OpenMP News
»SC08 OpenMP “Hands-On” Tutorial Available
Tim Mattson and Larry Meadows, both of Intel, presented a day-long tutorial introducing parallel programming with OpenMP at SC08 last week in Austin, TX.
The slides and class exercises from that tutorial are now available:
- Hands-On Introduction to OpenMP, Mattson and Meadows, from SC08 (Austin) (PDF)
- Code Exercises (zip)
»OpenMP 3.0 Status
»Christian Terboven reports:
SC08 brought us some pretty good news regarding availability of (full) support for OpenMP 3.0:
- Intel 11.0: Linux (x86), Windows (x86) and MacOS (x86)
- Sun Studio Express 11/08: Linux (x86) and Solaris (SPARC + x86)
- PGI 8.0: Linux (x86) and Windows (x86)
- IBM 10.1: Linux (POWER) and AIX (POWER)
GCC 4.4 will have support for OpenMP 3.0 as well, it is currently in regression fixes and docs only mode (see http://gcc.gnu.org/ml/gcc/2008-11/msg00007.html).
»OpenMP at SC08 Austin
Michael Wong of IBM Canada attended the OpenMP “Birds-of-a-Feather” meeting at SC08 in Austin, TX this week, and is blogging about it on his »C/C++ Cafe blog.
»OpenMP 3.0 Specs Summary Card
An 8 page summary card of the OpenMP 3.0 specifications is now available:
»Version 3.0 Summary Card (November, 2008) (PDF)
»New Book (in German)
OpenMP - Eine Einführung in die parallele Programmierung mit C/C++
(OpenMP - An Introduction into Parallel Programming with C/C++ )
von Rainer Lienhart und Simon Hofmann, (Universität Augsburg)
Has been published by »Springer (in German). Here is the description:
Über dieses Lehrbuch
OpenMP ist ein weit verbreiteter de-facto-Standard für High-Level Shared-Memory-Programmierung, der für viele Plattformen, u.a. Linux und Microsoft Windows, zur Verfügung steht. Das Programmiermodell von OpenMP ermöglicht einen einfachen, skalierbaren und flexiblen Ansatz zur Entwicklung paralleler Applikationen unter FORTRAN, C und C++. Seit seiner Einführung 1997 wird Open MP von der Mehrheit der High-performance Compiler- und Hardwarehersteller unterstützt.
In diesem Buch wird Open MP ausführlich vorgestellt und gezeigt, wie eine Implementierung paralleler C/C++ Algorithmen erfolgt.
This book can also be purchased as a PDF download from the Springer website.
»PGI Announces Release 8.0 With Full OpenMP 3.0 Support
Full support for the OpenMP 3.0 parallel programming standard for multi-core x64 CPUs and multi-socket servers, including full support for TASKs in PGF95 and PGCC in release 8.0.
More information is at the »PGI website.
»OpenMP 3.0 Implementation Completed in Sun Studio Express 11.2008
Sun Microsystems has announced the Sun Studio Express November 2008 release with complete implementation of the OpenMP 3.0 specifications.
Some OpenMP 3.0 features were first introduced in the Sun Studio Express July 2008 release. The implementation of OpenMP 3.0 is completed with the »Sun Studio Express November 2008 Release. The new OpenMP 3.0 features in the November release are:
- Pointer type loop control variables in C and C++
- C++ random access iterators
- Fortran allocatable arrays
More information at the »Sun Studio OpenMP Wiki
Sun Studio Express compilers and tools run on Solaris, OpenSolaris, and Linux, on SPARC and x86/x64 platforms, and is a free download from the »SDN Portal
»Japanese Translation Available
Our thanks go to the volunteers from Fujitsu who have completed a translation of the OpenMP 3.0 API Specifications into Japanese, and to Dr. Satoh at the University of Tsukuba and volunteers at NEC who reviewed the final document for accuracy.
