In this paper we present two variants of the DP procedure which overcome the problem outlined above. The idea underlying these variants is that splitting should occur only for the variables in the original formula. The CNF conversion methods employed ensure their correctness and completeness. As a consequence, we get two decision procedures for non-CNF formulas (i) which can exploit all the present and future sophisticated technology of current DP implementations, and (ii) whose space of assignments they have to search in, is limited in size by the number of variables in the original input formula.

In [11], it is showed that limiting the splitting step to a subset of the set of variables (the truth values of the others being consequentially determined) can lead to significant speeds up.