Output details
11 - Computer Science and Informatics
University of Edinburgh
Adaptive structured parallelism for distributed heterogeneous architectures: a methodological approach with pipelines and farms
<08> Originality: The use of extracts from the actual workload to calibrate and recalibrate dynamically is novel in the context of skeleton-based grid programming.
Significance: Distributed heterogeneous shared HPC resources are difficult to program efficiently. Our case study demonstrates that a pattern-oriented programming model, supported by a self-adaptive run-time, can address this challenge, all from behind a simple API. This opens up the field for the investigation of further patterns and applications.
Rigour: Experiments involve real applications on real distributed systems.