Today the OpenACC standards group announced growing support for OpenACC-supported development tools, and initial results from programmers who have been using the recently-released OpenACC compilers to accelerate research.
Using PGI’s OpenACC compiler, we ported a computational fluid dynamics (CFD) application benchmark to a general purpose GPU-based system,” reported NASA researchers in an upcoming research paper. “OpenACC is a much easier way to accelerate applications than other programming approaches, and we saw an immediate speed up of the benchmark on multiple tests, up to 10X faster compared with a single CPU core-based system.”
I met with members of the OpenACC standards group last week and it is clear to me that Directives are simply a much more efficient way to port codes to accelerators, especially on massive hybrid systems like the Titan Cray XK7.
As evidenced in this SC12 Show Guide, OpenACC will be all over the SC12 show floor next week. Be sure to check out the OpenACC tutorials and BoFs as well as demos from CAPS, PGI, and Rogue Wave. Read the Full Story.
The 5th PRACE Regular Call for Proposals, which opened on 17 April 2012 and closed on 30 May 2012, received 80 applications for time on one or more of the six Tier-0 PRACE systems offered. Of those applications, a record-breaking 57 withstood the in-depth scrutiny of the peer review process and were awarded the core hours needed to perform their research. The project allocation is as follows:
- Heavy ion phenomenology form lattice simulations by Szabolcs Borsanyi of the Bergische Universität Wuppertal (Germany) received 91.791.360 core hours on Juqueen, Germany.
- SHAKEIT – Physics-based evaluation of seismic shaking in Northern Italy by Andrea Morelli of the Istituto Nazionale di Geofisica e Vulcanologia (Italy) received 53.4 million core hours on SuperMUC, Germany.
- ENSING – Euler and Navier-Stokes SINGularities by Sergio Pirozzoli of the Sapienza, University of Rome (Italy) received 50 million core hours on Fermi, Italy.
- HiResClim: High Resolutions Climate Modeling by Colin Jones of the Swedish Metereological and hydrological Institute (SMHI, Sweden) received 38 million core hours on MareNostrum at BSC, Spain.
- Large scale molecular dynamics simulations of nucleation by Jürg Diemand of the University of Zürich (Switzerland) received 35.3 million core hours on Hermit, Germany.
- Electrophysiology – Atomistic modeling by Mounir Tarek of CNRS – Université de Lorraine (France) received 28 million core hours on Curie, France.
This story appears here as part of a cross-publishing agreement with Scientific Computing World.
A new Technical Report from the OpenMP Consortium details directives used for the execution of loops and regions of code on attached accelerators. Described as “a work in progress,” the report was released in order to get early feedback on the proposed directives.
We aim to provide what the marketplace has been looking for, a standard high-level way of programming accelerators across a broad base of languages and for all forms of accelerator devices”, said Michael Wong, OpenMP CEO.
This Technical Report describes a model for the offloading of code and data onto a target device. Any device may be a target device, including graphics accelerators, attached multiprocessors, co-processors and DSPs. The directives detailed in the Technical Report can be used in Fortran, C, and C++. Read the Full Story.
Also posted in HPC Software
Today Rogue Wave Software announced that their upcoming TotalView debugger release will include expanded support for the newest platforms that are advancing both the HPC and commercial markets. With the release of TotalView 8.11, debugging support will be offered for the IBM Blue Gene/Q platform, NVIDIA CUDA 4.2, and the OpenACC capability of Cray CCE 8.0. TotalView 8.11 will also offer early access support for the Intel Xeon Phi coprocessor.
TotalView is our tool of choice for debugging on the Blue Gene/Q platform, especially following our extremely positive experiences with it on the predecessor JuGene machine, which was the world’s largest Blue Gene/P, having approximately 300,000 cores. We need to support applications that use hybrid MPI and OpenMP, and TotalView gives us the ability to debug our very-scalable and most-complex, distributed, multi-threaded applications,” stated Bernd Mohr, Deputy Department Head of Application Support at Jülich Supercomputing Centre.
Read the Full Story or check out Rogue Wave at SC12 booth #3418.
This week Altair announced that its PBS Pro workload management software now supports Intel Xeon Phi co-processor scheduling and includes a configuration toolkit for simplified administration.
Our customers are among the most demanding high-performance computing consumers on the planet, and they expect their workload manager to support the latest innovations in hardware,” said Bill Nitzberg, chief technology officer for PBS Works at Altair. “With our support for Intel Xeon Phi including the simplified configuration toolkit, we are ensuring our users can run some of the most compute-intensive portions of their work on the most cutting-edge, high-performance architecture available.”
Read the Full Story.
Intel’s James Reinders writes that OpenMP 4.0 may offer important solutions for targeting and vectorization.
The “targeting directives” intended for OpenMP 4.0, have the challenge of spanning the likes of NVidia GPUs (SIMT oriented), and Intel Xeon Phi coprocessors (SMP-on-a-chip), and Intel HD Graphics (vector oriented GPUs), other GPUs (like AMD), and other potential attached processing devices and future ones as well. It is not easy to bring multiple companies together to find common ground.
Read the Full Story
In related news, there will be an OpenMP BoF at SC12 as well as a celebration of OpenMP’s 15th birthday at SC12 Booth #2237 next week.
Scaling applications at scale takes planning, work, and the right tools. At iVEC and CSIRO in Australia, users are busy preparing for their next-generation Cray Cascade supercomputer by deploying Allinea DDT in their existing environment.
The purchase of this advanced supercomputer and suite of development tools represents a great investment in the future of Australian research,” said Applications Specialist Rebecca Hartman-Baker. “Having access to such power will allow scientists to explore areas of research that were previously unavailable.”
