English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Scientific applications that run on leadership computing facilities oftenface the challenge of being unable to fit leading science cases ontoaccelerator devices due to memory constraints (memory-bound applications). Inthis work, the authors studied one such US Department of Energymission-critical condensed matter physics application, Dynamical ClusterApproximation (DCA++), and this paper discusses how device memory-boundchallenges were successfully reduced by proposing an effective "all-to-all"communication method -- a ring communication algorithm. This implementationtakes advantage of acceleration on GPUs and remote direct memory access (RDMA)for fast data exchange between GPUs.  Additionally, the ring algorithm was optimized with sub-ring communicatorsand multi-threaded support to further reduce communication overhead and exposemore concurrency, respectively. The computation and communication were alsoanalyzed by using the Autonomic Performance Environment for Exascale (APEX)profiling tool, and this paper further discusses the performance trade-off forthe ring algorithm implementation. The memory analysis on the ring algorithmshows that the allocation size for the authors' most memory-intensive datastructure per GPU is now reduced to 1/p of the original size, where p is thenumber of GPUs in the ring communicator. The communication analysis suggeststhat the distributed Quantum Monte Carlo execution time grows linearly assub-ring size increases, and the cost of messages passing through the networkinterface connector could be a limiting factor.

Memory Reduction using a Ring Abstraction over GPU RDMA for Distributed Quantum Monte Carlo Solver