Works like Array.Copy or Buffer.BlockCopy but with memcpy forward only operation vs memmove

ForwardCopy.cs

public static void ForwardCopy(byte[] s, uint si, byte[] d, uint di, uint cx) {
    unchecked {
        // pre copy to qword align to destination
        uint cxpre = (8 - (di & 7)) & 7;
        if (cx < cxpre) {
            // early out if we cannot even align to the first qword
            for (int i = 0; i < cx; i++)
                d[di++] = s[si++];
            return;
        }
        for (uint i = 0; i < cxpre; i++) d[di++] = s[si++];

        // qword copy
        var dl = MemoryMarshal.Cast<byte, ulong>(d.AsSpan((int)(di)));
        var sl = MemoryMarshal.Cast<byte, ulong>(s.AsSpan((int)(si)));
        uint cxloop = (cx - cxpre) >> 3;
        for (int i = 0; i < cxloop; i++)
            dl[i] = sl[i];
        si += cxloop << 3; 
        di += cxloop << 3;

        // post qword remainder copy
        uint cxpost = cx - (cxloop << 3) - cxpre;
        for (uint i = 0; i < cxpost; i++) 
            d[di++] = s[si++];
    }
}