»OpenMP Tutorial at Supercomputing 2008
Tutorial S08: A Hands-on Introduction to OpenMP (Full day)
Sunday, November 16, 2008 »SuperComputing ’08, Austin, Texas
Instructors: Larry Meadows, Intel Corporation, Tim Mattson, Intel Corporation
If you are planning on attending the 20th Anniversary of the SC Conference, to be held this year in Austin from November 15 through 21, and if you are interested in gaining a working knowledge of the OpenMP parallel programming model, then please consider registering for this tutorial, to be held on Sunday, November 16. This is a full day hands-on tutorial. We will alternate lectures with hands-on exercises designed to teach the basics and some advanced topics in OpenMP, including the new OpenMP 3.0 features
Participants are strongly encouraged to bring their own laptops and to pre-install a compilation environment that supports OpenMP. A few laptops will be available for those unable to bring their own
We will support laptops running Windows, Linux, or OSX. The main pre-requisite is a conforming OpenMP environment.
The exercises are all written in C or C++, so no Fortran compiler is required
For Linux, we have tested a distribution of Fedora Core 9 on an Intel64 processor (64-bit mode) with gcc 4.3.2
For Windows, we have tested the PGI compiler (7.2-5) and the Intel compiler (10.1) on Windows XP
For OSX, we have tested the PGI compiler (7.2-5) on Mac OS X Leopard (64-bit)
Note on OpenMP 3.0 : OpenMP 3.0 has introduced a new concept called tasks. One solution to one of the exercises uses tasks. The above mentioned versions of the PGI compiler and the Intel compiler do not support tasks, so that particular solution won’t compile. It appears that gcc 4.3.2 does support the syntax for tasks, but it isn’t clear if all the runtime library work is finished.
We suggest the use of the PGI compiler for Windows and OSX because it does not require installation of any additional packages. PGI has agreed to extend their standard 15-day evaluation period until the end of November for tutorial participants.
To take advantage of this offer please send an e-mail request to sales@pgroup.com (preferably after installing the PGI compiler and obtaining your temporary license key) and mention that you are attending the OpenMP tutorial at SC. PGI will send you special extended demo keys and instructions for downloading an advance copy of the PGI 8.0 release compilers including full support for OpenMP 3.0 tasks.
If you choose to use PGI 8.0, we suggest you co-install it with PGI 7.2-5 or another of the pre-tested compilers listed above.
If you have any questions or problems validating an OpenMP environment on your laptop, and you plan to attend the tutorial, please feel free to contact Larry Meadows at lawrence.f.meadows@intel.com
Go to the »SuperComputing ’08 website for more information.
»IBM Releases XL Compilers with OpenMP 3.0 Support
The IBM XL C/C++ V10.1 and IBM XL Fortran V12.1 compilers have been released for both the AIX and Linux platforms. The XL C/C++ V10.1 compilers include full support for OpenMP V3.0. The XL Fortran V12.1 compilers include support for a subset of OpenMP V3.0 features.
For details and trial downloads, please refer to
http://www-01.ibm.com/software/awdtools/fortran/ for XL Fortran and to
http://www-01.ibm.com/software/awdtools/xlcpp/ for XL C/C++.
»OpenMP at SC08
OpenMP will be at Supercomputing ‘08 in Austin, TX, Nov 17-21:
- A Birds of a Feather meeting OpenMP 3.0: Ushering in a New Era of Parallelism on Tuesday, Nov 18, 5.30-7.00 pm
- The OpenMP booth on the exhibit floor, manned by OpenMP experts giving demos and answering questions.
If you’re planning to attend SC’08, stop by and say hello!
»More Experiments from Christian Terboven on Tasking in 3.0
Christian Terboven at the University of Aachen continues his experiments with OpenMP 3.0 tasking:
»More on Tasking in OpenMP 3.0
We already played a lot with Tasking and are looking into how well it can be applied to *real* applications. In the “Introduction to Parallel Programming” lecture and also in the discussions following after some of my recent talks I noticed a strong interest in this new paradigm and I really hope that people will like it!
»Latest Topics in the General Forums
The following topics are currently being discussed over at the »OpenMP General Forum:
- Global scope & threadprivate leads to undefined symbol
- How can I install openMP packages on Ubuntu?
