GPU Technology Conference 2009 - Recorded Sessions
Please note some materials may not be available.
GTC 2010 Recordings Now Available
Session ID Title Abstract Speaker Names Key Topics Streamed Recording Download FLV or MP4 MP4 (Medium) MP4 (Large) PDF
1001 From Brook to CUDA While computing with the GPU may seem like a new technology, the foundations upon what is available today have a legacy which date back to the original parallel supercomputers.  The research community first identified the GPU as a computing platform and shown the promise to solve many of the world’s compute intensive problems, many orders of magnitude faster the conventional CPUs.  Once the opportunity became obvious, the challenge was how to best evolve a general purpose programming model to preserve the GPU’s architectural advantage.  In this talk, I will provide both a perspective on how GPU computing evolved from early published results of using GPU as a computing platform, to the early programming models and tools,  what the state of the art is today, and extrapolate where GPU computing is going.  Ian Buck General Interest View Streamed Recording  Download Video Download Video    Download PDF
1002 Accelerating Energy Exploration and Production with Seismic Processing The energy industry has always been at the forefront of technology and high-performance computing. Today, GPU Computing offers another step-change in the preceding benefits because high-capacity, high-throughput and massively parallel GPUs are well-applied to the compute intensive problems currently solved or proposed by the energy exploration and production industry.   This talk will show how using GPUs for problems like seismic migration-particularly Kirchhoff time migration ("KTM" or "PSTM") and reverse time migration ("RTM")-and reservoir simulation can reduce the runtime of today"s important problems by an order of magnitude (10X) or more.   In addition to the performance that the presenter has seen, in their customers' experience, it will also be possible to discuss how other organizations in the energy industry are harnessing GPUs. Ryan Schneider Energy Exploration View Streamed Recording  Download Video       
1003 Overview of Computational Finance on the GPU Financial institutions make extensive use of computers in solving models of the instruments they trade. In this talk we address the questions of how these instruments are modeled and what the results mean to the recipients. We explore applications of computational finance with a focus on the problems that are faced with existing systems today, why speed and accuracy are important, and how GPU technology can help. Thomas Bradley Computational Finance, High Performance Computing View Streamed Recording Download Video Download Video    Download PDF
1004 Derivative Price Modeling on the GPU To reproduce the implicit dynamics of stocks, increasingly complex models are used for pricing and hedging. Because of this complexity, the numerical methods mainly rely on Monte Carlo (MC) simulation.  Because these methods are easy to parallelize, GPU based hardware is an excellent candidate for accelerating financial algorithms. This talk will summarize the overall performance that one expects from an effective utilization of GPUs using the CUDA or CG Toolkit on financial problems. This will include pricing American and Exotic European options and some Exotic Fixed-Income options studied with CALYON Bank. Also some additional computational challenges which GPUs could help solve will be discussed. Lokman A. Abbas-Turki Computational Finance View Streamed Recording Download Video  Download Video     
1005 OPLib: A GPL Library of Elementary Pricing Functions in CUDA/OpenCL and OpenMP The talk includes a crash course in Derivative Pricing Theory to explain how to reduce nearly all pricing tasks to this handful of routines, spanning the full spectrum of asset classes and model types. High performance CUDA implementations for the key routines are given. As an application, we go through the code of a full-fledged scenario generation application for financial derivatives, also distributed as a C# project under GPL. Claudio Albanese Computational Finance, Tools View Streamed Recording Download Video  Download Video     
1006 Implementing 3D Finite Difference Codes on the GPU This presentation reviews GPU parallelization of 3D finite difference computation over regular grids.  3DFD is a fundamental computation in many applications, including Reverse Time Migration in seismic computing.  A single-GPU implementation is described first, followed up by a scalability study on a cluster of up to 8 GPUs.  Performance results are compared to the theoretical limits of the hardware. Paulius Micikevicius 3D, Tools, Physics Simulation, Energy Exploration, Algorithms & Numerical Techniques View Streamed Recording Download Video  Download Video    Download PDF
1007 Digital 3D Entertainment Digital 3D entertainment is suddenly catching fire. It is hard to find two letters that are changing the thinking and technology in so many industries in the digital entertainment value chain. This talk will cover digital 3D entertainment from the theater to the home. Along the way the audience will hear about some of the graphics intense challenges that people are facing. Gary Sasaki 3D View Streamed Recording Download Video Download Video    Download PDF
1008 Memory Saving Discrete Fourier Transformation on CUDA The Fourier transformation of large data sets is a computation intensive and memory consuming operation.
