Streamware

摘要: Recently, the number of cores on general-purpose processors has been increasing rapidly. Using conventional programming models, it is challenging to effectively exploit these for maximal performance. An interesting alternative candidate multiple stream model, which provides a framework writing programs in sequential-style while greatly simplifying task automatic parallelization. It shown that not only traditional media/image applications but also more data-intensive can be expressed style.In this paper, we investigate potential use model efficiently utilize commodity multicore (e.g., Intel/AMD). Although several languages and compilers have recently developed, they typically target special-purpose processors. In contrast, propose flexible software system, Streamware, automatically maps onto wide variety processor configurations. We leverage existing compilation design runtime environment takes as input output form machine-independent virtual machine code. The assigns work considering processor/cache configurations adapts workload variations. evaluate approach few scientific real hardware cycle-level simulator set-up showcase scaling contention issues. results show good choice exploiting modern future CPUs an important class applications.