Systolic generation of derangements

Iu recent years, a nurnber ofparallel algorithms have appeared for the generation ofcombinator-ial objects such as combinations, permutations, subsets, equivalence relations, etc. However, ltlauy of these algorithms are not cost-optimal, are designed for a powerful model of coutputirtion, or require a large rnemory per processor. To date, no parallel derangement generltion ulgorithrn possessing all of a number of desirable properties has been proposed. In this pirpe I we will show how the pernrutatiort generation algorithm proposed in [6] (which satisfies Irll orrr ploperties) rnotivates a derangemerrf generation algorithm, which also satisfies all our dcsirlble properties. In particular, both algorithms are designed for the (arguably) weakest nrodel of cornputation, the linear systolic array.