Performance of a GALS Single-Chip Multiprocessor

摘要

The performance of a single-chip shared-memory multiprocessor built with a globally-asynchronous locally-synchronous (GALS) design strategy is explored. The performance of the multiprocessor is found to be nearly as good as that of a globally-clocked equivalent multiprocessor. Since the GALS strategy results in a much more easily scaled design and a large reduction in design effort for the multiprocessor's clock tree, the GALS methodology appears to be an attractive alternative to the standard globally-clock methodology.