While a heterogeneous system offers substantial aggregate processing power to exploit parallelism in GLU programs, its processors and links exhibit fluctuating loads and can fail arbitrarily. On such a system, effective exploitation of coarse grain parallelism requires the runtime strategy for load distribution to adapt to system dynamics. In this paper, we describe two eduction-based architectures for implementing GLU on heterogeneous systems that use different approaches to cope with dynamically fluctuating loads, processor failures, as well as processor additions. The first implementation architecture, which is based on multiple distributed generators, deals with fluctuating loads as well as processor changes using an adaptive value-demand distribution function. The second implementation architecture, which is also based on multiple distributed generators, uses dynamic inter-generator term-demand migration to adapt to fluctuating loads as well as processor changes.

• glu-dga.ps