OpenMP task is the most significant feature in the new specification, which provides us with a way to handle unstructured parallelism. This paper presents a runtime library of task model on Cell heterogeneous multicore, which attempts to maximally utilize architectural advantages. Moreover, we propose two optimizations, an original scheduling strategy and an adaptive cutoff technique. The former combines breadth-first with the work-first scheduling strategy. While the latter adaptively chooses the optimal cut-off technique between max number of tasks and max task recursion level according to application characteristics. Performance evaluations indicate that our scheme achieves a speedup factor from 3.4 to 7.2 compared to serial executions. © Springer-Verlag Berlin Heidelberg 2010.
CITATION STYLE
Cao, Q., Hu, C., He, H., Huang, X., & Li, S. (2010). Support for OpenMP tasks on cell architecture. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 6082 LNCS, pp. 308–317). https://doi.org/10.1007/978-3-642-13136-3_32
Mendeley helps you to discover research relevant for your work.