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 Plus annual

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

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Hardware heterogeneity is here to stay for high-performance computing.Large-scale systems are currently equipped with multiple GPU accelerators percompute node and are expected to incorporate more specialized hardware. Thisshift in the computing ecosystem offers many opportunities for performanceimprovement; however, it also increases the complexity of programming for sucharchitectures. This work introduces a runtime framework that enables effortlessprogramming for heterogeneous systems while efficiently utilizing hardwareresources. The framework is integrated within a distributed and scalableruntime system to facilitate performance portability across heterogeneousnodes. Along with the design, this paper describes the implementation andoptimizations performed, achieving up to 300% improvement on a single deviceand linear scalability on a node equipped with four GPUs. The framework in adistributed memory environment offers portable abstractions that enableefficient inter-node communication among devices with varying capabilities. Itdelivers superior performance compared to MPI+CUDA by up to 20% for largemessages while keeping the overheads for small messages within 10\%.Furthermore, the results of our performance evaluation in a distributed Jacobiproxy application demonstrate that our software imposes minimal overhead andachieves a performance improvement of up to 40%. This is accomplished by theoptimizations at the library level as well as by creating opportunities toleverage application-specific optimizations like over-decomposition.

Runtime Support for Performance Portability on Heterogeneous Distributed Platforms