
}
}
/* set the max used and clamp */
x->used = n->used + 1;
mp_clamp (x);
/* if A >= m then A = A - m */
if (mp_cmp_mag (x, n) != MP_LT) {
return s_mp_sub (x, n, x);
}
return MP_OKAY;
}
/* low level squaring, b = a*a, HAC pp.596-597, Algorithm 14.16 */
int s_mp_sqr (mp_int * a, mp_int * b)
{
mp_int t;
int res, ix, iy, pa;
mp_word r;
mp_digit u, tmpx, *tmpt;
pa = a->used;
if ((res = mp_init_size (&t, 2*pa + 1)) != MP_OKAY) {
return res;
}
/* default used is maximum possible size */
t.used = 2*pa + 1;
for (ix = 0; ix < pa; ix++) {
/* first calculate the digit at 2*ix */
/* calculate double precision result */
r = ((mp_word) t.dp[2*ix]) +
((mp_word)a->dp[ix])*((mp_word)a->dp[ix]);
/* store lower part in result */
t.dp[ix+ix] = (mp_digit) (r & ((mp_word) MP_MASK));
/* get the carry */
u = (mp_digit)(r >> ((mp_word) DIGIT_BIT));
/* left hand side of A[ix] * A[iy] */
tmpx = a->dp[ix];
/* alias for where to store the results */
tmpt = t.dp + (2*ix + 1);
for (iy = ix + 1; iy < pa; iy++) {
/* first calculate the product */
r = ((mp_word)tmpx) * ((mp_word)a->dp[iy]);
/* now calculate the double precision result, note we use
* addition instead of *2 since it's easier to optimize
*/
r = ((mp_word) *tmpt) + r + r + ((mp_word) u);
/* store lower part */
*tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK));
/* get carry */
u = (mp_digit)(r >> ((mp_word) DIGIT_BIT));
}
/* propagate upwards */
while (u != ((mp_digit) 0)) {
r = ((mp_word) *tmpt) + ((mp_word) u);
*tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK));
u = (mp_digit)(r >> ((mp_word) DIGIT_BIT));
}
}
mp_clamp (&t);
mp_exch (&t, b);
mp_clear (&t);
return MP_OKAY;
}
/* low level addition, based on HAC pp.594, Algorithm 14.7 */
int
s_mp_add (mp_int * a, mp_int * b, mp_int * c)
{
mp_int *x;
int olduse, res, min, max;
/* find sizes, we let |a| <= |b| which means we have to sort
* them. "x" will point to the input with the most digits
*/
if (a->used > b->used) {
min = b->used;
max = a->used;
x = a;
} else {
min = a->used;
max = b->used;
x = b;
}
/* init result */
if (c->alloc < max + 1) {
if ((res = mp_grow (c, max + 1)) != MP_OKAY) {
return res;
}
}
/* get old used digit count and set new one */
olduse = c->used;
c->used = max + 1;
{
register mp_digit u, *tmpa, *tmpb, *tmpc;
register int i;
/* alias for digit pointers */
/* first input */
tmpa = a->dp;
/* second input */
tmpb = b->dp;
/* destination */
tmpc = c->dp;
/* zero the carry */
u = 0;
for (i = 0; i < min; i++) {
/* Compute the sum at one digit, T[i] = A[i] + B[i] + U */
*tmpc = *tmpa++ + *tmpb++ + u;
/* U = carry bit of T[i] */
u = *tmpc >> ((mp_digit)DIGIT_BIT);
/* take away carry bit from T[i] */
*tmpc++ &= MP_MASK;
}
/* now copy higher words if any, that is in A+B
* if A or B has more digits add those in
*/
if (min != max) {
for (; i < max; i++) {
/* T[i] = X[i] + U */
*tmpc = x->dp[i] + u;
/* U = carry bit of T[i] */
u = *tmpc >> ((mp_digit)DIGIT_BIT);
/* take away carry bit from T[i] */
*tmpc++ &= MP_MASK;
}
}
/* add carry */
*tmpc++ = u;
/* clear digits above oldused */
for (i = c->used; i < olduse; i++) {
*tmpc++ = 0;
}
}
mp_clamp (c);
return MP_OKAY;
}
int mp_toom_cook_5_mul(mp_int *a, mp_int *b, mp_int *c)
{
mp_int w1, w2, w3, w4, w5, w6, w7, w8, w9;
mp_int tmp1, tmp2;
mp_int a0, a1, a2, a3, a4;
mp_int b0, b1, b2, b3, b4;
int e = MP_OKAY;
int B, count, sign;
B = (MAX(a->used, b->used)) / 5;
sign = (a->sign != b->sign) ? MP_NEG : MP_ZPOS;
if (MIN(a->used, b->used) < TOOM_COOK_5_MUL_CO) {
if ((e = mp_mul(a, b, c)) != MP_OKAY) {
return e;
}
c->sign = sign;
return MP_OKAY;
}
if ((e =
mp_init_multi(&w1, &w2, &w3, &w4, &w5, &w6, &w7, &w8, &w9, &tmp1, &tmp2,
//&a0, &a1, &a2, &a3, &a4, &b0, &b1, &b2, &b3, &b4,
NULL)) != MP_OKAY) {
goto ERR0;
//goto ERR;
}
if ((e = mp_init_size(&a0, B)) != MP_OKAY) {
goto ERRa0;
}
if ((e = mp_init_size(&a1, B)) != MP_OKAY) {
goto ERRa1;
}
if ((e = mp_init_size(&a2, B)) != MP_OKAY) {
goto ERRa2;
}
if ((e = mp_init_size(&a3, B)) != MP_OKAY) {
goto ERRa3;
}
if ((e = mp_init_size(&a4, B)) != MP_OKAY) {
goto ERRa4;
}
if ((e = mp_init_size(&b0, B)) != MP_OKAY) {
goto ERRb0;