- Any good 3.0 examples?
- Does OpenMP implementation clean up the threads itself?
- Custom reduction operators in OpenMP
- Calculate CPU Utilization
- Only uses 2 processors?
- Totally lost when compiling
- flush an array element
- OpenMP vs. sieve
- Simple Parallel Program Issue
- Mixing OpenMP 2.5 and OpenMP 3
- Reduction for arrays
- A question about OMP3.0 spec
- Streams in C++
- openMP in Leopard
- Nested for loops
- Whats the problem with this code?
- open mp performance degradation
- Rounding off problem(maybe..)
- Please Help. MFC + OpenMP?
- Is OpenMP Event Driven
- OpenMP makes programs the same speed
- Dynamic momory allocation in a parallel region
»OpenMP 3.0 in Sun Studio Express 7.08 Compilers
Sun Studio Express July 2008 released today implements some of the new OpenMP 3.0 features in their C, C++, and Fortran compilers and tools.
Runs on Solaris, OpenSolaris, and Linux, on SPARC and x86/x64 platforms, and it’s a free download from the Sun Studio portal.
Details on what OpenMP 3.0 features are implemented in this “beta” release is on the Sun Studio Wiki.
»IWOMP 2009 Announced
The 5th International Workshop on OpenMP: Evolving OpenMP in an Age of Extreme Parallelism - will take place in Dresden (Germany) from the 3rd June through the 5th June 2009.
The workshop will be hosted by the Center for Information Services & High Performance Computing (ZIH) at the Dresden University of Technology.
The International Workshop on OpenMP is an annual series of international workshops dedicated to the promotion and advancement of all aspects focusing on parallel programming with OpenMP. The workshop serves as a forum to present the latest research ideas and results related to this shared memory programming model.
More information is at the IWOMP website, »http://www.iwomp.org/
The technical program will be announced on the IWOMP website (and here) in March, 2009.
»International Journal of Parallel Programming, OpenMP Edition
»The International Journal of Parallel Programming June 2008 edition is devoted entirely to OpenMP.
• Guest Editors Introduction: Special Issue on OpenMP
Guang R. Gao, Mitsuhisa Sato and Eduard Ayguadé
• Supporting OpenMP on Cell
Kevin O’Brien, Kathryn O’Brien, Zehra Sura, Tong Chen and Tao Zhang
• Performance Evaluation of a Multi-Zone Application in Different OpenMP Approaches
Haoqiang Jin, Barbara Chapman, Lei Huang, Dieter an Mey and Thomas Reichstein
• Nebelung: Execution Environment for Transactional OpenMP
Miloš Milovanović, Roger Ferrer, Vladimir Gajinov, Osman S. Unsal, Adrian Cristal, Eduard Ayguadé and Mateo Valero
• Performance Advantage of Reconfigurable Cache Design on Multicore Processor Systems
Jie Tao, Marcel Kunze, Fabian Nowak, Rainer Buchty and Wolfgang Karl
»Christian’s First Experiments with Tasking in OpenMP 3.0
From Christian Terboven’s blog:
OpenMP 3.0 is out, maybe a bit later than we hoped for, but I think that we got a solid standard document. At IWOMP 2008 a couple of weeks ago, there was an OpenMP tutorial which included a talk by Alex Duran (from UPC in Barcelona, Spain) on what is new in OpenMP 3.0 - which is really worth a look! My talk was on some OpenMP application experiences, including a case study on Windows, and I really think that many of our codes can profit from Tasks. Motivated by Alex’ talk I tried the updated Nanos compiler and prepared a couple of examples for my lectures on Parallel Programming in Maastricht and Aachen. In this post I am walking through the simplest one: Computing the Fibonacci number in parallel.
»New Forum Created
The OpenMP 3.0 API Specifications forum is now open for discussing the specs document itself.
»New Links
New links and information have been added to the OpenMP Compilers and the OpenMP Resources pages.
