1 / 7

Exploitation of modern CPU architectures and its impact on the computing model of HEP experiments

Exploitation of modern CPU architectures and its impact on the computing model of HEP experiments. Goals of the Workshop. Review of current industrial technologies Hardware Software Trend for the future Short term Next decade Review of current activities in HEP. Presentations.

retha
Download Presentation

Exploitation of modern CPU architectures and its impact on the computing model of HEP experiments

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. Exploitation of modern CPU architectures and its impact on the computing model of HEP experiments

  2. Goals of the Workshop • Review of current industrial technologies • Hardware • Software • Trend for the future • Short term • Next decade • Review of current activities in HEP

  3. Presentations • Reviews (hardware, software, activities) • Vincenzo: Multi-Core • Leone: GPUs • Specific topics • Alfio: “final” physics analysis • Gerri: I/O (root) and analysis of large datasets • Karen: physics analysis on GPUs

  4. Goals of the Workshop Exploitation of modern CPU means parallelization • New Evaluate impact on • SuperB computing model • Software Architecture • Identify the most affected areas • Propose a short term R&D program

  5. Issues for discussion R&D • Optimization of scalar code • It will not go faster because of technology • We need to go back to trading accuracy for speed • Vectorization (in core parallelism) • Compiler level auto-vectorization is not enough • We may need to rethink algorithms to exploit it Challenge here is to achieve a significant speed-up while keeping the code flexible and readable/maintainable by a large group of people

  6. Issues for discussion R&D • High grain parallelism (inside algorithms) • Many paradigms around (OMP, TBB, Cilk++, Cl, std::thread,…) • We need to evaluate the various approaches and see which best fit our environment • Try with existing algorithms (pattern recognition and tracking for instance)

  7. Issues for discussion R&D • Framework level parallelism • Beyond the well established event by event parallelism • Algorithm pipelines, parallel I/0 • Needs to establish the actual gain of different scheduling strategies • Leverage Babar’s software and data to build “emulators” • Keep an eye on emerging technology • GPGPU: established for specialized use-case, can be used for general computing? • Experiment with (HL)T algorithms

More Related