📄 http:^^www.cs.yale.edu^html^yale^cs^hyplans^deshpande-ashish.html
字号:
Date: Thu, 21 Nov 1996 23:47:58 GMTServer: NCSA/1.4.1Content-type: text/htmlLast-modified: Wed, 24 Jan 1996 20:59:40 GMTContent-length: 7271<head><TITLE>Ashish Deshpande's Home Page</TITLE><H1><!WA0><img align=middle src="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/home.gif">Ashish's Home Page</H1></head><body><h1>(deshpande-ashish@cs.yale.edu)</h1><p><HR><P><h2><!WA1><img src="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/redball.gif"> Address</h2><address> Yale University Department of Computer Science<br> P.O. Box 208285<br> New Haven, CT 06520-8285<br> Tel: (203)-432-1203<br> Fax: (203)-432-0593<br></address><p><HR><p><!WA2><img src="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/at_work_icon.gif">This page is still under construction.<p><HR><p><h2><!WA3><img src="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/greenball.gif">Personal Information</H2>I'm a graduate student in the <!WA4><a href="http://www.cs.yale.edu/HTML/YALE/CS/FrontDoor.html">Department of Computer Science</a> at<!WA5><a href="http://www.cs.yale.edu/HTML/YALE/FrontDoor.html"> YaleUniversity</a>. I expect to get a Ph.D. in Computer Science in May 1996. I'm from <!WA6><a href="http://spiderman.bu.edu/misc/india/">India</a> and movedto the US in 1988. I hold a Bachelor's degree in Computer Science and Engineeringfrom the Indian Institute of Technology, Bombay and an MS in Computer Science from the<!WA7><a href="http://www.virginia.edu">University of Virginia</a>.<p>I'm married to <!WA8><a href=http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/golf.html> Rashmi</a> and we have a baby boy <!WA9><a href=http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/rohan.html>Rohan.</a> Here's some<!WA10><a href=#cool>cool stuff</a> I enjoy.<p><HR><p><h2><!WA11><img src="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/orangeball.gif"> Research</H2><ul><h3><li>Dissertation Research</h3>We investigate a novel procedure for the numerical solution ofscalar partial differential equations, which are in conservation form.Such equations are most commonlysolved on a computer by approximating the derivatives using finitedifferences and advancing the solution forward in time using either anexplicit or an implicit method.Explicit schemes have the advantage that each time step is very fastto compute. They are also highly parallelizable. However,stability requirements necessitate a very small time step, whichincreases the total amount of work unacceptably.Implicit schemes remove the necessity of taking a very small timestep. However, each time step is far more expensive in terms ofcomputer time. Furthermore, the computation process is not easilyparallelizable.<p>In this thesis, we propose a new procedure intermediate between theexplicit and implicit schemes described above. It results frommodifying the explicit method using the Gauss-Seidel principle. Wesimply use the latest available information at any point in thecomputation process. We retain the computational efficiencyof the explicit scheme, and at the same time obtain a stabilityrequirement that imposes an acceptable restriction on the length ofthe time step. However, the truncation error is of a form that againimposes an unacceptable restriction on the length of the time step. Wecircumvent this difficulty by using one or more additional sets ofapproximations to cancel out the undesirable terms in the truncationerror. The new procedure is not as easily parallelized as the explicitmethod but is more parallel than the implicit method. Hence, itscharacteristics lie between those of the explicit and implicit methods.We use the new method to solve the 1,2 and 3 dimensional Burgers'equation, which is frequently used as a model for the incompressibleNavier-Stokes equations. We present proofs ofstability along with experimental results thatvalidate the theory.