A power-aware GALS architecture for real-time algorithm-specific tasks

We propose an adaptive scalable architecture suitable for performing real-time algorithm-specific tasks. The architecture is based on the globally asynchronous and locally synchronous (GALS) design paradigm. We demonstrate that for different real-time commercial applications with algorithm-specific jobs like online transaction processing, Fourier transform etc., the proposed architecture allows dynamic load-balancing and adaptive inter-task voltage scaling. The architecture can also detect process-shifts for the individual processing units and determine their appropriate operating conditions. Simulation results for two representative applications show that for a random job distribution, we obtain up to 67% improvement in MOPS/W (millions of operations per second per watt) over a fully synchronous implementation.