http:^^http.cs.berkeley.edu^~ssmith^gator.html

来自「This data set contains WWW-pages collect」· HTML 代码 · 共 154 行

HTML
154
字号
Date: Wed, 20 Nov 1996 19:25:33 GMTServer: NCSA/1.4Content-type: text/htmlLast-modified: Tue, 17 Sep 1996 13:47:47 GMTContent-length: 5698<Title> GATOR </Title><H1>GATOR</H1><P> This work is part of an <!WA0><a href = "http://www.netlib.org/nse/home.html">HPCC </a> project funded by <!WA1><a href = "http://www.netlib.org/nse/curr_hpcc.html"> NASA </a>  to develop astate-of-the-art Earth System Model (ESM) that will be comprised of a coupledatmosphere and ocean system including chemical tracersthat are found in, and may be exchanged between, the atmosphereand the oceans.  The starting point of the ESM model is the<!WA2><a href = "http://www.atmos.ucla.edu/~vwk206/esm.html"> UCLA AGCM</a>, which has been parallelized on different platforms andwill be coupled with GATOR, a Gas, Aerosol, Transport, andRadiation Chemistry model, developed by M. Jacobson, O. Toon, R. Turco, and R. Lu.  The UCLA ESM model will archive and retrieve model output via the<!WA3><a href = "http://s2k-ftp.cs.berkeley.edu:8000/cgi-bin/imagemap/s2khome?357,117"> Sequoia </a> database.  Coupled systems are an important tool for helping scientiststo understand complex phenomenasuch as El Nino and stratospheric ozone depletion.  <P>A parallel version of GATOR is being developed at Berkeley by<!WA4><a href = "http://http.cs.berkeley.edu/~demmel"> Jim Demmel</a>and <!WA5><a href = "http://http.cs.berkeley.edu/~ssmith"> Sharon Smith </a>.GATOR modelsatmospheric chemistry in the Los Angeles Basin, and has been used fordetailed air pollution studies.  Our task is to parallelize GATOR andscale it to the globe.<P>GATOR includesboth gases and aerosols, modeling<DL><DD>gas-phase chemistry; <DD>aqueous-phase chemistry; <DD>radiation transfer;<DD>aerosol nucleation, coagulation, growth and evaporation; <DD>horizontal advection and vertical convection;<DD>dry and wet deposition; <DD>visibility and emissions. </DL><P><H2> The Parallel Version of GATOR </H2><P>A primary motivation for this research is to provide atmospheric scientistswith  the best computational means to further their studies of theearth's future climate.  Since the UCLA ESM is still evolving,it is important that the codes also be portable, so our efforts  are thus focused on developing portable, parallel code thatcan be performance tuned for different parallel architectures.<P>The main challenges in parallelizing GATOR areovercoming problems due to load imbalance and minimizing the communication costs.   The figure below showsthe differences in computation that can result in loadimbalance:<P><A NAME=205>&#160;</A><A NAME=Figglob>&#160;</A><!WA6><IMG  ALIGN=BOTTOM ALT="" SRC="http://http.cs.berkeley.edu/~ssmith/SC95/SP2_Load_3d.gif"><P>In the picture,the z-axis shows time, while the x and y axis show thelatitudinal and longitudinal partitioning of the atmosphere intoblock columns.  Each processor owns one set of block columns.  Thedifference in time among the processors illustrates the amountof computation due to solving ODEs that arise from chemicalkinetic equations.  The largest amount of computation occurs inparts of the globe that are in the summer and daylight, whereas the processors holding atmospheric cells in winter and night     require the least amount of computation.  Click <!WA7><a href = "http://http.cs.berkeley.edu/~ssmith/world.gif">here</a> to see the ODE solver computation costs for the world.<P>Our solution to the load balancing problem is to use a block-cycliclayout which will collect different atmospheric cells from the different parts of the globe together in a single processor.  Thehope is that each processor will have cells requiring different amounts of computation, but not all of the computationallyintensive cells at once.   The pictures below illustrate a blocklayout, which does not prevent against load imbalance, and a block-cyclic layout.<P><A NAME=204>&#160;</A><A NAME=Figpipeline>&#160;</A><!WA8><IMG  ALIGN=BOTTOM ALT="" SRC="http://http.cs.berkeley.edu/~ssmith/mapping_all.gif"><H2>Performance Modeling and the ESM</H2>Throughout the design of the parallel implementation of GATOR,we used a timing model to help choose between differentdesign alternatives that would benefit GATOR both as a stand-alonesystem and as part of the UCLA AGCM.The results of this work are reported in: <P><DL><DT><A NAME=DemmelSmith><STRONG></STRONG></A><DD>J. Demmel and S. L. Smith.<!WA9><a href="http://http.cs.berkeley.edu/~ssmith/shpcc.ps">Parallelizing a global atmospheric chemical tracer model</a>.In <em> Proceedings of the Scalable High Performance ComputingConf.</em>, pages 718--725, Knoxville, TN., May 1994.</DL><P><DL><DD>J. Demmel and S. L. Smith.<!WA10><a href="http://http.cs.berkeley.edu/~ssmith/SC95/paper.html"> Performance of a parallel global atmospheric chemical tracer model.</a>Submitted to <em> Supercomputing 95 </em>.</DL><P><H2>EOSDIS</H2><p>We are extending our modeling effort to help explore an architecturefor the <!WA11><A HREF="http://eos.nasa.gov"> Earth Observing System (EOS)</a>. This  effort is part of <!WA12><A HREF="http://epoch.CS.Berkeley.EDU:8000/nasa_e2e/">"End-to-End Problems in EOSDIS"</A>, a NASA-sponsored multi-year projectto investigate alternative data management strategies for the EOS. The project involves researchers at the <!WA13><a href="http://http.CS.Berkeley.EDU">Berkeley</a>, <!WA14><a href="http://www.cs.ucla.edu">Los Angeles</a>,<!WA15><a href="http://www-cse.ucsd.edu:80/users/pasquale/ResearchProjects/EOSDIS.html">San Diego</a>, and <!WA16><a href="http://www.icess.ucsb.edu/eosdis/eosdis.html">Santa Barbara</a>campuses of the University of California. <p>For some sample views of environmental and climate data,courtesy of <!WA17><a href="http://www.icess.ucsb.edu/eosdis/eosdis.html">Jeff Dozier</a> at UC Santa Barbara, click<!WA18><a href="http://http.cs.berkeley.edu/~ssmith/colorfigs.ps"> here</a>.<HR>last updated on December 12, 1995.<ADDRESS>ssmith@CS.Berkeley.EDU</ADDRESS>

⌨️ 快捷键说明

复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?