1 / 4

How to Rev Up Parallel Algorithms and Many-Core Architectures?

How to Rev Up Parallel Algorithms and Many-Core Architectures?. Uzi Vishkin. State of the Art. - Since ~2005 parallel processing is the only way for improving performance of an application.

bdaniels
Télécharger la présentation

How to Rev Up Parallel Algorithms and Many-Core Architectures?

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. How to Rev Up Parallel Algorithms and Many-Core Architectures? Uzi Vishkin

  2. State of the Art - Since ~2005 parallel processing is the only way for improving performance of an application. "We are dedicating all of our future product development to multicore designs“-- Paul Otellini, CEO, Intel, 2005. - In 2005: Intel expected coupling 1000 cores on-chip by 2015. - But, where are we in 2019? General success stories • iGPUs (Intel’s integrated GPU) • Dedicated GPUs (NVIDIA) Deprecated after major investment since 2005 • Many-Integrated-Core, known as MIC or Xeon Phi (Intel) So revving up is important and timely

  3. Possible Explanation What went right? • Computer graphics is a much needed application in the general-purpose CPU space. Hence: iGPU • Deep learning has become a “killer app” for dedicated GPUs What went wrong • MIC failed to grow killer apps • Possibly because the architecture made it too challenging to develop apps for it; i.e., it was too difficult to program effectively the architecture

  4. Conclusion • Research question: develop a killer-app for a many-core architecture • The architecture can be • An off-the-shelf one, or • A research-stage one such as the UMD XMT architecture that my team developed. See http://users.umiacs.umd.edu/~vishkin/XMT/index.shtml • This is much more challenging than meets the eye • I can discuss with you candidate killer apps, but will also be glad to hear your suggestions

More Related