📄 http:^^www.crpc.rice.edu^fortran-tools^fortran-tools.html
字号:
Date: Tue, 14 Jan 1997 19:21:58 GMT
Server: NCSA/1.5.1
Last-modified: Thu, 21 Nov 1996 18:45:06 GMT
Content-type: text/html
Content-length: 11211
<!-- comments--><HTML><HEAD><title>Fortran Parallel Programming Systems</title></HEAD><BODY bgcolor="#ffffff"><h1>Fortran Parallel Programming Systems</h1><hr><p>The Fortran Parallel Programming Systems, or Fortran Tools, project seeks to make parallel computer systems truly usable for Fortran programmers. In this effort, special emphasis is placed on data-parallel programming and scalable parallelism. <p><h3>On This Page...</h3><p><ul><li> <!WA0><a href = "#OO">Project Overview</a><li> <!WA1><a href = "#LC">Fortran D Language & Compilers</a><ul><li> <!WA2><a href = "#HPF">High Performance Fortran (HPF)</a><li> <!WA3><a href = "#IP">Irregular Problems</a></ul><li> <!WA4><a href = "#DS">The D System</a><li> <!WA5><a href = "#RP">Related Projects</a><li> <!WA6><a href = "#P">People</a></ul><p><h3>Project Home Pages</h3><p> <ul><li> <!WA7><a href = "http://www.mcs.anl.gov/adifor">ADIFOR</a><li> <!WA8><a href = "http://www.crpc.rice.edu/fortran-tools/DSystem/DSystem.html">The D System</a><li> <!WA9><a href = "http://www.crpc.rice.edu/fortran-tools/ParaScope/">ParaScope</a></ul><p><h3>Related Links</h3><p><ul><li> <!WA10><a href = "http://riceinfo.rice.edu/">Rice University home page</a><li> <!WA11><a href = "http://www.cs.rice.edu/CS/compilers/">Rice Compiler Group home page</a><li> <!WA12><a href = "http://www.cs.rice.edu/">Rice Computer Science home page</a><li> <!WA13><a href = "http://www.crpc.rice.edu/CRPC/">Center for Research on Parallel Computation (CRPC) home page</a></ul><p><hr><p><a name = "OO"><h2>Project Overview</h2></a>The Fortran Tools effort seeks to make parallel computer systems usable for Fortran programmers. In this effort, special emphasis is placed on data-parallel programming and scalable parallelism. <p>To achieve this goal, researchers are developing a coordinatedprogramming system that includes compilers and tools for Fortran D, anextended dialect of Fortran that supports machine-independent data-parallelprogramming. The tools support a variety of parallel programmingactivities, including intelligent editing and program transformation,parallel debugging, performance estimation, performance visualization andtuning, and automatic data partitioning. <p>Research efforts also include validation of the compilers and tools onrealistic applications, as well as investigations of new functionality tohandle irregular computations, parallel I/O, and automatic differentiationusing the program analysis infrastructure developed for the project.<p><a name = "LC"><h2>Fortran D Language and Compilers</h2></a><p>Existing languages for parallel programming on scalable parallel systemsare primitive and hard to use. They are primitive in the sense that eachone reflects the architecture of the target machine for which it isintended, making programs written for current parallel systems highlymachine-dependent. As a result, there is no protection of the programminginvestment on parallel machines -- a program written for one target machinemay need to be completely rewritten when the next-generation machine isavailable. This situation is the principal impediment to widespread use ofscalable parallel systems for science and engineering problems. <p>To address this problem, researchers have developed Fortran D, a setof extensions to Fortran 77 and Fortran 90 that permit the programmer tospecify, in a machine-independent way, how to distribute a program'sprincipal data structures among the processors of a parallel system. Inaddition, Fortran D makes programming easier than it is with explicitmessage-passing, because programmers can write codes that use a shared namespace, independent of the target architecture. Programmers find a sharedname space easier to use than a distributed name space because dataplacement and access issues can be ignored. Using sophisticated compilertechniques, these "high-level" programs can be compiled for both SIMD andMIMD parallel architectures.<p>The Fortran D research effort has led to prototype compilers for the IntelParagon and Thinking Machines CM-5 for both Fortran 77D and Fortran 90D.In addition, the Fortran 90D compiler has been ported to a number of othermachines, including the Intel Paragon, nCube/2, and a network ofworkstations. Compilers for other machines, such as the SIMD MasPar MP-2,are under development. The strategy for all these compilers is based upondeep program analysis, aggressive communication optimization, advancedcode-generation techniques and the use of sophisticated computation andcommunication libraries. The effectiveness of these methods is beingevaluated, using a suite of scientific programs developed by affiliatedresearchers at Syracuse University. <p><a name = "HPF"><h3>High Performance Fortran (HPF)</h3></a><p>Fortran D was a major impetus behind the definition of High Performance Fortran (HPF). The High Performance Fortran Forum, which produced the definition of HPF, includes representatives from industry, academia, and government laboratories. TheFortran D compilers produced in part by the Rice Compiler Group are being used as models for several commercial HPF compilers. Thus, the project has established an efficient technology transfer mechanism by which new features in Fortran D,once demonstrated, may be included in a future round of HPF definition. (<!WA14><a href = "http://www.erc.msstate.edu/hpff/home.html">HPF home page</a>)<p><a name = "IP"><h3>Irregular Problems</h3></a><p>The Fortran group also works closely with applications scientists andengineers working on "irregular" scientific problems, such