next up previous
Next: ``Hybrid'' Parallelism Up: Parallel Cluster Algorithms Previous: Independent Parallelism

Independent Monte Carlo Simulations

This is very easy to do. Just set up different random number streams on each processor (which is done in any case on a parallel machine, but here different processors could be different workstations in a network), set up different random initial configurations on each processor, and then run completely independent simulations on each processor. This gives a parallel efficiency of 100%!

This simple idea works very well for cluster algorithms, as long as the lattice is small enough that all the data will fit into the memory of a single processor. This was a problem for early parallel machines, but is usually no problem for workstation networks and machines like the IBM SP2. Note that this only works for MIMD machines, since cluster algorithms are irregular.



Paul Coddington, Northeast Parallel Architectures Center at Syracuse University, paulc@npac.syr.edu