Mapping Monte Carlo-Metropolis algorithm onto a double ring architecture

Mathematical models are frequently used to evaluate the physical properties of matter; they involve intensive computation: processing times and loan costs soar, using general purpose computers. Remarkable improvements are promised by special purpose architectures, in particular for the parallel execution of the most time consuming routines; the Monte Carlo-Metropolis method has been examined, and parallelization approaches have been identified. A double ring architecture of processing elements (PE) has been conceived, which optimizes efficiency, yet saves the algorithm principles. A network of 16 purposely built PEs has been assembled, and modeling programs have been run on it. The main architectural choices, both for hardware and software are described; they refer to the computing system and to each PE as well; preliminary results are also given