eaton-watts
Uploaded by
2 SLIDES
165 VUES
20LIKES

CMAQ Parallel Performance

DESCRIPTION

This document summarizes the advancements in the CMAQ Sept 2003 release, focusing on parallel performance improvements utilizing a 118x118 xy grid with 21 z-layers across 84 species. Notable upgrades include a new version tested on a dual Xeon cluster, achieving significant reductions in computation time—1 day in just 6.7 hours and 1 year in 102 days. Key enhancements involve alternative numerical methods for aerosol computations, an improved Newton-Raphson solver for chemistry, and optimized I/O operations, facilitating improved performance for both serial and parallel processing.

1 / 2

Télécharger la présentation

CMAQ Parallel Performance

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. CMAQ Parallel Performance • CMAQ Sept 2003 release, cb4_ae3_aq chemical mechanism • 118 x 118 xy grid (2 km), 21 z-layers, 84 species • 24-hour run, output every 2 hrs • HP Linux cluster, 3.0 GHz Xeon (dual-proc), 4 Gb/node, Myrinet, PVFS • New version on 1 proc: 1 day in 6.7 hrs 1 year in 102 days • New version on 32 procs: 1 day in 26 min 1 year in 6.8 days Original full Computation Original write Modified write Modified full

  2. Performance Improvement Ideas May affect serial or parallel performance: Aerosol: alternative to numerical quadrature Chemistry: try damped Newton-Raphson solver Vdiff: array orderings for cache effects Xadv/Yadv: alternate formulation of PPM for reduced communication I/O: binary vs NetCDF (10x) multiple procs read & write pre- and post-processing of files 6.7 hrs 40 min

More Related