}
if ((e = mp_init_size(&b1, B)) != MP_OKAY) {
goto ERRb1;
}
if ((e = mp_init_size(&b2, B)) != MP_OKAY) {
goto ERRb2;
}
if ((e = mp_init_size(&b3, B)) != MP_OKAY) {
goto ERRb3;
}
if ((e = mp_init_size(&b4, B)) != MP_OKAY) {
goto ERRb4;
}
// A = a4*x^4 + a3*x^3 + a2*x^2 + a1*x + a0
for (count = 0; count < a->used; count++) {
switch (count / B) {
case 0:
a0.dp[count] = a->dp[count];
a0.used++;
break;
case 1:
a1.dp[count - B] = a->dp[count];
a1.used++;
break;
case 2:
a2.dp[count - 2 * B] = a->dp[count];
a2.used++;
break;
case 3:
a3.dp[count - 3 * B] = a->dp[count];
a3.used++;
break;
case 4:
a4.dp[count - 4 * B] = a->dp[count];
a4.used++;
break;
default:
a4.dp[count - 4 * B] = a->dp[count];
a4.used++;
break;
}
}
mp_clamp(&a0);
mp_clamp(&a1);
mp_clamp(&a2);
mp_clamp(&a3);
mp_clamp(&a4);
// B = b4*x^4 + b3*x^3 + b2*x^2 + b1*x + b0
for (count = 0; count < b->used; count++) {
switch (count / B) {
case 0:
b0.dp[count] = b->dp[count];
b0.used++;
break;
case 1:
b1.dp[count - B] = b->dp[count];
b1.used++;
break;
case 2:
b2.dp[count - 2 * B] = b->dp[count];
b2.used++;
break;
case 3:
b3.dp[count - 3 * B] = b->dp[count];
b3.used++;
break;
case 4:
b4.dp[count - 4 * B] = b->dp[count];
b4.used++;
break;
default:
b4.dp[count - 4 * B] = b->dp[count];
b4.used++;
break;
}
}
mp_clamp(&b0);
mp_clamp(&b1);
mp_clamp(&b2);
mp_clamp(&b3);
mp_clamp(&b4);
/*
if ((e = mp_mod_2d(a, DIGIT_BIT * B, &a0)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_copy(a, &a1)) != MP_OKAY) {
goto ERR;
}
mp_rshd(&a1, B);
mp_mod_2d(&a1, DIGIT_BIT * B, &a1);
if ((e = mp_copy(a, &a2)) != MP_OKAY) {
goto ERR;
}
mp_rshd(&a2, B * 2);
mp_mod_2d(&a2, DIGIT_BIT * B, &a2);
if ((e = mp_copy(a, &a3)) != MP_OKAY) {
goto ERR;
}
mp_rshd(&a3, B * 3);
mp_mod_2d(&a3, DIGIT_BIT * B, &a3);
if ((e = mp_copy(a, &a4)) != MP_OKAY) {
goto ERR;
}
mp_rshd(&a4, B * 4);
if ((e = mp_mod_2d(b, DIGIT_BIT * B, &b0)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_copy(a, &b1)) != MP_OKAY) {
goto ERR;
}
mp_rshd(&b1, B);
mp_mod_2d(&b1, DIGIT_BIT * B, &b1);
if ((e = mp_copy(b, &b2)) != MP_OKAY) {
goto ERR;
}
mp_rshd(&b2, B * 2);
mp_mod_2d(&b2, DIGIT_BIT * B, &b2);
if ((e = mp_copy(b, &b3)) != MP_OKAY) {
goto ERR;
}
mp_rshd(&b3, B * 3);
mp_mod_2d(&b3, DIGIT_BIT * B, &b3);
if ((e = mp_copy(b, &b4)) != MP_OKAY) {
goto ERR;
}
mp_rshd(&b4, B * 4);
*/
// S1 = a4*b4
if ((e = mp_mul(&a4, &b4, &w1)) != MP_OKAY) {
goto ERR;
}
// S9 = a0*b0
if ((e = mp_mul(&a0, &b0, &w9)) != MP_OKAY) {
goto ERR;
}
// S2 = (a0- 2*a1 +4*a2 -8*a3 +16*a4)
if ((e = mp_mul_2d(&a1, 1, &tmp1)) != MP_OKAY) {
goto ERR;
} // 2*a1 = tmp1
if ((e = mp_sub(&a0, &tmp1, &w2)) != MP_OKAY) {
goto ERR;
} // a0- 2*a1 = a0 - tmp1 = w2
if ((e = mp_mul_2d(&a2, 2, &tmp1)) != MP_OKAY) {
goto ERR;
} // 4*a2 = tmp1
if ((e = mp_add(&w2, &tmp1, &w2)) != MP_OKAY) {
goto ERR;
} // a0- 2*a1 +4*a2 = w2 + tmp1 = w2
if ((e = mp_mul_2d(&a3, 3, &tmp1)) != MP_OKAY) {
goto ERR;
} // 8*a3 = tmp1
if ((e = mp_sub(&w2, &tmp1, &w2)) != MP_OKAY) {
goto ERR;
} // a0- 2*a1 +4*a2 -8*a3 = w2 - tmp1 = w2
if ((e = mp_mul_2d(&a4, 4, &tmp1)) != MP_OKAY) {
goto ERR;
} // 16*a4 = tmp1
if ((e = mp_add(&w2, &tmp1, &w2)) != MP_OKAY) {
goto ERR;
} // a0- 2*a1 +4*a2 -8*a3 +16*a4 = w2 + tmp1 = w2
// * (b0- 2*b1 +4*b2 -8*b3 +16*b4)
if ((e = mp_mul_2d(&b1, 1, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&b0, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b2, 2, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b3, 3, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b4, 4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul(&tmp2, &w2, &w2)) != MP_OKAY) {
goto ERR;
}
// S5 = (a0+ 2*a1+ 4*a2+ 8*a3+ 16*a4)
if ((e = mp_mul_2d(&a1, 1, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&a0, &tmp1, &w5)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&a2, 2, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w5, &tmp1, &w5)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&a3, 3, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w5, &tmp1, &w5)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&a4, 4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w5, &tmp1, &w5)) != MP_OKAY) {
