private void RoundBCA(
long x,
long mul)
{
a ^= x ;
b -= t1[(int)a & 0xff] ^ t2[(int)(a >> 16) & 0xff]
^ t3[(int)(a >> 32) & 0xff] ^ t4[(int)(a >> 48) & 0xff];
c += t4[(int)(a >> 8) & 0xff] ^ t3[(int)(a >> 24) & 0xff]
^ t2[(int)(a >> 40) & 0xff] ^ t1[(int)(a >> 56) & 0xff];
c *= mul;
}