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/

Global: Zendy Tools annual

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

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

The trend towards highly parallel multi-processing is ubiquitous in allmodern computer architectures, ranging from handheld devices to large-scale HPCsystems; yet many applications are struggling to fully utilise the multiplelevels of parallelism exposed in modern high-performance platforms. In order torealise the full potential of recent hardware advances, a mixed-mode betweenshared-memory programming techniques and inter-node message passing can beadopted which provides high-levels of parallelism with minimal overheads. Forscientific applications this entails that not only the simulation code itself,but the whole software stack needs to evolve. In this paper, we evaluate themixed-mode performance of PETSc, a widely used scientific library for thescalable solution of partial differential equations. We describe the additionof OpenMP threaded functionality to the library, focusing on sparsematrix-vector multiplication. We highlight key challenges in achieving goodparallel performance, such as explicit communication overlap using task-basedparallelism, and show how to further improve performance by explicitly loadbalancing threads within MPI processes. Using a set of matrices extracted fromFluidity, a CFD application code which uses the library as its linear solverengine, we then benchmark the parallel performance of mixed-mode PETSc acrossmultiple nodes on several modern HPC architectures. We evaluate the parallelscalability on Uniform Memory Access (UMA) systems, such as the FujitsuPRIMEHPC FX10 and IBM BlueGene/Q, as well as a Non-Uniform Memory Access (NUMA)Cray XE6 platform. A detailed comparison is performed which highlights thecharacteristics of each particular architecture, before demonstrating efficientstrong scalability of sparse matrix-vector multiplication with significantspeedups over the pure-MPI mode.

Benchmarking mixed-mode PETSc performance on high-performance architectures