goto ERR;
}
// *(b0+ 2*b1+ 4*b2+ 8*b3+ 16*b4)
if ((e = mp_mul_2d(&b1, 1, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&b0, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b2, 2, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b3, 3, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b4, 4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul(&tmp2, &w5, &w5)) != MP_OKAY) {
goto ERR;
}
// S3 = (a4+ 2*a3+ 4*a2+ 8*a1+ 16*a0)
if ((e = mp_mul_2d(&a3, 1, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&a4, &tmp1, &w3)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&a2, 2, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w3, &tmp1, &w3)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&a1, 3, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w3, &tmp1, &w3)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&a0, 4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w3, &tmp1, &w3)) != MP_OKAY) {
goto ERR;
}
// * (b4+ 2*b3+ 4*b2+ 8*b1+ 16*b0)
if ((e = mp_mul_2d(&b3, 1, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&b4, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b2, 2, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b1, 3, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b0, 4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul(&tmp2, &w3, &w3)) != MP_OKAY) {
goto ERR;
}
// S8 = (a4- 2*a3+ 4*a2- 8*a1+ 16*a0)
if ((e = mp_mul_2d(&a3, 1, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&a4, &tmp1, &w8)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&a2, 2, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w8, &tmp1, &w8)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&a1, 3, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&w8, &tmp1, &w8)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&a0, 4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w8, &tmp1, &w8)) != MP_OKAY) {
goto ERR;
}
//* (b4- 2*b3+ 4*b2- 8*b1+ 16*b0)
if ((e = mp_mul_2d(&b3, 1, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&b4, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b2, 2, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b1, 3, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b0, 4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul(&tmp2, &w8, &w8)) != MP_OKAY) {
goto ERR;
}
// S4 = (a0+ 4*a1+ 16*a2+ 64*a3+ 256*a4)
if ((e = mp_mul_2d(&a1, 2, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&a0, &tmp1, &w4)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&a2, 4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w4, &tmp1, &w4)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&a3, 6, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w4, &tmp1, &w4)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&a4, 8, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w4, &tmp1, &w4)) != MP_OKAY) {
goto ERR;
}
//* (b0+ 4*b1+ 16*b2+ 64*b3+ 256*b4)
if ((e = mp_mul_2d(&b1, 2, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&b0, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b2, 4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b3, 6, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul_2d(&b4, 8, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp2, &tmp1, &tmp2)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul(&tmp2, &w4, &w4)) != MP_OKAY) {
goto ERR;
}
// S6 = (a0- a1+ a2- a3 +a4)
if ((e = mp_sub(&a0, &a1, &w6)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w6, &a2, &w6)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&w6, &a3, &w6)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w6, &a4, &w6)) != MP_OKAY) {
goto ERR;
}
// * (b0- b1+ b2- b3+ b4)
if ((e = mp_sub(&b0, &b1, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &b2, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&tmp1, &b3, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &b4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul(&tmp1, &w6, &w6)) != MP_OKAY) {
goto ERR;
}
// S7 = (a0+ a1+ a2+ a3+ a4)
