Performance evaluation of parallel branch and bound search with the Intel IPSC (Intel Personal Supercomputer) hypercube computer. Master's thesis
With the recent availability of commercial parallel computers, researchers are examining new classes of problems for benefits from parallel processing. This report presents results of an investigation of the set of problems classified as search-intensive. The specific problems discussed in this report are the backtracking search method of the N-queens problem and the Least-Cost Branch and Bound search of deadline job scheduling. The object-oriented design methodology was used to map the problem into a parallel solution. While the initial design was good for a prototype, the best performance resulted from fine tuning the algorithms for a specific computer. The experiments of the N-queens and deadline job scheduling included an analysis of the computation time to first solution, the computation time to all solutions, the speedup over a VAX 11/785, and the load balance of the problem when using an Intel Personal SuperComputer(IPSC). The IPSC is a loosely couple multiprocessor system based on a hypercube architecture. Results are presented that compare the performance of the IPSC and VAX 11/785 for these classes of problems.
- Research Organization:
- Air Force Inst. of Tech., Wright-Patterson AFB, OH (USA). School of Engineering
- OSTI ID:
- 6399714
- Report Number(s):
- AD-A-179384/3/XAB; AFIT/GCE/ENG-86D-2
- Country of Publication:
- United States
- Language:
- English
Similar Records
Parallel implementation of VHDL simulations on the Intel iPSC/2 hypercube. Master's thesis
PRASE (Parallel Resource Analysis Software Environment) instrumentation software for the Intel IPSC hypercube. Master's thesis