Connection machine

Beazley D M and Lomdahl P S 1993 Message-passing multi-cell molecular dynamics on the Connection Machine 5 Parallel Comput. 20 173-95... 

The massively parallel approach adopted in the Connection Machine has been termed data parallel. Whereas a uniprocessor must sequentially step through large amounts of data, a data parallel machine moves processors to the data. Aggregate memory to processor bandwidth in the Connection Machine is more than 700 megabytes per second. 

B. Murray, P. Bash, and M. Katplus, Molecular Dynamics on the Connection Machine, Thinking Machines Cotp., Cambridge, Mass., 1989. 

A.C. Robinson, C.T. Vaughan, H.E. Fang, C.F. Diegert, and K. Cho, Hydrocode Development on the nCUBE and the Connection Machine Hypercubes, Proceedings of the 1991 APS Topical Conference on Shock Compression of Condensed Matter, Williamsburg, VA, 1991. 

Whaley, R. and Hodes, L. Clustering a large number ofcompounds. 2. Using the Connection Machine. J. Chem. Inf. Comput. Sci. 1991,31,345-347... 

Reczko, M Hatzigeorgiou, A., Mache, N., Zell, A. Suhai, S. (1995). A parallel neural network simulator on the connection machine CM-5. ComputAppl Biosci 11,309-15. 

Windemuth and Schulten described a further variant of the systolic loop algorithm, the replicated systolic loop for the efficient calculation of macromolecular force fields on the Connection Machine (e.g., CM-1, CM-2, CM-5). The full force field was separated into bond interactions and nonbonding interactions only the latter were implemented on the Connection Machine, and parallelized by the replicated systolic loop algorithm. The former, less computationally intensive tasks were performed by an existing, conventional MD code on the front end. 

B. M. Boghosian, Comput. Phys., 4, 14 (1990). Computational Physics on the Connection Machine. 

D. M. Beazley and P. S. Lomdahl, Technical Report LA-UR-92-3158, Los Alamos National Laboratories, Los Alamos, NM, 1992. Message-Passing Multi-Cell Molecular Dynamics on the Connection Machine 5. 

Nov. 12-16, 1990, IEEE Computer Society Press, Los Alamitos, CA, 1990, pp. 748-752. An Optimal Hypercube Direct N-Body Solver on the Connection Machine. 

A. Windemuth and K. Schulten, Mol. Simulation, 5, 353 (1991). Molecular Dynamics Simulation on the Connection Machine. 

D. Chasman, R. J. Silbey, and M. D Mello, Theor. Chim. Acta, 84,335 (1993). Exact Three-Dimensional Time-Dependent Wave-Packet Calculations on the Connection Machine. 

J. Bailey, The Ghosts of Computers Past understanding data parallel architecture in the context of the human computing era, in Scientific Applications of the Connection Machine, H. D. Simon, ed.. World Scientific, 1989, 3-23. 

Hardware. Computers built to work like neural nets are called "parallel processors". A parallel processor uses a large number of small, interconnected processing units rather than a single CPU. Prominent among these is Thinking Machines Corporation s "Connection Machine", which can realize a variety of neural net models. It is programmable in LISP, and is well-suited to database tasks. Hecht-Nielsen produces the ANZA board, a co-processor which allows the PC/AT to emulate a parallel processor. 

Interestingly, the Internet (a major computer network linking millions of machines around the world) is growing at a phenomenal rate. Between January 1993 and January 1994, the number of connected machines grew from 1,313,000 to 2,217,000. More than 70 countries have full Internet connectivity, and about 150 have at least some electronic mail services. As I write this, there are about 20 to 25 million users of the Internet (Goodman et al The Global Diffusion of the Internet, Communications of the ACM, 37[8][1994] 27). 

Consider the material-hemdling configuration where each link (i.j) that connects machine centers i and j(i,j = 1,. . . , m i + j) has a sufficient number of transporters or a conveyor system such that when a job completes processing at machine center i, it can immediately begin movement from machine center i to its destination. The average time taken to transfer a job from machine center i to machine center j is I/mo, ). >J = 1,. m i j. Suppose the job transfers from one machine center to another are represented by a job transfer probability matrix P = (p )/,y=i If we now incorporate the transporters on each link (ij) as a service center, the job transfer probabilities are P ii — Pu< Pi.ii.n Pij P(i.j)j 1> J = 1,. . . , tif, i i= j. All Other transfer probabilities are zero. Here i,j) represents the service center corresponding to the transporters on link = 1,. . . , ... 

---