The modeling of large integrated circuits leads to a system of differential-algebraic equations, typically with several million equations and unknowns. In addition, these systems of equations contain implicit constraints, which have to be considered during discretization. Solving these equations numerically leads to very large, sparse systems of equations. The aim is to provide robust and efficient preconditioners and to apply them to the linear systems of equations from circuit simulation. Furthermore, implicit constraints are used for index reduction using the topology of the circuit.