1 / 6

Fine-Grain Concurrency Widespread in HPEC platforms

Multicore, Multithreaded, Multi-GPU Kernel VSIPL Standardization, Implementation, & Programming Impacts Anthony Skjellum, Ph.D. http://www.runtimecomputing.com. Fine-Grain Concurrency Widespread in HPEC platforms. SMP (not new) Multicore (e.g., Core i7, 8641D)

ethel
Télécharger la présentation

Fine-Grain Concurrency Widespread in HPEC platforms

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. Multicore, Multithreaded, Multi-GPU Kernel VSIPL Standardization, Implementation, & Programming Impacts Anthony Skjellum, Ph.D. http://www.runtimecomputing.com

  2. Fine-Grain Concurrency Widespread in HPEC platforms • SMP (not new) • Multicore (e.g., Core i7, 8641D) • Hyperthreading and other weaker internal core concurrency • SMP VxWorks (tasks vs. POSIX processes… use case new, issue not new) • Customers are demanding standards-based approaches to programming VSIPL in multicore environments • Thread-safe VSIPL increasingly in demand

  3. Thread Safe User-Level Portability must be maintained Users may need to control/hint use of concurrency internally in a library vs. their own task concurrency Affinity issues and choices arise Single-threaded users don’t want to pay for overhead of multithreaded models VSIPL APIs are a good start

  4. Model VSIPL Model MT1 Rules Every thread that uses VSIPL calls vsip_init() and vsip_finalize() Every thread works with independent objects and memory All existing VSIPL syntax is valid No new VSIPL syntax is needed or provided Pthreads (POSIX) as threading library High quality implementations allow internal concurrent execution of user threads Low overhead

  5. What’s in the poster Why Thread-safe VSIPL important A practical multithreaded programming model for VSIPL defined (MT1) An example of MT1-based FFT example Explanations of issues, concerns, and barriers going to multithreaded Discussion of GPU-related (offload-engine) related issues

More Related