Search

This paper presents an effort to mitigate overheads, latencies, and limitations observed in message-driven runtime frameworks, by utilizing lightweight threads tightly integrated with message passing. It also introduces new abstractions and features for group communication as well as fine-grained concurrency on top of remote method invocations to improve workload balancing in shared and distributed memory. We observe up to 100% difference in performance behavior for task creation and handling message passing. Evaluations on 1000 cores (25 nodes) of a distributed memory machine showed that the integration of fine-grained concurrency with the runtime achieves performance improvements of 12% on a seismic wave simulation benchmark, as opposed to 50% degradation with OpenMP. Moreover, a 3D mesh refinement application showed 50% improvement, exploiting multi-grain parallelism at data and task levels. [ABSTRACT FROM AUTHOR]

Handling the ever-increasing complexity of mesh generation codes along with the intricacies of newer hardware often results in codes that are both difficult to comprehend and maintain. Different facets of codes such as thread management and load balancing are often intertwined, resulting in efficient but highly complex software. In this work, we present a framework which aids in establishing a core principle, deemed separation of concerns, where functionality is separated from performance aspects of various mesh operations. In particular, thread management and scheduling decisions are elevated into a generic and reusable tasking framework. The results indicate that our approach can successfully abstract the load balancing aspects of two case studies, while providing access to a plethora of different execution back-ends. One would expect, this new flexibility to lead to some additional cost. However, for the configurations studied in this work, we observed up to 13 % speedup for some meshing operations and up to 5.8 % speedup over the entire application runtime compared to hand-optimized code. Moreover, we show that by using different task creation strategies, the overhead compared to straight-forward task execution models can be improved dramatically by as much as 1200 % without compromises in portability and functionality. [ABSTRACT FROM AUTHOR]