Using universe knowledge and arithmetic to get faster parallel algorithms

The authors provide algorithms that achieve better upper bounds for three classical parallel problems, i.e., searching, merging and maximum finding. The algorithms assume that the p parallel processors can use arithmetic and that the universe U of the input elements is bounded and known. They show that parallel searching is done in O (loglogU/logp), while parallel merging and maximum finding in O(logloglogU). The results hold for the CRCW and CREW PRAM models