It can be speeded up by transferring the computation to a graphics card where many operations can be done in parallel. As today even the low class consumer products are CUDA enabled, the transformation can be computed on the graphics device and offload the main processor. Our approach (called FlexFT) intends to make the benefit of parallel computing available on low end hardware running demanding image processing applications.
In the talk we will show our algorithm and present the evaluated results in comparison to existing approaches.  We will show, that our algorithm performs equal to CUFFT but does not suffer from the memory limitation at all.
Steffen Frey, Harald Sanftmann Computational Imaging View Streamed Recording Download Video Download Video     
1009 Real Time Multi-Channel, Multi-Camera Image Processing using CUDA At The Johns Hopkins University Applied Physics Laboratory we are developing and deploying advanced image processing techniques implemented in CUDA and running on off-the-shelf NVIDIA GPUs. Our experience and confidence with CUDA has allowed us to migrate existing image processing techniques from CPU to GPU implementations. The GPU implementations execute with such speed that we are now able to deploy custom-built CUDA implementations that take full advantage of the speed and robustness of the latest CUDA releases. GPUs and the CUDA architecture have allowed us to deploy more sophisticated processing techniques and run multiple techniques simultaneously on additional sensors, multiplying by many times our capability. We will discuss in presentation format the evolution of our product from CPU to GPU and beyond. Sean Happel, Todd Neighoff Computational Imaging View Streamed Recording Download Video Download Video     
1010 Volunteer Computing for GPUs: Petaflops for Free Volunteer computing is a paradigm for high-throughput computing in which computer  owners donate computing resources to science projects. About 500,000 people currently  participate, donating the use of about 800,000 computers and video game consoles. Together  these devices provide over 6 PetaFLOPS of computing power to over 40 science projects, whose  research areas include climate research, astronomy, high-energy physics, epidemiology,  molecular biology, biomedicine, quantum computing, nanotechnology, and environmental  studies.  This session will cover how scientists with applications for NVIDIA GPUs can use BOINC  to quickly and economically gain access to large numbers of GPUs. David Anderson Algorithms & Numerical Techniques, Astro GPU, General Interest View Streamed Recording Download Video  Download Video     
1011 Have You Heard What a GPU Can Do? - A Revolution in Audio  This talk is divided into two parts and details how GPUs are now being used as well as delivering significant benefits for audio processing and editing. The first section explains what features of GPUs and the CUDA language provide key benefits for audio processing. The second part will be given by Rudy Sarzo, a world renowned recording and performing professional rock bassist, and focuses on how the GPU can be integrated into a production audio setup as well as the benefits in doing so. Rudy Sarzo, Ian Williams General Interest View Streamed Recording Download Video Download Video    Download PDF
1012 MetaSL and the Evolution of Shading Languages The performance of the GPU as a 3D geometry engine has revolutionized our  expectations of rendered image complexity.  However, attempts to tap the power of new  hardware in 3D rendering have encouraged the design of a variety of complex and incompatible  programming languages.  MetaSL is a "meta shading language" that defines an abstraction layer  above the specifics of any given hardware implementation.  Rendering plugins (called "shaders)  written in MetaSL are translated to the target shading language of a renderer through "back- end" compilers.  The development of these compilers is supported by a software library and SDK  that provide the intermediate data structures typical of software compilers, including the  construction of an abstract syntax tree.   This talk will provide an overview of MetaSL: its design  philosophy, practical examples of shader implementation, and the custom development of new  back-end compilers for current and future rendering pipelines. Andy Kopra Film, Visualization View Streamed Recording Download Video  Download Video     
1013 Large Scale Visualization Current GPUs can create incredibly realistic images for digital prototyping, however, sometimes realism alone is not sufficient to gain sufficient understanding of a proposed solution or design alternative. In many industries it is very important to further enhance realism in the image with large scale display technologies and immersive environments. This talk discusses many of the technologies and factors associated with large scale visualization, including synchronized multi-channel graphics, 3D stereo configurations, high bit-depth displays, as well as highlighting considerations when many such technologies are combined together.  Ian Williams Visualization View Streamed Recording Download Video Download Video    Download PDF
1015 DirectCompute  This talk will describe the features and advantages of DirectCompute, give performance advice for optimizing DirectCompute Shaders on NVIDIA hardware, and discuss the various applications of DirectCompute. Tianyun Ni Tools View Streamed Recording Download Video Download Video    Download PDF
1016 MATLAB GPU Computing Essentials Tutorial In this tutorial, we will discuss AccelerEyes’ Jacket software which connects MATLAB to the graphics processing unit (GPU).  With the GPU as a backend computation engine, Jacket brings together the best of three important computational worlds:  computational speed, visualization, and the user-friendliness of MATLAB programming.
