We demonstrate that in a coupled two-qubit system any single-qubit gate can be decomposed into ͉0͘-controlled and ͉1͘-controlled two-qubit gates which can be implemented by manipulations analogous to that used for a controlled NOT ͑CNOT͒ gate. Based on this we present a unified approach to implement universal single-qubit and two-qubit gates in a coupled two-qubit system with fixed always-on coupling. This approach requires neither supplementary circuit or additional physical qubits to control the coupling nor extra hardware to adjust the energy level structure. The feasibility of this approach is demonstrated by numerical simulation of single-qubit gates and creation of two-qubit Bell states in rf-driven inductively coupled two superconducting quantum interference device flux qubits with realistic device parameters and constant always-on coupling. However, building a practical quantum computer requires the simultaneous operation of a large number of multiqubit gates in a coupled multiqubit system. It has been proved theoretically that any type of multiqubit gate can be decomposed into a set of universal single-qubit gates and a twoqubit gate, such as the controlled-NOT ͑CNOT͒ gate. 16,17 Thus it is imperative to implement the universal single-qubit and two-qubit gates in a multiqubit system with the minimum resource and maximum efficiency. 18 Implementing universal single-qubit gates and two-qubit gates in coupled multiqubit systems can be achieved by turning off and on the coupling between qubits. [18][19][20][21] In these schemes, supplementary circuits were required to control interqubit coupling. However, rapid switching of the coupling results in two serious problems. The first one is gate errors caused by population propagation between qubits. Because the computational states of the single-qubit gates are not a subset of the eigenstates of the two-qubit gates the populations of the computational states propagate from one qubit to another when the coupling is changed, resulting in additional gate errors. The second one is additional decoherence introduced by the supplementary circuits. 22,23 This is one of the biggest obstacles for quantum computing with solid-state qubits, particularly in coupled multiqubit systems. 2 In addition, the use of supplementary circuits also significantly increases the complexity of fabrication and manipulation of the coupled qubits.To circumvent these problems, a couple of alternative schemes, such as those with untunable coupling 22 and always-on interaction, 24 have been proposed. In the first scheme, each logic qubit is encoded by extra physical qubits and coupling between the encoded qubits is constant but can be turned off and on effectively by putting the qubits in and driving them out of the interaction free subspace. In the second scheme, the coupling is always on but the transition energies of the qubits are tuned individually or collectively. These schemes can overcome the problem of undesired population propagation but still suffer from those caused by the supplemen...