Introduction to Computational Resources at NCSA
This short talk will give an overview of NCSA's current and future computational resources with the emphasis on industrial projects and applications.
The Blue Waters System
The Blue Waters sustained petascale system will be described, including details on the POWER7 processor, the interconnection network, the file system, the external network, and the software stack.
XL Compilers and Optimizations
This talk will describe how to use the IBM XL Fortran and C compilers to optimize and tune the generated code. The available optimization levels and the optimizations performed will be described, together with the additional options available for tuning the optimization process to the characteristics of the application and of the target system. The effectof tuning on performance will be illustrated with examples. Profile directed feedback and simultaneous multithreading will also be discussed.
Performance Tools for the IBM POWER7
Galen Arnold, Greg Bauer
This presentation will provide an overview of application performance and current performance tools available to users on Linux on POWER7. As HPC technologies evolve, the need for tools to characterize and improve performance remains constant. These tools provide information on application performance from the single core up to the 10,000s of tightly coupled tasks. After a survey of the familiar and not so familar, IBM POWER-specific tools will be highlighted.
Using Computational Libraries to Accelerate Applications on Blue Waters
Using Computational Libraries to Accelerate Applications on Blue Waters. A description of the computational and I/O libraries that will be available on the Blue Waters petascale system.
Enablers: Manufacturing and High Performance Computing
Manufacturing has the possibility of using high-performance computing to great effect. While relative HPCfast, big computers relative to their timeshave been a part of manufacturing for at least three decades, amazing new levels of capability are becoming excitingly visible. That new capability, though, is through the agency of massively parallel machines. While we have leadership-class computers at the cutting-edge of performance, and we have important and complex application needs, we are missing a key enabler: advanced software expressing the application solutions. How do we easily create leadership software that properly taxes the capability of the machine, producing verifiably accurate solutions? What is needed to close the software gap? Some thoughts to this challenge are the focus of this talk.
Caterpillar Virtual Product Development & HPC
The use of high-performance computing for modeling and simulation will continue to grow as a key driver in product development. The motivation for HPC modeling and simulation will be presented followed by real examples of applications and a potential vision for the future.
State of the Union: Modeling and Simulation
Structural barriers and market failures have limited the rate of adoption and further use of high-performance computing by the U. S. manufacturing sector. In particular, penetration of HPC modeling and simulation into small and medium size U. S. manufacturers is hindered by the investment of human capital and finance required to create greater productivity and sustainability through this methodology. A national strategy and implementation must be created to support manufacturing innovation through modeling and simulation for global competitiveness. This imperative can only be achieved through federal investment and national coordinated effort. The Council on Competitiveness over the past 16 months, through the HPC & U. S. Manufacturing Working Group, advanced this strategy to the Obama administration. What is the State of the Union?
HPC Technology Leadership: An ISV Perspective
This presentation will highlight the strategic focus on HPC at ANSYS, with special emphasis on the potential to leverage connections between government, academic, and commercial stakeholders.
Data Integration Challenges and Solutions
Emerging sensing and information technology are rapidly creating a new paradigm for fusing data from multiple sensors and information sources to adaptively guide real-time observation and decision making. This talk will provide an overview of emerging technologies and case studies in the sustainability arena. An advanced information system for real-time decision making will be presented as an example of cyberinfrastructure that enables easier implementation of numerous real-time applications. Recent research findings on the use of machine learning for automating anomaly detection in real-time sensor data and fusing multiple data sources will also be highlighted.
Building a Successful Partner Ecosystem for HPC
Microsoft is building an HPC partner ecosystem that includes ISVs, open-source software, systems integrators and OEMs. This talk will focus on their work and strategy regarding commercial and open source applications, codes, libraries and tools for HPC.
GPUs in High-Performance Computing: Closing the Software Gap
Modern GPUs, such as the NVIDIA GTX285 and ATI Radeon HD 5870, are massively parallel, many-core compute engines. According to the semiconductor industry scaling roadmap, these compute engines could scale up to 10,000x the peak performance of a circa-2007 microprocessor core by the end of 2016. Such a dramatic increase in computation power will likely trigger multiple major science discoveries as well as revolutions in consumer applications. We are experiencing a once-in-a-lifetime opportunity in our profession. However, the programming and code optimization models of GPU computing are quite different from those of parallel systems based on traditional CPUs. Hwu will describe the recent progress by the HPC community in building GPU-based clusters, enhanced software stacks, educational materials, and porting applications to close the software gap for current and future GPU computing platforms.
Petascale Software Issues
Making effective use of parallel systems requires careful attention to the many aspects of application design. This talk will review the software issues in building petascale applications and will discuss which are common to smaller-scale systems and which are (nearly) unique to petascale systems such as Blue Waters. Approaches to address these issues include new programming models such as UPC and the CoArray features in Fortran 2008, as well as more established techniques such as the use of tuned computational science libraries. The use of performance modeling to guide software tuning will also be covered. Approaches to parallel I/O, often left to the end (and hence the weak link in parallel applications) will also be discussed.
IBM HPC for a Smarter Planet
The recently announced IBM POWER7 server portfolio will be reviewed at a high level, with an emphasis on the High Performance POWER7 system that is targeted for use in the NCSA Blue Waters project. The highly integrated design of this product will be discussed, along with the close integration of this product into the National Petascale Computing Facility.