Jacket enables developers to write and run code on the GPU in the native M-Language used in MATLAB.  Jacket accomplishes this by automatically wrapping the M-Language into a GPU compatible form.  By simply casting input data to Jacket’s GPU data structure, MATLAB functions are transformed into GPU functions.  Jacket also preserves the interpretive nature of the M-Language by providing real-time, transparent access to the GPU compiler.

The tutorial will provide examples of running MATLAB code on the GPU for image and signal processing, life science, finance, and other applications.  Also, a Q/A session will enable audience members to ask specific questions about the Jacket project and MATLAB GPU computing.
John Melonakos Tools View Streamed Recording  Download Video  Download Video     
1017 The Art of Debugging for the CUDA Architecture This session is an open discussion on CUDA debugging tips and discussion of existing debugger features in the CUDA architecture. Members of this panel have several years of experience developing various debuggers and will share their experience building the debugger for the CUDA architecture. This is excellent opportunity for existing and future customers and developers to come together to share their questions and comments on how to make CUDA debugger better and how to take CUDA debugging into every developer tool. Ed Hinkel, Dave LeComber, Satish Salian General Interest, Tools View Streamed Recording Download Video Download Video     
1018 Sparse Linear Algebra - Iterative Solvers and Preconditioners on GPUs for Engineering Simulations Sparse linear algebra solvers are used in many areas of scientific computing and are a key target for GPU acceleration.  In this session, Acceleware will present an overview of its linear algebra libraries including support for direct and sparse iterative solvers.  We will discuss the supported algorithms, performance results, and their application to engineering simulations.  We will also look at some of the more commonly used preconditioners and some of the challenges faced when accelerating these algorithms on the GPU. Michal Okoniewski Algorithms & Numerical Techniques, Tools View Streamed Recording Download Video Download Video     
1019 CULA: Robust GPU Accelerated Linear Algebra Libraries EM Photonics has been working in the field of accelerated linear algebra functions for several years. In a recent project funded by NASA, we have developed GPU versions of the most widely used LAPACK routines in partnership with NVIDIA. Available to the public in the summer of 2009, CULA will be a valuable asset to an extremely broad audience as it is easy to use and provides speedups to many applications.   Our session at the GPU Developer Summit will show the broad range of functionality present in CULA We will demonstrate the ease of using CULA and then present some technical details and "lessons learned" while implementing our routines.  CULA will also be on display at our booth, featuring live-running demos of some CULA-based applications. John Humphrey, Eric Kelmelis, Daniel Price Algorithms & Numerical Techniques, Tools Sorry - this session is not available.        
1020 Advances in GPU-based Image Processing and Computer Vision As a massively parallel processor, the GPU is well-suited for performing 'per-pixel' operations in image processing and computer vision. However, new changes in hardware, software, and algorithm mappings allow entire vision algorithms to be performed solely on GPU. In this session, we'll discuss how GPU programming now goes beyond per-pixel mappings, and is providing speedups in image feature processing and handling, frequency domain processing, graph cuts segmentation, and more. James Fung Computational Imaging, Computer Vision View Streamed Recording        
1021 CUDA Research Roundtable: Mixed Precision GPU Computing Many algorithms used in computational physics can be greatly accelerated by the use of GPUs.  However, the full double-precision floating point operations to which scientists are accustomed can prove costly, especially in compute-intensive applications where floating-point computations rather than memory bandwidth limit performance.  In fact, many scientific problems actually require double precision in only a small subset of the code. In these cases, the development of mixed-precision algorithms can bring substantial improvements without sacrificing overall accuracy. Double precision is used only where it is required; the remaining calculations are carried out in single precision.  Although the task of identifying "necessary"
double-precision code may be non-trivial, the performance payoff can be considerable.  This approach also facilitates software emulation of double precision in key portions of the code, which can be effective in accelerating GPU double-precision performance.  In practice, the emulation is neither complete nor IEEE-compliant, and is cumbersome to code without support at the compiler or processor levels, but impressive improvements in speed have been obtained.  This roundtable will discuss progress made so far in mixed-precision calculations and emulation techniques for GPUs, and will consider the prospects for future development of these approaches.