<p>Details in Postscript format (these are incomplete drafts I'm currentlyworking on)<ul><li> <!WA12><a href="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/conslaw.ps"> Numerical solution of conservation laws</a> (169 K)<li> <!WA13><a href="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/burgers.ps"> Parallel numerical solution of Burgers' equation </a> (643 K)</ul><h3><li>Time Domain Parallelism</h3>Time dependent PDE's are often solved by parallelizing the solutionprocess in the spatial domain. This often limits efficiency as thenumber of processors increases. In this paper, we examine a classof algorithms, which incorporate parallelism in the temporal domainto increase efficiency. We discuss conditions under which thesealgorithms may be useful.<p>A Postscript version can be found<!WA14><a href="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/norms.ps">here</a> (133 K).<h3><li>Parallel CFD Simulations</h3>A general purpose, 3D, incompressible Navier-Stokes algorithm is implementedon a network of workstations using PVM. Performance comparisons with theCray Y-MP and Intel IPSC/860 are made.<p>A Postscript version can be found<!WA15><a href="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/reno93.ps">here</a> (395 K).<h3><li>Efficient Parallel Programming with Linda</h3>We discuss C-Linda's performance in solving a representative scientificcomputing problem, the shallow water equations, and make comparisons withalternatives available on various shared and distributed memory machines.<p>A Postscript version can be found<!WA16><a href="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/p1.ps">here</a> (173 K).<h3><li>Biological Sequence Comparison</h3>We describe a general, efficient and flexible platform for searching biologicalsequence databases that runs on Intel hypercubes. It can easily be adapted toevaluate a wide variety of sequence comparison algorithms and shows close tooptimal performance characteristics.<p>I don't have a Postscript version :-(</ul><p><hr><p><h2><!WA17><img src="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/whiteball.gif"><a name="technical">Other Technical Interests</a></h2><ul><li>Check out Intel's fantastic <!WA18><a href="http://www.intel.com/procs/p6/index.html">P6 processor presentation</a>.<li>Info on tons of CPU's at the<!WA19><a href="http://infopad.eecs.berkeley.edu/CIC/">CPU Info Center</a><li>Here's a very nice<!WA20><a href="http://gams.nist.gov/">GAMS : Guide to Available Mathematical Software</a>.<li>Lots of MPI stuff is<!WA21><a href="http://www.mcs.anl.gov/Projects/mpi/index.html">here</a>.</ul><p><HR><p><h2><!WA22><img src="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/purpleball.gif"><a name="cool">Cool Stuff</a></H2><ul><li><!WA23><a href="http://www.msstate.edu/Movies/">Movies</a><li><!WA24><a href="http://humper.student.princeton.edu/floyd/">Pink Floyd</a><li>Sci-fi (esp. Star Trek)<li><!WA25><a href="http://www.cis.ohio-state.edu/hypertext/faq/usenet/radio/top.html">World band radio</a><li>15-minute delayed quotes can be found at the<!WA26><a href="http://www.secapl.com/cgi-bin/qs">Security APL Quote Server</a><li> Check out the <!WA27><a href="http://update2.wsj.com/"> Money and Investing Update</a>from the Wall Street Journal.<li> Here's <!WA28><a href="http://pathfinder.com/@@tqbCs2GwFgMAQDFR/money/"> Money Magazine</a>on the Internet.<li>Stock Market Data from MIT is <!WA29><a href="http://www.ai.mit.edu:80/stocks/">here</a>.<li>Deep discount broker services from <!WA30><a href="http://www.etrade.com">E*Trade</a>.(This is not an endorsement). Also has delayed stock quotes.</ul><p><HR><p><h2><!WA31><img src="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/yellowball.gif"><a name="sports">Sports</a></h2><ul><li><!WA32><a href="http://www.cs.yale.edu/HTML/YALE/CS/HyPlans/deshpande/golf.html"> Golf</a><li>Lots of cool sports stuff at <!WA33><a href="http://pathfinder.com/si/greet.html">Sports Illustrated</a>.<li>Check out Jacqueline Dupree's fabulous <!WA34><a href="http://www.access.digex.net/~jd/tennis.html">Site that the ATP hates</a>.<li>Here's the home page of the <!WA35><a href="http://atptour.com/"> Tennis ATP Tour</a>.<li><!WA36><a href="http://sunsite.unc.edu/drears/running/running.html"> Running</a></ul></body>⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -