Swendsen-Wang Monte Carlo Simulation of Spin Models

These programs do a Monte Carlo simulation of the two dimensional ferromagnetic Ising spin model. This is a model of magnetic materials that is also widely used to study the properties of phase transitions.

The programs use the non-local cluster Monte Carlo algorithm of Swendsen and Wang, Phys. Rev. Lett. 58, 86 (1987). The Swendsen-Wang algorithm, is difficult to parallelize efficiently, since it is highly irregular and requires a lot of non-local communication. The basic computational task is connected component labeling.

For the message-passing (MIMD) version we have used the local label propagation (or ``self-labeling'') algorithm, however we may implement other algorithms in the future. For details of these algorithms, see the papers:

• Caltech technical report C3P-862, P. Coddington and C. Baillie, "Cluster Algorithms for Spin Models on MIMD Parallel Computers", Proc. of the 5th Distributed Memory Computing Conference, Charleston (April 1990), eds. D. Walker and Q. Stout (IEEE Computer Society Press, Los Alamitos, California, 1990),
• Caltech technical report C3P-855, C. Baillie and P. Coddington, "Cluster Identification Algorithms for Spin Models - Sequential and Parallel", Concurrency: Practice and Experience 3, 129 (1991),

For a description of the SIMD labeling algorithms, including the Get/Send algorithm that is used in the CMF/HPF codes, see the paper:

• NPAC technical report SCCS-279, J. Apostolakis, P. Coddington and E. Marinari, "New SIMD Algorithms for Cluster Labeling on Parallel Computers", Int. J. Mod. Phys. C 4, 749 (1993).

F77	Sequential Fortran 77 code for Swendsen-Wang algorithm. 

F77mp	F77 + message passing code for parallel Swendsen-Wang 
	algorithm on MIMD parallel computers, using the Express 
	communications library. Local label propagation is used 
	for the connected component labeling. 

CMF	CMFortran code for parallel Swendsen-Wang algorithm
	on the Connection Machine. Get/Send algorithm is used 
	for the connected component labeling.