if ((e = mp_add(&a0, &a1, &w7)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w7, &a2, &w7)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w7, &a3, &w7)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w7, &a4, &w7)) != MP_OKAY) {
goto ERR;
}
// * (b0+ b1+ b2+ b3+ b4)
if ((e = mp_add(&b0, &b1, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &b2, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &b3, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &b4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_mul(&tmp1, &w7, &w7)) != MP_OKAY) {
goto ERR;
}
// S6 -= S7
if ((e = mp_sub(&w6, &w7, &w6)) != MP_OKAY) {
goto ERR;
}
// S2 -= S5
if ((e = mp_sub(&w2, &w5, &w2)) != MP_OKAY) {
goto ERR;
}
// S4 -= S9
if ((e = mp_sub(&w4, &w9, &w4)) != MP_OKAY) {
goto ERR;
}
// S4 -= (2^16*S1)
if ((e = mp_mul_2d(&w1, 16, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&w4, &tmp1, &w4)) != MP_OKAY) {
goto ERR;
}
// S8 -= S3
if ((e = mp_sub(&w8, &w3, &w8)) != MP_OKAY) {
goto ERR;
}
// S6 /= 2
if ((e = mp_div_2d(&w6, 1, &w6, NULL)) != MP_OKAY) {
goto ERR;
}
// S5 *= 2
if ((e = mp_mul_2d(&w5, 1, &w5)) != MP_OKAY) {
goto ERR;
}
// S5 += S2
if ((e = mp_add(&w5, &w2, &w5)) != MP_OKAY) {
goto ERR;
}
// S2 = -S2
if ((e = mp_neg(&w2, &w2)) != MP_OKAY) {
goto ERR;
}
// S8 = -S8
if ((e = mp_neg(&w8, &w8)) != MP_OKAY) {
goto ERR;
}
// S7 += S6
if ((e = mp_add(&w7, &w6, &w7)) != MP_OKAY) {
goto ERR;
}
// S6 = -S6
if ((e = mp_neg(&w6, &w6)) != MP_OKAY) {
goto ERR;
}
// S3 -= S7
if ((e = mp_sub(&w3, &w7, &w3)) != MP_OKAY) {
goto ERR;
}
// S5 -= (512*S7)
if ((e = mp_mul_2d(&w7, 9, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&w5, &tmp1, &w5)) != MP_OKAY) {
goto ERR;
}
// S3 *= 2
if ((e = mp_mul_2d(&w3, 1, &w3)) != MP_OKAY) {
goto ERR;
}
// S3 -= S8
if ((e = mp_sub(&w3, &w8, &w3)) != MP_OKAY) {
goto ERR;
}
// S7 -= S1
if ((e = mp_sub(&w7, &w1, &w7)) != MP_OKAY) {
goto ERR;
}
// S7 -= S9
if ((e = mp_sub(&w7, &w9, &w7)) != MP_OKAY) {
goto ERR;
}
// S8 += S2
if ((e = mp_add(&w8, &w2, &w8)) != MP_OKAY) {
goto ERR;
}
// S5 += S3
if ((e = mp_add(&w5, &w3, &w5)) != MP_OKAY) {
goto ERR;
}
// S8 -= (80*S6)
if ((e = mp_mul_d(&w6, 80, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&w8, &tmp1, &w8)) != MP_OKAY) {
goto ERR;
}
// S3 -= (510*S9)
if ((e = mp_mul_d(&w9, 510, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&w3, &tmp1, &w3)) != MP_OKAY) {
goto ERR;
}
// S4 -= S2
if ((e = mp_sub(&w4, &w2, &w4)) != MP_OKAY) {
goto ERR;
}
// S3 *= 3
if ((e = mp_mul_d(&w3, 3, &w3)) != MP_OKAY) {
goto ERR;
}
// S3 += S5
if ((e = mp_add(&w3, &w5, &w3)) != MP_OKAY) {
goto ERR;
}
// S8 /= 180 \\ division by 180
if ((e = mp_div_d(&w8, 180, &w8, NULL)) != MP_OKAY) {
goto ERR;
}
// S5 += (378*S7)
if ((e = mp_mul_d(&w7, 378, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w5, &tmp1, &w5)) != MP_OKAY) {
goto ERR;
}
// S2 /= 4
if ((e = mp_div_2d(&w2, 2, &w2, NULL)) != MP_OKAY) {
goto ERR;
}
// S6 -= S2
if ((e = mp_sub(&w6, &w2, &w6)) != MP_OKAY) {
goto ERR;
}
// S5 /= (-72) \\ division by -72
if ((e = mp_div_d(&w5, 72, &w5, NULL)) != MP_OKAY) {
goto ERR;
}
if (&w5.sign == MP_ZPOS)
(&w5)->sign = MP_NEG;
(&w5)->sign = MP_ZPOS;
// S3 /= (-360) \\ division by -360
if ((e = mp_div_d(&w3, 360, &w3, NULL)) != MP_OKAY) {
goto ERR;
}
if (&w3.sign == MP_ZPOS)
(&w3)->sign = MP_NEG;
(&w3)->sign = MP_ZPOS;
// S2 -= S8
if ((e = mp_sub(&w2, &w8, &w2)) != MP_OKAY) {
goto ERR;
}
// S7 -= S3
if ((e = mp_sub(&w7, &w3, &w7)) != MP_OKAY) {
goto ERR;
}
// S4 -= (256*S5)
if ((e = mp_mul_2d(&w5, 8, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&w4, &tmp1, &w4)) != MP_OKAY) {
goto ERR;
}
// S3 -= S5
if ((e = mp_sub(&w3, &w5, &w3)) != MP_OKAY) {
goto ERR;
}
// S4 -= (4096*S3)
if ((e = mp_mul_2d(&w3, 12, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&w4, &tmp1, &w4)) != MP_OKAY) {
goto ERR;
}
// S4 -= (16*S7)
if ((e = mp_mul_2d(&w7, 4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_sub(&w4, &tmp1, &w4)) != MP_OKAY) {
goto ERR;
}
// S4 += (256*S6)
if ((e = mp_mul_2d(&w6, 8, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w4, &tmp1, &w4)) != MP_OKAY) {
goto ERR;
}
// S6 += S2
if ((e = mp_add(&w6, &w2, &w6)) != MP_OKAY) {
goto ERR;
}
