felem_diff64
felem_diff64(ftmp, delta); /* ftmp[i] < 2^60 + 2^58 + 2^44 */
felem_diff64(beta, x_out); /* beta[i] < 2^60 + 2^58 + 2^44 */
felem_diff64(ftmp3, x_out); /* ftmp3[i] < 2^60 + 2^56 + 2^44 */
felem_diff64(ftmp, delta);
felem_diff64(beta, x_out);
felem_diff64(ftmp3, x_out);