Experience with a Personal Sized Supercomputer Implications for Algorithm Development
While the use of supercomputers for computational fluid dynamics, CFD, has received considerable attention, alternatives to batch use of supercomputers for CFD have received little discussion. In the current paper, a class of systems which is inexpensive enough that they may be dedicated full time to single projects is discussed to demonstrate that such systems are a useful alternative to batch supercomputers for CFD research projects. Low-cost, dedicated systems that can perform computations of complexity equivalent to supercomputers in reasonable time will be termed personal sized supercomputers. A research project that has successfully constructed a personal-sized supercomputer from minicomputer system components will be discussed along with the computational model and message-passing algorithm representation required for implementation on the minicomputer system and future multiprocessing systems. Several other projects for developing personal-sized systems exist, and the successful use of such systems for significant computational projects will depend in large measure on the ability to efficiently implement the same numerical algorithm on several different computers or architectures. The message-passing style developed for the minicomputer system is suggested as a sound programming tool to achieve the required degree of architecture independence.
Unable to display preview. Download preview PDF.
- Thompkins, W.T. and Haimes, R., “A Minicomputer/Array Processor/Memory System for Large-Scale Fluid Dynamic Calculations,” Symposium on Impact of New Computing Systems on Computational Mechanics, ASME Winter Annual Meeting, Nov. 1983, Boston MAGoogle Scholar
- Cebeci, T. and Smith, A.M.O., ANALYSIS OF TURBULENT BOUNDARY LAYERS, Academic Press, New York, 1974Google Scholar
- Norton, R.J.G., Thompkins, W.T., and Haimes, R., “Implicit, Finite Volume Schemes with Non-simply Connected Grids: A Novel Approach,” AIAA paper 84–003, presented at AIAA 22nd Aerospace Sciences Meeting, Jan. 1984, Reno, Nevada.Google Scholar
- Hoare, C.A.R., “Communicating Sequential Processes,” Comm. ACM, August 1978, pp 666–677Google Scholar
- Thompkins, W.T. and Dirks, P.W., “The Nemesis System, Simultaneous Simulation of Computer Architecture, Numerical Algorithms, and Operating Systems, MIT report CFDL-TR-84–5, September 1984Google Scholar
- Nicholson, J.H., Forest, A.E., Oldfield, M.L.G., and Schultz, D.L., “Heat Transfer Optimised Turbine Rotor Blades-An Experimental Study Using Transient Techniques,” ASME 82-GT-304, 1982.Google Scholar