// S2 *= 180
if ((e = mp_mul_d(&w2, 180, &w2)) != MP_OKAY) {
goto ERR;
}
// S2 += S4
if ((e = mp_add(&w2, &w4, &w2)) != MP_OKAY) {
goto ERR;
}
// S2 /= 11340 \\ division by 11340
if ((e = mp_div_d(&w2, 11340, &w2, NULL)) != MP_OKAY) {
goto ERR;
}
// S4 += (720*S6)
if ((e = mp_mul_d(&w6, 720, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&w4, &tmp1, &w4)) != MP_OKAY) {
goto ERR;
}
// S4 /= (-2160) \\ division by -2160
if ((e = mp_div_d(&w4, 2160, &w4, NULL)) != MP_OKAY) {
goto ERR;
}
if (&w4.sign == MP_ZPOS)
(&w4)->sign = MP_NEG;
(&w4)->sign = MP_ZPOS;
// S6 -= S4
if ((e = mp_sub(&w6, &w4, &w6)) != MP_OKAY) {
goto ERR;
}
// S8 -= S2
if ((e = mp_sub(&w8, &w2, &w8)) != MP_OKAY) {
goto ERR;
}
// P = S1*x^8 + S2*x^7 + S3*x^6 + S4*x^5 + S5*x^4 + S6*x^3 + S7*x^2 + S8*x + S9
if ((e = mp_copy(&w9, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_lshd(&w8, B)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &w8, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_lshd(&w7, 2 * B)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &w7, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_lshd(&w6, 3 * B)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &w6, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_lshd(&w5, 4 * B)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &w5, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_lshd(&w4, 5 * B)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &w4, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_lshd(&w3, 6 * B)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &w3, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_lshd(&w2, 7 * B)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &w2, &tmp1)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_lshd(&w1, 8 * B)) != MP_OKAY) {
goto ERR;
}
if ((e = mp_add(&tmp1, &w1, c)) != MP_OKAY) {
goto ERR;
}
// P - A*B \\ == zero
c->sign = sign;
ERR:
ERRb4:
mp_clear(&b4);
ERRb3:
mp_clear(&b3);
ERRb2:
mp_clear(&b2);
ERRb1:
mp_clear(&b1);
ERRb0:
mp_clear(&b0);
ERRa4:
mp_clear(&a4);
ERRa3:
mp_clear(&a3);
ERRa2:
mp_clear(&a2);
ERRa1:
mp_clear(&a1);
ERRa0:
mp_clear(&a0);
ERR0:
mp_clear_multi(&w1, &w2, &w3, &w4, &w5, &w6, &w7, &w8, &w9, &tmp1, &tmp2,
// &a0, &a1, &a2, &a3, &a4, &b0, &b1, &b2, &b3, &b4,
NULL);
return e;
}
/* this is a modified version of fast_s_mul_digs that only produces
* output digits *above* digs. See the comments for fast_s_mul_digs
* to see how it works.
*
* This is used in the Barrett reduction since for one of the multiplications
* only the higher digits were needed. This essentially halves the work.
*
* Based on Algorithm 14.12 on pp.595 of HAC.
*/
int fast_s_mp_mul_high_digs (mp_int * a, mp_int * b, mp_int * c, int digs)
{
int olduse, res, pa, ix, iz;
mp_digit W[MP_WARRAY];
mp_word _W;
/* grow the destination as required */
pa = a->used + b->used;
if (c->alloc < pa) {
if ((res = mp_grow (c, pa)) != MP_OKAY) {
return res;
}
}
/* number of output digits to produce */
pa = a->used + b->used;
_W = 0;
for (ix = digs; ix < pa; ix++) {
int tx, ty, iy;
mp_digit *tmpx, *tmpy;
/* get offsets into the two bignums */
ty = MIN(b->used-1, ix);
tx = ix - ty;
/* setup temp aliases */
tmpx = a->dp + tx;
tmpy = b->dp + ty;
/* this is the number of times the loop will iterrate, essentially its
while (tx++ < a->used && ty-- >= 0) { ... }
*/
iy = MIN(a->used-tx, ty+1);
/* execute loop */
for (iz = 0; iz < iy; iz++) {
_W += ((mp_word)*tmpx++)*((mp_word)*tmpy--);
}
/* store term */
W[ix] = ((mp_digit)_W) & MP_MASK;
/* make next carry */
_W = _W >> ((mp_word)DIGIT_BIT);
}
/* setup dest */
olduse = c->used;
c->used = pa;
{
mp_digit *tmpc;
tmpc = c->dp + digs;
for (ix = digs; ix < pa; ix++) {
/* now extract the previous digit [below the carry] */
*tmpc++ = W[ix];
}
/* clear unused digits [that existed in the old copy of c] */
for (; ix < olduse; ix++) {
*tmpc++ = 0;
}
}
mp_clamp (c);
return MP_OKAY;
}
/* computes xR**-1 == x (mod N) via Montgomery Reduction */
int
mp_montgomery_reduce (mp_int * x, mp_int * n, mp_digit rho)
{
int ix, res, digs;
mp_digit mu;
/* can the fast reduction [comba] method be used?
