/* P=u0.Q0+u1*Q1+u2*Q2+u3*Q3 */
public static ECP2 mul4(ECP2[] Q, BIG[] u)
{
int i, j, nb;
int[] a = new int[4];
ECP2 T = new ECP2();
ECP2 C = new ECP2();
ECP2 P = new ECP2();
ECP2[] W = new ECP2[8];
BIG mt = new BIG();
BIG[] t = new BIG[4];
sbyte[] w = new sbyte[ROM.NLEN * ROM.BASEBITS + 1];
for (i = 0; i < 4; i++)
{
t[i] = new BIG(u[i]);
Q[i].affine();
}
/* precompute table */
W[0] = new ECP2();
W[0].copy(Q[0]);
W[0].sub(Q[1]);
W[1] = new ECP2();
W[1].copy(W[0]);
W[2] = new ECP2();
W[2].copy(W[0]);
W[3] = new ECP2();
W[3].copy(W[0]);
W[4] = new ECP2();
W[4].copy(Q[0]);
W[4].add(Q[1]);
W[5] = new ECP2();
W[5].copy(W[4]);
W[6] = new ECP2();
W[6].copy(W[4]);
W[7] = new ECP2();
W[7].copy(W[4]);
T.copy(Q[2]);
T.sub(Q[3]);
W[1].sub(T);
W[2].add(T);
W[5].sub(T);
W[6].add(T);
T.copy(Q[2]);
T.add(Q[3]);
W[0].sub(T);
W[3].add(T);
W[4].sub(T);
W[7].add(T);
multiaffine(8, W);
/* if multiplier is even add 1 to multiplier, and add P to correction */
mt.zero();
C.inf();
for (i = 0; i < 4; i++)
{
if (t[i].parity() == 0)
{
t[i].inc(1);
t[i].norm();
C.add(Q[i]);
}
mt.add(t[i]);
mt.norm();
}
nb = 1 + mt.nbits();
/* convert exponent to signed 1-bit window */
for (j = 0; j < nb; j++)
{
for (i = 0; i < 4; i++)
{
a[i] = (sbyte)(t[i].lastbits(2) - 2);
t[i].dec(a[i]);
t[i].norm();
t[i].fshr(1);
}
w[j] = (sbyte)(8 * a[0] + 4 * a[1] + 2 * a[2] + a[3]);
}
w[nb] = (sbyte)(8 * t[0].lastbits(2) + 4 * t[1].lastbits(2) + 2 * t[2].lastbits(2) + t[3].lastbits(2));
P.copy(W[(w[nb] - 1) / 2]);
for (i = nb - 1; i >= 0; i--)
{
T.select(W, w[i]);
P.dbl();
P.add(T);
}
P.sub(C); // apply correction
P.affine();
return(P);
}