H. Johansson (Sweden) and J. Steensland (USA)
: Load balancing, structured adaptive mesh re finement, partitioning.
Significantly improving the scalability of large structured adaptive mesh refinement (SAMR) applications is chal lenging. It requires sophisticated capabilities for using the underlying parallel computer's resources in the most effi cient way. This is non-trivial, since the basic conditions for how to allocate the resources change dramatically during run-time due to the dynamics inherent in these applications. This paper presents a first characterization of a hy brid and dynamic partitioner for parallel SAMR applica tions. Specifically, we investigate parameter settings for trade-offs like communication vs. load balance and speed vs. quality. The key contribution is that the characterization shows that the partitioner is able to respond accurately to stimuli from system and application state, and hence adapt to various SAMR scenarios. This potentially reduces the run-time for large SAMR applications.
Important Links:
Go Back