as computationalfluid dynamics, computational chemistry, computational biology, structuralmechanics, and electrical power grid calculations. Key aspects of theresearch associated with irregular scientific problems focuses on thedevelopment of portable runtime support libraries which (1) coordinateinterprocessor data movement, (2) manage the storage of, and access to,copies of off-processor data, (3) support a shared name space, and (4)couple runtime data and workload partitioners to compilers. These runtimesupport libraries are being used to port application codes to a variety ofmultiprocessor architectures and are being incorporated into the Fortran Ddistributed-memory compilers. For a list of technical papers on irregular problems, see the <!WA15><a href = "http://www.crpc.rice.edu/fortran-tools/DSystem/techPapers.html#DCirreg">D System:Technical Papers</a> Web pages.<p><a name = "DS"><h2>The D System</h2></a>In 1992, researchers at Rice University set out under ARPA funding to build a suite of prototype tools that support development of programs in Fortran D, an abstract, machine-independent parallel programming language. The tools emerging from this research are collectively being called the D System. To date, research has focused on four key areas: an intelligent editor for Fortran D that provides feedback on the analysis and parallelization performed by a Fortran D compiler developed at Rice; advances in automatic data distribution research performed at Rice; joint work on performance analysis of Fortran D programs with collaborators at the University of Illinois; and overhaul and integration of a program analysis repository with the interprocedural analysis system to provide a firm basis for the development of efficient whole-program analysis tools. (<!WA16><a href = "http://www.crpc.rice.edu/fortran-tools/DSystem/DSystem.html">D System home page</a>)<p>The D System grew out of a collection of tools called ParaScope, which was initially designed to support development of Fortran programs with explicit parallelism in the form of parallel loops. One difference between ParaScope and the D System is that ParaScope focuses on shared memory machines, whereas the D System is targeted to distributed shared memory (DSM) machines. (<!WA17><a href = "http://www.crpc.rice.edu/fortran-tools/ParaScope/">ParaScope home page</a>)<p><a name = "RP"><h2>Related Projects</h2></a>Members of the Fortran group are involved in several additionalcollaborations that are capitalizing on the available softwareinfrastructure. For instance, researchers at Rice University and ArgonneNational Laboratory are continuing to enhance ADIFOR, an automaticdifferentiation tool for Fortran built upon the ParaScope infrastructure,to support sensitivity analysis of large simulation codes for use inmultidisciplinary design optimization by members of the CRPC ParallelOptimization group. (<!WA18><a href = "http://www.mcs.anl.gov/adifor">ADIFOR home page</a>)<p>The Massively Scalar Compiler Project (MSCP) at Rice is exploiting the interprocedural analysis engine developed for ParaScope and is also investigating interactions between parallelizing transformations and scalar node performance. (<!WA19><a href="http://www.cs.rice.edu/MSCP/MSCP.html">MSCP home page</a>)<p>The Fortran group is also collaborating with the CRPC Parallel Paradigm Integration project to investigate ways of integrating ParaScope and Fortran D-style data decomposition directives into Fortran M, a modular version of Fortran. Syracuse University is coordinating an ARPA activity to set up a Parallel Compiler Runtime Consortium. This involves other sites and aims to design and implementcommon runtime support for parallel Fortran, C++ and Ada for both data andtask parallelism. Finally, Rice University is an active collaborator in a project by the Intel Delta Consortium to develop software support for parallel I/O.The Fortran project researchers will develop and implement extensions toFortran D that support "out-of-core" arrays, which are too large to fitinto the main memory of even a massively parallel computer system. <p><a name = "P"><h2>People</h2></a>Present members of the Fortran Tools effort include <!WA20><a href = "http://www.cs.rice.edu/~adve/">Vikram Adve</a>, <!WA21><a href = "http://www.cs.rice.edu/~carle/">Alan Carle</a>, <!WA22><a href = "http://www.cs.rice.edu/~keith/">Keith Cooper</a>, <!WA23><a href = "http://www.cs.rice.edu/~ken/">Ken Kennedy</a>, <!WA24><a href = "http://www.cs.rice.edu/~chk/">Charles Koelbel</a>, <!WA25><a href = "http://www.cs.rice.edu/~johnmc/">John Mellor-Crummey</a>, <!WA26><a href = "http://www.cs.rice.edu/~linda/">Linda Torczon</a>, and <!WA27><a href = "http://www.rosetta.com/Resume.html">Scott Warren</a>. To "meet" the people behind a specific Fortran Tools project, see the home page for that project.<p><hr><p><center> <!WA28><a href = "http://www.mcs.anl.gov/adifor">ADIFOR</a> - <!WA29><a href = "http://www.crpc.rice.edu/fortran-tools/DSystem/DSystem.html">The D System</a> - <!WA30><a href = "http://www.crpc.rice.edu/fortran-tools/ParaScope/">ParaScope</a> - <!WA31><a href = "http://www.cs.rice.edu/">Rice University Computer Science Department</a> <br><!WA32><a href = "http://www.cs.rice.edu/CS/compilers/">Rice Compiler Group</a> - <!WA33><a href = "http://www.crpc.rice.edu/CRPC/">Center for Research on Parallel Computation (CRPC)</a> - <!WA34><a href = "http://riceinfo.rice.edu/">Rice University</a> </center><p><hr><p>Updated by <!WA35><a href = "http://www.cs.rice.edu/~dcamp/">Debbie Campbell</a> (<!WA36><A HREF="mailto:dcamp@cs.rice.edu">dcamp@cs.rice.edu</A>)<br>http://www.crpc.rice.edu/fortran-tools/fortran-tools.html</BODY></HTML>⌨️ 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 + -