Experimental Parallel Algorithmics

    Experimental Data Sets:

    h-Relation Personalized Communication

    Institute for Advanced Computer Studies

    University of Maryland, College Park

    Variables

    n total number of elements
    p number of processors
    i the processor number from (0 <= i <= p-1)
    h an h-relation, where h is a multiple of n/p

    • Our input of size n is initially distributed cyclically across the p processors such that each processor i initially holds n/p keys, for (0 <= i <= p-1).
    • For h = n/p the input consists of v[0] = n/p keys labelled for P0, followed by v[1] = n/p keys labelled for P1, and so forth, (with v[i] = n/p keys labelled for Pi), with the last v[p-1] = n/p keys labelled for P[p-1].
    • For h > n/p, instead of an equal number of elements destined for each processor, the function v[i], for (0 <= i <= p-1), is characterized by
    floor(h × {1 - [h / (2n-h)] i}) if i < 2n/h and i < p-1
    0 if 2n/h <= i < p-1
    n - Sum(from j=0 to 2n/h - 1) [floor(h × {1 - [h / (2n-h)] j})] if i = p-1

    Return to the Experimental Parallel Algorithmics Experimental Data Sets page.

    dbader@umiacs.umd.edu