»Recent Forum Posts
»Tutorial Slides from IWOMP 2008
Here are slides from some of the tutorials presented at IWOMP 2008:
- Important concepts in parallelization (PDF) - Ruud van der Pas, Sun Microsystems, Menlo Park, CA, USA
- OpenMP under the hood (PDF) - Barbara Chapman, University of Houston, Houston, TX, USA
- OpenMP 3.0: What’s new? (PDF) - Alejandro Duran, Barcelona Supercomputing Center, Barcelona, Spain
- OpenMP in the real world (PDF) - Christian Terboven and Dieter an Mey, RWTH Aachen University, Aachen, Germany
- OpenMP and performance (PDF) - Ruud van der Pas, Sun Microsystems, Menlo Park, CA, USA
»OpenMP 3.0 Specs Now Available
The OpenMP Architecture Revew Board has approved the 3.0 specifications.
The 3.0 draft specifications were out for public review from October thru January. The final 3.0 specs is now available. It corrects a number of typos, clarifies some issues raised during the review period, and contains completed appendicies and lists of examples and implementation-defined features.
»OpenMP 3.0 API Specifications (May 2008) (PDF)
»OpenMP at ISC’08 Dresden
Sessions discussing OpenMP at ISC 08 in Dresden in June:
- Harnessing the Potential of Multicore/Manycore Processors
- MPI has been the prevailing programming interface for parallel computing. With the widespread use of multi/many-core computers, OpenMP will gain in importance. Professor Lutz Gross, the Director of High Performance Computing in the Earth Systems Science Computational Center (ESSCC
...
http://www.supercomp.de/isc08/content/e2/e2437/index_eng.html
- Parallel FEM Solvers on Multicore Xeon Clusters: Do We Need OpenMP?
- Prof. Dr. Lutz Gross Director HPC Earth Systems Science Computational Center The University of Queensland, Australia “To write parallel software
...
http://www.supercomp.de/isc08/content/e2/e10/e22/e1671/e2931/index_eng.html
»Recent User Forum Posts
»Recent Posts in the General Topics Forum
- Drop of performance for large arrays/matrices
- parallelized for_each
- OPENMP + Mixing FORTRAN and C code
- traverse stl containers in parallel
- how to use “private” ?
- Messages in OpenMP
- default(private)
- OMP ON NEC SX-8
- omp_get_max_threads in parallel region
- Calling functions in parallel For-Loop
»OpenMP in a New Era of Parallelism
New workshop proceedings from Springer to be published soon:
OpenMP in a New Era of Parallelism
4th International Workshop, IWOMP 2008 West Lafayette, IN, USA, May 12-14, 2008, Proceedings
Series: Lecture Notes in Computer Science , Vol. 5004
Sublibrary: Programming and Software Engineering
Eigenmann, Rudi; Supinski, Bronis R. de (Eds.)
2008, X, 191 p. With online files/update., Softcover
ISBN: 978-3-540-79560-5
This book constitutes the refereed proceedings of the 4th International Workshop on OPenMP, IWOMP 2008, held in West Lafayette, IN, USA, in May 2008.
The 16 revised full papers presented together with 1 keynote talk and 1 invited lecture were carefully reviewed and selected for inlusion in the book. The papers are organized in topical sections on OpenMP overheads and hybrid models; OpenPM for clusters; OpenPM tasking models and extensions, as well as applications, scheduling, and tools.
»S-Lang Interface Package 1.9.4 Supports OpenMP
»SLIRP is a vectorizing code generator aimed primarily at simplifying the process of creating modules for the S-Lang scripting language. It supports making C, C++, and Fortran code callable directly from the S-Lang interpreter and can automatically vectorize functions to take advantage of the abilities of S-Lang. SLIRP can also generate parallelizable wrappers for OpenMP-aware compilers.
»SC07 OpenMP BOF
The SC07 OpenMP BOF was held at the SC07 conference in Reno on November 15, 2007. There were about 70 attendees. Mark Bull of EPCC presented the draft OpenMP 3.0 standard. Marty Itzkowitz presented a modification to the Performance API to reduce overhead for short parallel regions. Both presentations are available on the resources page.
RSS