Steve McMillan Algorithms & Numerical Techniques, General Interest View Streamed Recording  Download Video       
1022 High Performance Remote Graphics GPU computing is rapidly becoming a key technology for many industries. When associated to the latest virtualization technology, graphics intensive applications (CAD, DCC) can be centralized and published to offer users a solution that can be rapidly deployed and easily maintained. Also, multi-GPU architectures can be leveraged for greater scalability and cost effective solutions. This talk will provide an overview of the hardware and software stack that will allow near desktop experience of a remote enterprise class graphics application. Phillippe Rollin Visualization View Streamed Recording  Download Video Download Video    
1023 NEXUS: A Powerful IDE for GPU Computing on Windows Unleash your productivity with NEXUS, NVIDIA's new development environment for GPU Computing and graphics applications that use CUDA C, OpenCL, DirectCompute, Direct3D, or OpenGL. NEXUS introduces native GPU debugging and platform-wide performance analysis tools for both computing and graphics developers, fully integrated into Visual Studio 2008. 
In this session, you will learn to use the powerful NEXUS debugger to set breakpoints and step through your CUDA C kernels or HLSL shader source code. Easily catch hard-to-find bugs using the NEXUS memory inspection tools and set data breakpoints on your running application. Visualize your application's workloads and performance characteristics across the CPU, GPU, and operating system, and then dig deeper using the NEXUS profilers. NEXUS also includes state of the art graphics debugging features and convenient API state inspection.

This session will include information on how to register for the NEXUS Beta Program.
Sebastien Domine Tools View Streamed Recording Download Video Download Video   Download PDF 
1024 Par4All: Auto-Parallelizing C and Fortran for the CUDA Architecture Par4All in CUDA is a new tool to translate C and Fortran code to CUDA to help programmers accelerate their codes using GPU computing. It is a tool based on the PIPS source-to-source framework that is developed by Mines ParisTech and others for 20 years and is one component of the new Par4All open-source initiative from HPC Project to promote parallelism.   Par4All in CUDA uses an abstract interpretation framework based on linear algebra to compute many interesting properties in the program being analyzed. Region analysis is able to compute the array regions used and produced by any program statement. This information is used to parallelize nested loops, allocate the needed data on the GPU and generate communication primitives between the GPU and the host.  Christopher Carothers, Beatric Creusillet, Serge Guelton, Ronan Keryell, Francois Irigoin High Performance Computing, Tools View Streamed Recording Download Video Download Video   Download PDF 
1025 Large-Scale Text Mining on the GPU We are quickly reaching an age in which a capability is needed for knowledge discovery from terabyte-scale unstructured text collection for prompt decision-making. Text mining (TM) is an important field for knowledge discovery in large scale text collection. In terms of speedup, utilizing the GPU for TM processing offers a much more cost-effective choice in comparison with using a high-performance supercomputer. Our research mainly focused on exploring GPU-enhanced solutions of three TM functions, text document encoding, text vector dimension reduction, and document clustering functions. In our experiments, the average speedup ranging from six to eight times for all three TM functions. Our successful results proofed the GPU-enhanced computer’s ability in text mining the large-scale text dataset.  Xiaohui Cui Databases & Data Mining View Streamed Recording Download Video Download Video   Download PDF 
1026 Particle Filters and Sequential Monte Carlo Techniques Using GPUs Sequential Monte Carlo (SMC) techniques, which include "particle filters", have become important tools in a variety of signal processing and estimation applications over the past decade. The basic approach involves using a collection of point masses (particles) to create a non-parametric representation of a probability density of interest. SMC allows non-linear state dynamics and arbitrary (including multi-modal) densities, thus providing an important generalisation of previous approaches such as Kalman filters.