*
* Note that unlike in mul you're safely allowed *less*
* than the available columns [255 per default] since carries
* are fixed up in the inner loop.
*/
digs = n->used * 2 + 1;
if ((digs < MP_WARRAY) &&
n->used <
(1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) {
return fast_mp_montgomery_reduce (x, n, rho);
}
/* grow the input as required */
if (x->alloc < digs) {
if ((res = mp_grow (x, digs)) != MP_OKAY) {
return res;
}
}
x->used = digs;
for (ix = 0; ix < n->used; ix++) {
/* mu = ai * rho mod b
*
* The value of rho must be precalculated via
* montgomery_setup() such that
* it equals -1/n0 mod b this allows the
* following inner loop to reduce the
* input one digit at a time
*/
mu = (mp_digit) (((mp_word)x->dp[ix]) * ((mp_word)rho) & MP_MASK);
/* a = a + mu * m * b**i */
{
register int iy;
register mp_digit *tmpn, *tmpx, u;
register mp_word r;
/* alias for digits of the modulus */
tmpn = n->dp;
/* alias for the digits of x [the input] */
tmpx = x->dp + ix;
/* set the carry to zero */
u = 0;
/* Multiply and add in place */
for (iy = 0; iy < n->used; iy++) {
/* compute product and sum */
r = ((mp_word)mu) * ((mp_word)*tmpn++) +
((mp_word) u) + ((mp_word) * tmpx);
/* get carry */
u = (mp_digit)(r >> ((mp_word) DIGIT_BIT));
/* fix digit */
*tmpx++ = (mp_digit)(r & ((mp_word) MP_MASK));
}
/* At this point the ix'th digit of x should be zero */
/* propagate carries upwards as required*/
while (u) {
*tmpx += u;
u = *tmpx >> DIGIT_BIT;
*tmpx++ &= MP_MASK;
}
}
}
/* at this point the n.used'th least
* significant digits of x are all zero
* which means we can shift x to the
* right by n.used digits and the
* residue is unchanged.
*/
/* x = x/b**n.used */
mp_clamp(x);
mp_rshd (x, n->used);
/* if x >= n then x = x - n */
if (mp_cmp_mag (x, n) != MP_LT) {
return s_mp_sub (x, n, x);
}
return MP_OKAY;
}
/* single digit subtraction */
int
mp_sub_d (mp_int * a, mp_digit b, mp_int * c)
{
mp_digit *tmpa, *tmpc, mu;
int res, ix, oldused;
/* grow c as required */
if (c->alloc < a->used + 1) {
if ((res = mp_grow(c, a->used + 1)) != MP_OKAY) {
return res;
}
}
/* if a is negative just do an unsigned
* addition [with fudged signs]
*/
if (a->sign == MP_NEG) {
a->sign = MP_ZPOS;
res = mp_add_d(a, b, c);
a->sign = c->sign = MP_NEG;
/* clamp */
mp_clamp(c);
return res;
}
/* setup regs */
oldused = c->used;
tmpa = a->dp;
tmpc = c->dp;
/* if a <= b simply fix the single digit */
if ((a->used == 1 && a->dp[0] <= b) || a->used == 0) {
if (a->used == 1) {
*tmpc++ = b - *tmpa;
} else {
*tmpc++ = b;
}
ix = 1;
/* negative/1digit */
c->sign = MP_NEG;
c->used = 1;
} else {
/* positive/size */
c->sign = MP_ZPOS;
c->used = a->used;
/* subtract first digit */
*tmpc = *tmpa++ - b;
mu = *tmpc >> (sizeof(mp_digit) * CHAR_BIT - 1);
*tmpc++ &= MP_MASK;
/* handle rest of the digits */
for (ix = 1; ix < a->used; ix++) {
*tmpc = *tmpa++ - mu;
mu = *tmpc >> (sizeof(mp_digit) * CHAR_BIT - 1);
*tmpc++ &= MP_MASK;
}
}
/* zero excess digits */
while (ix++ < oldused) {
*tmpc++ = 0;
}
mp_clamp(c);
return MP_OKAY;
}
/* Fast (comba) multiplier
*
* This is the fast column-array [comba] multiplier. It is
* designed to compute the columns of the product first
* then handle the carries afterwards. This has the effect
* of making the nested loops that compute the columns very
* simple and schedulable on super-scalar processors.
*
* This has been modified to produce a variable number of
* digits of output so if say only a half-product is required
* you don't have to compute the upper half (a feature
* required for fast Barrett reduction).
*
* Based on Algorithm 14.12 on pp.595 of HAC.
*
*/
int fast_s_mp_mul_digs(mp_int * a, mp_int * b, mp_int * c, int digs)
{
int olduse, res, pa, ix;
extern mp_word *W;
/* grow the destination as required */
if (c->alloc < digs) {
if ((res = mp_grow(c, digs)) != MP_OKAY) {
return res;
}
}
/* clear temp buf (the columns) */
memset(W, 0, sizeof(mp_word) * digs);
/* calculate the columns */
pa = a->used;
for (ix = 0; ix < pa; ix++) {
/* this multiplier has been modified to allow you to
* control how many digits of output are produced.
* So at most we want to make upto "digs" digits of output.
