A fast algorithm for complete subcube recognition
Hal Burch, Fikret Erçal · 2002
The complete subcube recognition problem is defined as, given a collection of available processors on an n-dimensional hypercube, locate a subcube of dimension k that consists entirely of available processors, if one exists. Despite many algorithms proposed so far on this subject, improving the time complexity of this problem remains a challenge. Efficiency limits that can be reached have not been exhausted yet. This paper proposes a novel algorithm to recognize all the overlapping subcubes available on an n-dimensional hypercube whose processors are partially allocated. Given P=2/sup n/, as the total number of processors in the hypercube, the new algorithm runs in O(n-3/sup n/) or O(P(log/sub 2/) /sup 3/log/sub 2/ P) time which is an improvement over previously proposed strategies, such as multiple-graycode, missing combination, maximal set of subcubes, and tree collapsing.