Parallelizing computations of full disjunctions