127x Filetype PDF File size 0.08 MB Source: www.hpc2n.umu.se
Detailed description ciency and techniques to boost the performance will be pre- Goals sented throughout the short course. This tutorial will give an example-oriented introduction TheteachersatPARA’06willbeProfessorsXingCai,Hans Python in High-Performance to using Python as a programming platform for high- Petter Langtangen, and Kent-Andre Mardal from Simula Computing performance computing. The participants will learn how Research Laboratory (Oslo, Norway) and the University of their daily work can benefit from the flexibility of the Oslo. The course material will be made available as hand- Python language, and from several powerful add-on tools. outs. Parts of the material are also covered by the book Are M. Bruaset, Xing Cai, Also those already familiar with the Python programming H.P.Langtangen: Python Scripting for Computational Sci- HansP.Langtangen, Kent-Andre Mardal, environment can expect to improve their skills and pick up ence, Springer, 2nd edition, 2006. Halvard Moe, Ola Skavhaug, Åsmund Ødegård useful techniques and procedures. After the lectures, participants may try out some exercises on their own laptop with guidance from the course instruc- Targeted audience tors. Necessary software will be provided, both at ❤tt♣✿ ✴✴✇✇✇✳s✐♠✉❧❛✳♥♦✴⑦❤♣❧✴P②❚✉t♦r✐❛❧P❆❘❆✵✻✴andona Abstract Weaddress both researchers doing a lot of “Matlab-style” CD. The virtue of Python is flexibility, making it an ideal numerical computing and plotting, and researchers doing tool for the scientist who wants to experiment with differ- scientific application development or adaption in low-level Whyisthistopicrelevant to PARA’06 attendees? ent models, do rapid prototyping, and even do large scale languages like Fortran, C, or C++. Interactive computing Python has become a popular language over recent years, simulations. The syntax and functionality seem much like and plotting can easily make use of applications and li- and is already in widespread use within the scientific com- Matlab, but Python is a full-fledged advanced program- braries written in compiled languages, while program de- munity. Still, we think that the capabilities of Python for ming language with excellent support for user-defined velopment in Fortran, C, or C++ can benefit from a mixed high-performance computing are under-utilized, especially data types, operator overloading, object orientation, and language approach where Python is used for data manage- as a tool for driving numerical simulations. The language generic (“template”) programming. Moreover, it is con- ment tasks while the most time-critical number crunching is ideal for gluing together existing codes and programs in venient to migrate Python code to Fortran, C, or C++, or takes place in compiled code. various ways. This topic is relevant to PARA’06 attendees equiplegacylibrariesinthoselanguageswithmodernand No previous knowledge about Python is required, but in user-friendly interfaces. Python also has cross-platform such a short course it will be an advantage to have some fa- whoare running numerical simulations as part of their sci- support for most operating system and file management entific work. tasks. This tutorial presents concepts and software that miliarity with basic Python syntax. Newcomers to Python will give the scientist adaptable, powerful, and efficient should, however, be able to get an overview of what the tools. language can be used for and how the technical solutions to various problems look like. Coursecontents Thetutorial starts with a crash course on basic Python syn- tax and programming. Then we cover simple (“Matlab- style”) numerical computing with matrices and vectors us- ing the Numerical Python package. Combination of Python and Fortran, C, or C++ for increased performance is also discussed. In particular we show how conveniently Python integrates with Fortran using the F2PY tool. An overview of plotting and visualization packages callable from Python is also given. Furthermore, we discuss high-level parallel computing using the PyPar module. Computational effi- High erfP ormance Computing Pythonin
no reviews yet
Please Login to review.