1 / 8

The Source-Sink Model: A DataFlow Implementation (work in progress)

University of Belgrade School of Electrical Engineering (ETF) A project co-funded by Serbian Government and SMEs. The Source-Sink Model: A DataFlow Implementation (work in progress). Marko Stanković, Siniša Ivković (ETF) Radomir Radojičić, Luka Ilić (Meteos). Contents. Eta model

leane
Download Presentation

The Source-Sink Model: A DataFlow Implementation (work in progress)

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. University of BelgradeSchool of Electrical Engineering (ETF)A project co-funded by Serbian Government and SMEs The Source-Sink Model:A DataFlow Implementation(work in progress) Marko Stanković, Siniša Ivković (ETF) Radomir Radojičić, Luka Ilić (Meteos)

  2. Contents • Eta model • Source-sink model • Method of implementation • Results so far • Next Steps

  3. The Eta Model • State-of-the-art atmospheric model • Developed in the ’70s in the ex-YU • η (eta) – the vertical coordinate • ηs = (pref(zs)-pt) / (pr(z=0)−pt) • Horizontal coordinates: Arakawa E-grid • Preventing pressure-gradient force errors due to steep topography

  4. The Source-Sink Model • Shallow-water equations model • Defined on Arakawa E-grid • Process of geostrophic adjustment • Four main components: • Continuity equation (Mesinger, Janjić) • Updating the height • Equation of motion • Updating the speed

  5. Method of Implementation • Move all the calculation to the DFE • All data stored on DFE • Initial iteration is streamed from host • Its results go to RAM on card • Iterations are done locally • Final iteration’s results streamed to host

  6. Results so Far • Loop calculating four variables • Same execution time on host and DFE • After implementing more code • Expected DFE exec time: Almost constant • Expected host exec time: Rises linearly More code -> More speed-up!

  7. Next Steps (due: March 31, 2013) • Finishing the implementation • Performing the final measurements Future Work: • Arbitrary number of sources and sinks

  8. Thank you!Questions?Marko Stanković (stankovic.marko@gmail.com)Siniša Ivković (sinisa.ivkovic@gmail.com)Radomir Radojičić (rrbata@meteos.rs)Luka Ilić (luka.ilic@sewa-weather.com)

More Related