1 / 4

Innovative Programming Models for Asymmetric Multi-Core Systems

This paper explores the potential of asymmetric multi-core systems for data-intensive applications, highlighting the benefits of leveraging multiple dimensions of parallelism. It examines the role of heterogeneity and application-specific accelerators, as well as ample memory bandwidth and software control flexibility. While acknowledging limitations, it introduces new concepts such as polymorphic parallelism, event-driven scheduling, and unification of execution modes. The work presents models for multi-grain concurrency and runtime adaptation, showcasing the integration of DMA and communication parallelism, with results from various Cell BE system software modules.

norah
Télécharger la présentation

Innovative Programming Models for Asymmetric Multi-Core Systems

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Synthesizing Parallel Programming Models for Asymmetric Multi-core Systems Dimitris Nikolopoulos, Kirk Cameron Center for High-End Computing Systems Department of Computer Science Virginia Tech {dsn,cameron}@cs.vt.edu

  2. Asymmetric Multi-Core Systems • Potential for data-intensive applications • Multiple dimensions of parallelism • Heterogeneity, application-specific accelerators • Ample memory bandwidth • Flexibility via delegating control to software • Limitations • Programming models ignore polymorphic parallelism • Runtime systems ignore polymorphic parallelism

  3. Contributions • Polymorphic parallelism on the Cell BE • Event-driven scheduling • Unification of task, data, pipeline parallel execution • Integration of DMA and communication parallelism • Runtime performance prediction • Model of multi-grain concurrency • Model-driven adaptation in the runtime system • Cell system software modules • EDTLP, MMGP

  4. Results

More Related