*
* this adds products to distinct columns (at ix+iy) of W
* note that each step through the loop is not dependent on
* the previous which means the compiler can easily unroll
* the loop without scheduling problems
*/
{
register mp_digit tmpx, *tmpy;
register mp_word *_W;
register int iy, pb;
/* alias for the the word on the left e.g. A[ix] * A[iy] */
tmpx = a->dp[ix];
/* alias for the right side */
tmpy = b->dp;
/* alias for the columns, each step through the loop adds a new
term to each column
*/
_W = W + ix;
/* the number of digits is limited by their placement. E.g.
we avoid multiplying digits that will end up above the # of
digits of precision requested
*/
pb = MIN(b->used, digs - ix);
for (iy = 0; iy < pb; iy++) {
*_W++ +=
((mp_word) tmpx) * ((mp_word) *
tmpy++);
}
}
}
/* setup dest */
olduse = c->used;
c->used = digs;
{
register mp_digit *tmpc;
/* At this point W[] contains the sums of each column. To get the
* correct result we must take the extra bits from each column and
* carry them down
*
* Note that while this adds extra code to the multiplier it
* saves time since the carry propagation is removed from the
* above nested loop.This has the effect of reducing the work
* from N*(N+N*c)==N**2 + c*N**2 to N**2 + N*c where c is the
* cost of the shifting. On very small numbers this is slower
* but on most cryptographic size numbers it is faster.
*
* In this particular implementation we feed the carries from
* behind which means when the loop terminates we still have one
* last digit to copy
*/
tmpc = c->dp;
for (ix = 1; ix < digs; ix++) {
/* forward the carry from the previous temp */
W[ix] += (W[ix - 1] >> ((mp_word) DIGIT_BIT));
/* now extract the previous digit [below the carry] */
*tmpc++ =
(mp_digit) (W[ix - 1] & ((mp_word) MP_MASK));
}
/* fetch the last digit */
*tmpc++ = (mp_digit) (W[digs - 1] & ((mp_word) MP_MASK));
/* clear unused digits [that existed in the old copy of c] */
for (; ix < olduse; ix++) {
*tmpc++ = 0;
}
}
mp_clamp(c);
return MP_OKAY;
}
/* Karatsuba squaring, computes b = a*a using three
* half size squarings
*
* See comments of karatsuba_mul for details. It
* is essentially the same algorithm but merely
* tuned to perform recursive squarings.
*/
int mp_karatsuba_sqr (mp_int * a, mp_int * b)
{
mp_int x0, x1, t1, t2, x0x0, x1x1;
int B, err;
err = MP_MEM;
/* min # of digits */
B = USED(a);
/* now divide in two */
B = B >> 1;
/* init copy all the temps */
if (mp_init_size (&x0, B) != MP_OKAY)
goto ERR;
if (mp_init_size (&x1, USED(a) - B) != MP_OKAY)
goto X0;
/* init temps */
if (mp_init_size (&t1, USED(a) * 2) != MP_OKAY)
goto X1;
if (mp_init_size (&t2, USED(a) * 2) != MP_OKAY)
goto T1;
if (mp_init_size (&x0x0, B * 2) != MP_OKAY)
goto T2;
if (mp_init_size (&x1x1, (USED(a) - B) * 2) != MP_OKAY)
goto X0X0;
{
register int x;
register mp_digit *dst, *src;
src = DIGITS(a);
/* now shift the digits */
dst = DIGITS(&x0);
for (x = 0; x < B; x++) {
*dst++ = *src++;
}
dst = DIGITS(&x1);
for (x = B; x < USED(a); x++) {
*dst++ = *src++;
}
}
SET_USED(&x0,B);
SET_USED(&x1,USED(a) - B);
mp_clamp (&x0);
/* now calc the products x0*x0 and x1*x1 */
if (mp_sqr (&x0, &x0x0) != MP_OKAY)
goto X1X1; /* x0x0 = x0*x0 */
if (mp_sqr (&x1, &x1x1) != MP_OKAY)
goto X1X1; /* x1x1 = x1*x1 */
/* now calc (x1+x0)**2 */
if (s_mp_add (&x1, &x0, &t1) != MP_OKAY)
goto X1X1; /* t1 = x1 - x0 */
if (mp_sqr (&t1, &t1) != MP_OKAY)
goto X1X1; /* t1 = (x1 - x0) * (x1 - x0) */
/* add x0y0 */
if (s_mp_add (&x0x0, &x1x1, &t2) != MP_OKAY)
goto X1X1; /* t2 = x0x0 + x1x1 */
if (s_mp_sub (&t1, &t2, &t1) != MP_OKAY)
goto X1X1; /* t1 = (x1+x0)**2 - (x0x0 + x1x1) */
/* shift by B */
if (mp_lshd (&t1, B) != MP_OKAY)
goto X1X1; /* t1 = (x0x0 + x1x1 - (x1-x0)*(x1-x0))<<B */
if (mp_lshd (&x1x1, B * 2) != MP_OKAY)
goto X1X1; /* x1x1 = x1x1 << 2*B */
if (mp_add (&x0x0, &t1, &t1) != MP_OKAY)
goto X1X1; /* t1 = x0x0 + t1 */
if (mp_add (&t1, &x1x1, b) != MP_OKAY)
goto X1X1; /* t1 = x0x0 + t1 + x1x1 */
err = MP_OKAY;
X1X1:mp_clear (&x1x1);
X0X0:mp_clear (&x0x0);
T2:mp_clear (&t2);
T1:mp_clear (&t1);
X1:mp_clear (&x1);
X0:mp_clear (&x0);
ERR:
return err;
}
/* Karatsuba squaring, computes b = a*a using three
* half size squarings
*
* See comments of karatsuba_mul for details. It
* is essentially the same algorithm but merely
* tuned to perform recursive squarings.