Read the Full Story.
Also posted in HPC Software, Tools
Today Samplify announced the availability of its APAX Software Development Kit (SDK) and Web-based Profiler analysis tool for high-performance computing (HPC), Big Science, Big Data, and cloud computing applications. According to the company, HPC users and independent software developers can accelerate their applications, 3-8X, by linking the APAX SDK into their applications.
Our APAX numerical encoding technology reduces the cost of every aspect of cloud computing by reducing data upload, storage and interprocessor communication bottlenecks by 75 percent,” said Allan Evans, Samplify CEO. “By making APAX available as an SDK, we now make cloud computing resources more accessible to companies, scientists, and academic researchers, and expand the capability of existing data centers.”
The new APAX Profiler analyzes the inherent accuracy in the user’s data set and makes a recommendation of encoding rates to maximize the acceleration of their algorithm with no effect on results. Read the Full Story or check out this RichReport podcast on APAX.
Also posted in HPC Software
A new International Consortium on sustained data management is a planning a launch event at SC12. Called E-iRODS, the consortium will be formed by the University of North Carolina at Chapel Hill and members of the Max Planck Society with a mission to develop Rule-Oriented Data System (iRODS) into a sustained, production-quality technology for data management, sharing, and integration.
SC12 will be our first chance to talk to research communities about their data challenges and how they could benefit from being involved in the E-iRODS Consortium,” said Charles Schmitt, director of data initiatives at RENCI. “We invite any group looking for data solutions—whether they are current iRODS users or not—to attend this reception.”
Participants in SC12 interested in learning more about the E-iRODS Consortium are encouraged to attend the presentation “Enterprise iRODS and the E-iRODS Consortium,” in the RENCI/North Carolina booth (3640) at 1:30 p.m. Tuesday, Nov. 13, or 10:30 a.m. Wednesday, Nov. 14. Additionally, an informational reception about iRODS and E-iRODS will be held from 5 p.m. – 7 p.m. Wednesday, Nov. 14, in the Solitude Room at the Salt Lake Marriott Downtown, across the street from the Salt Palace Convention Center. Read the Full Story.
Over at The Register, Dan Olds writes that winning the SC12 Student Cluster Competition is all about how you approach the applications.
The real challenge for students in this competition is how to best make use of the hardware they have. On Monday evening, the teams will receive the data sets for each task. This is a critical time in the competition. Some teams just jump in and start running code as quickly as they can. Others will take some time, get a feel for the size and complexity of the various tasks, and then plan out how they’re going to attack each application.
Olds goes on to suggest that next year, the competition be expanded to include some business applications centered around Big Data. Read the Full Story.
Also posted in Events, SC12
Today Penguin Computing announced that Altus rackmount systems are now shipping with AMD Opteron 6300 Series processors. Known as “Piledriver,” the new AMD chips deliver improved performance and power efficiency.
The new line of AMD Opteron 6300 Series processors offers superior performance and scalability,” said John Williams, vice president, Server Marketing and Business Development, AMD. “With a high core density and balanced processor architecture, the AMD Opteron 6300 Series processors offer a compelling valuable proposition for our joint customers.”
Penguin Computing will showcase their latest HPC technologies at SC12 booth #1217 next week in Salt Lake City. Read the Full Story.
The Arctic Region Supercomputing Center (ARSC) is seeking an HPC User Services Support Engineer in our Job of the Week.
This position primarily provides user support for ARSC’s high performance computation, storage, and networks. It includes all aspects of research and supercomputing user support, from help desk coverage to application performance tuning, account management to system testing, and developing the tools or procedures to carry out those tasks. User support staff provide training and support to users by phone, email and in person, as needed. They represent users in ARSC planning and development activities. Also, they provide information about ARSC activities to current and potential users.
Are you paying too much for your job ads? Not only do we offer ads for a fraction of what the other guys charge, our insideHPC Job Board is powered by SimplyHIred, the world’s largest job search engine.
As a reminder, we are offering FREE job listings for .EDU and .GOV domains, so email us at info @ insideHPC.com for a special discount code.
James Hamilton from Amazon Web Services writes that AMD’s plans to build a 64 bit, 4-core, 2Ghz ARM processor points to a recurring trend in microservers.
Invariably what we see happening about once a decade is a high-volume, lower-priced technology takes over the low end of the market. When this happens many engineers correctly point out that these systems can’t hold a candle to the previous generation server technology and then incorrectly believe they won’t get replaced. The new generation is almost never better in absolute terms but they are better price/performers so they first are adopted for the less performance critical applications. Once this happens, the die is cast and the outcome is just about assured. The high-volume parts move up market and eventually take over even the most performance critical workloads of the previous generation. We see this same scenario play out roughly once a decade.
Read the Full Story.
In this podcast, Steve Henn from NPR talks to Buddy Bland from ORNL and others about the new Titan supercomputer and how its powered by the same GPU technology that drives the video gaming market. Read the Full Story. Download the MP3.
Energy efficiency will again be a hot topic at SC12, with at least 38 Technical Program sessions focused on energy efficiency. Natalie Bates from the Energy Efficient HPC Working Group (EE HPC WG) has written up a terrific Guide to Energy Efficiency Topics at SC12.
Demand for HPC is growing in both the public and private sectors. It is also highly energy-intensive. The Federal government is required by the Energy Independence and Security Act of 2007 (EISA) to reduce energy intensity in all facilities, including laboratories and industrial buildings, by 30% by 2015. The increasing need for HPC and the attendant energy intensity threatens to derail the progress toward this and other goals. Through meeting mandated energy reductions, the Federal government is poised to lead by example in energy efficiency.
Read the Full Story.
Also posted in Events, Green HPC, SC12