Implementation of SMC algorithms on GPUs is attractive due to the fact that particle computations are highly data-parallel. Some operations however, such as weight re-normalisation and some resampling algorithms, are more difficult to parallelise. The proposed session will discuss implementation issues surrounding SMC algorithms, and facilitate sharing and evaluating implementation ideas.
W. James MacLean Algorithms & Numerical Techniques, Computational Finance View Streamed Recording Download Video  Download Video    
1027 Banking on Monte Carlo Monte Carlo simulation is a common technique used to find solutions to a wide range of problems where it is impossible or impractical to compute a result with a deterministic method. Monte Carlo relies on repeated computation of random numbers and this can be accomplished extremely efficiently on GPUs. In this presentation we give an example from the Finance community of extreme performance gains using the random number generators (RNGs) in the NAG GPU Library. We go on to discuss why techniques such as Monte Carlo might play an increasing role in advancing science during the current step change in hardware technology. Ian Reid Algorithms & Numerical Techniques, Computational Finance View Streamed Recording Download Video  Download Video    
1028 Performance Primitives for Video Codec and Image Processing  The NVIDIA Performance Primitives library (NPP) is a library of highly optimized data-processing functions that execute on the CUDA architecture. It currently comprises 75 functions for image processing. The presentation will give an overview of the problem domains covered by those functions, what image types they support, and how they can be used to implement CUDA based image-processing solutions. The talk also gives data on how NPP's performance compares to CPU based implementations as well as how to design for maximum performance using NPP.  Video encoding and decoding tasks are among the most computationally demanding features of consumer and professional applications today.  NVIDIA's CUDA architecture for GPU Computing was designed to handle the large computational complexity of these tasks with relative ease. In addition to a hardware video decoder, we will present several software codec libraries written in CUDA and\or OpenCL. Frank Jargstorff, Anton Obukhov Computational Imaging, Tools View Streamed Recording Download Video Download Video   Download PDF 
1029 C for CUDA - Advanced This presentation covers the major CUDA optimizations.  Topics will include: maximizing memory throughput, kernel launch configuration, using shared memory, and improving GPU/CPU interaction.  We will also describe how the CUDA Visual Profiler and code instrumentation can be used to assess performance.  While C for CUDA is used for illustration, the concepts covered will apply equally to programs written with OpenCL and DirectCompute APIs.

Thomas Bradley, Paulius Micikevicius Tools View Streamed Recording Download Video Download Video    
1030 GPU-based Libraries:  Accelerating the Next Generation of Applications Through continued technological advances, GPUs have become massively-parallel vector processors streamlined for high-performance computations. The increasing flexibility of the GPU has allowed it to enter new markets, and the vendors behind this hardware have increased GPU usability and opened up the underlying hardware constructs to general computing uses. GPUs are now allowing software normally targeted at high-performance computing platforms to be run on desktops and in GPU-enhanced clusters. This creates an opportunity for GPU-accelerated libraries that can be used to increase the performance of numerous applications. The panel will discuss some of the libraries that are currently being developed and answer questions concerning their future and integration into other applications.  Eric Kelmelis, Peter Messmer, Gallagher Pryor, Ryan Schneider High Performance Computing, Tools View Streamed Recording Download Video  Download Video    
1031 Designed for CUDA Program Overview As the leader in Visual Computing, NVIDIA provides extensive development and marketing support for key ISV partners.  ISVs can take advantage of NVIDIA’s  “Designed for CUDA” program, which provides  a wide range of marketing and development tools available exclusively to developers of compelling new CUDA-enabled consumer applications.  You’ll learn how to leverage NVIDIA’s worldwide marketing influence in PR, distribution, promotions, online presence and more.   In addition, we will discuss details of OpenAutomate, a technical resource that provides comprehensive compatibility testing, performance measurement and tuning across the full spectrum of NVIDIA configurations.  Technical prerequisites for certification of the program will be explained in detail, including automation, installation and licensing requirements.  John Spitzer, Michael Steele General Interest View Streamed Recording Download Video Download Video   Download PDF 
1032 Compute Application Testing  As the leader in Visual Computing, NVIDIA provides extensive development and marketing support for key ISV partners. ISVs can take advantage of NVIDIA’s “Designed for CUDA” program, which provides a wide range of marketing and development tools available exclusively to developers of compelling new CUDA-enabled consumer applications. You’ll learn how to
leverage NVIDIA’s worldwide marketing influence in PR, distribution, promotions, online presence and more. In addition, we will discuss details of OpenAutomate, a technical resource that provides comprehensive compatibility testing, performance measurement and tuning across the full spectrum of NVIDIA configurations. Technical prerequisites for certification of the program will be explained in detail, including automation,installation and licensing requirements.