*/
int mp_karatsuba_sqr(const mp_int *a, mp_int *b)
{
mp_int x0, x1, t1, t2, x0x0, x1x1;
int B, err;
err = MP_MEM;
/* min # of digits */
B = a->used;
/* now divide in two */
B = B >> 1;
/* init copy all the temps */
if (mp_init_size(&x0, B) != MP_OKAY)
goto LBL_ERR;
if (mp_init_size(&x1, a->used - B) != MP_OKAY)
goto X0;
/* init temps */
if (mp_init_size(&t1, a->used * 2) != MP_OKAY)
goto X1;
if (mp_init_size(&t2, a->used * 2) != MP_OKAY)
goto T1;
if (mp_init_size(&x0x0, B * 2) != MP_OKAY)
goto T2;
if (mp_init_size(&x1x1, (a->used - B) * 2) != MP_OKAY)
goto X0X0;
{
int x;
mp_digit *dst, *src;
src = a->dp;
/* now shift the digits */
dst = x0.dp;
for (x = 0; x < B; x++) {
*dst++ = *src++;
}
dst = x1.dp;
for (x = B; x < a->used; x++) {
*dst++ = *src++;
}
}
x0.used = B;
x1.used = a->used - B;
mp_clamp(&x0);
/* now calc the products x0*x0 and x1*x1 */
if (mp_sqr(&x0, &x0x0) != MP_OKAY)
goto X1X1; /* x0x0 = x0*x0 */
if (mp_sqr(&x1, &x1x1) != MP_OKAY)
goto X1X1; /* x1x1 = x1*x1 */
/* now calc (x1+x0)**2 */
if (s_mp_add(&x1, &x0, &t1) != MP_OKAY)
goto X1X1; /* t1 = x1 - x0 */
if (mp_sqr(&t1, &t1) != MP_OKAY)
goto X1X1; /* t1 = (x1 - x0) * (x1 - x0) */
/* add x0y0 */
if (s_mp_add(&x0x0, &x1x1, &t2) != MP_OKAY)
goto X1X1; /* t2 = x0x0 + x1x1 */
if (s_mp_sub(&t1, &t2, &t1) != MP_OKAY)
goto X1X1; /* t1 = (x1+x0)**2 - (x0x0 + x1x1) */
/* shift by B */
if (mp_lshd(&t1, B) != MP_OKAY)
goto X1X1; /* t1 = (x0x0 + x1x1 - (x1-x0)*(x1-x0))<<B */
if (mp_lshd(&x1x1, B * 2) != MP_OKAY)
goto X1X1; /* x1x1 = x1x1 << 2*B */
if (mp_add(&x0x0, &t1, &t1) != MP_OKAY)
goto X1X1; /* t1 = x0x0 + t1 */
if (mp_add(&t1, &x1x1, b) != MP_OKAY)
goto X1X1; /* t1 = x0x0 + t1 + x1x1 */
err = MP_OKAY;
X1X1:
mp_clear(&x1x1);
X0X0:
mp_clear(&x0x0);
T2:
mp_clear(&t2);
T1:
mp_clear(&t1);
X1:
mp_clear(&x1);
X0:
mp_clear(&x0);
LBL_ERR:
return err;
}
/* low level subtraction (assumes |a| > |b|), HAC pp.595 Algorithm 14.9 */
int
s_mp_sub (mp_int * a, mp_int * b, mp_int * c)
{
int olduse, res, min, max;
/* find sizes */
min = b->used;
max = a->used;
/* init result */
if (c->alloc < max) {
if ((res = mp_grow (c, max)) != MP_OKAY) {
return res;
}
}
olduse = c->used;
c->used = max;
{
register mp_digit u, *tmpa, *tmpb, *tmpc;
register int i;
/* alias for digit pointers */
tmpa = a->dp;
tmpb = b->dp;
tmpc = c->dp;
/* set carry to zero */
u = 0;
for (i = 0; i < min; i++) {
/* T[i] = A[i] - B[i] - U */
*tmpc = *tmpa++ - *tmpb++ - u;
/* U = carry bit of T[i]
* Note this saves performing an AND operation since
* if a carry does occur it will propagate all the way to the
* MSB. As a result a single shift is enough to get the carry
*/
u = *tmpc >> ((mp_digit)(CHAR_BIT * sizeof (mp_digit) - 1));
/* Clear carry from T[i] */
*tmpc++ &= MP_MASK;
}
/* now copy higher words if any, e.g. if A has more digits than B */
for (; i < max; i++) {
/* T[i] = A[i] - U */
*tmpc = *tmpa++ - u;
/* U = carry bit of T[i] */
u = *tmpc >> ((mp_digit)(CHAR_BIT * sizeof (mp_digit) - 1));
/* Clear carry from T[i] */
*tmpc++ &= MP_MASK;
}
/* clear digits above used (since we may not have grown result above) */
for (i = c->used; i < olduse; i++) {
*tmpc++ = 0;
}
}
mp_clamp (c);
return MP_OKAY;
}