John Spitzer   View Streamed Recording Download Video Download Video    
1033 Using CUDA within Mathematica Mathematica comes with many extremely optimized numerical libraries integrated into the application, but they don’t yet take advantage of the GPU. Thankfully, Mathematica provides an easy to use API for communicating between a large variety of external resources, called MathLink. This tutorial will provide a hands-on introduction to start using CUDA within Mathematica, an introduction to the cuda mathematica plugin, as well as the different issues one has to keep in mind when writing MathLink applications using the CUDA Toolkit. Finally we will showcase a few real-world examples. Kashif Rasul Tools View Streamed Recording Download Video Download Video    
1034 You Might Also Like: A Multi-GPU Recommendation System A recommendation system utilizes customers' known preferences to predict unknown preferences. We propose a multi-GPU algorithm for such recommendation systems. By an experiment, we have shown that our algorithm, when implemented on just two GPUs, runs more than 260 times faster than a single core implementation on a latest CPU. Kimikazu Kato Databases & Data Mining View Streamed Recording Download Video Download Video   Download PDF
1035 Harnessing the GPU for Surgical Training and Preoperative Planning Surgical simulation and preoperative planning systems require high performance solutions to complex spatial and physically based problems often over large datasets. We detail the use of NVIDIA CUDA and GPGPU techniques in surgical training and pre-operative planning systems currently under development.  Josh Passenger Life Sciences, Visualization, Medical Imaging & Visualization, Physics Simulation View Streamed Recording Download Video Download Video   Download PDF
1036 GPU Accelerated Solvers for ODEs Describing Cardiac Membrane Equations Mathematical models describing cellular membranes form the basis of whole tissue models to describe the electrical activity of entire organs, such as the heart.  Numerical simulations based on these models are useful for both basic science and increasingly for clinical diagnostic and therapeutic applications such as targeting ablation therapy for atrial arrhythmias, defibrillator design and cardiac resynchronization therapy. A common bottleneck in such simulations arises from solving large stiff systems of ordinary differential equations (ODEs) thousands of times for numerous integration points (representing cells) throughout a three-dimensional tissue or organ model. For some electrophysiology simulations, over 80% of the time is spent solving these systems of ODEs. While a cluster provides the required interactive response time to solve the ODEs, a desktop sized platform would enhance usability of the software in a laboratory setting.   The audience will benefit by learning how a real-world, complex, HPC application can directly benefit by the use of CUDA technology.  Participants will learn which optimization techniques yielded the best performance results on an actual application.  We will also explore the benefits and limits of the use of single precision in certain scientific applications. Fred Lionetti Algorithms & Numerical Techniques, Life Sciences, Visualization, Medical Imaging & Visualization View Streamed Recording Download Video Download Video   Download PDF
1037 Mapping Satellite Imagery on the GPU: Fast Orthorectification and Pan-Sharpening In the mapping and remote sensing industries, the ever-increasing availability of high-resolution satellite and aerial imagery in recent years has had two main effects. First, the data volume required for a given project size (for example, a county or state) has increased due to improved spatial resolution of the data. Second, the improved availability of coverage is leading geospatial data producers to handle larger and larger projects. To better serve our customers in these industries, we have improved the throughput in our core image processing modules by taking advantage of the high performance offered by NVIDIA GPUs. Numerous software modules in our new ProLines GeoImaging Server product are being reengineered to support this technology, but the first two completed modules (orthorectification and pansharpening) have already demonstrated a significant performance benefit (65x and 14x over conventional methods).  Teodor C. Hanchevici, James Lutes Computational Imaging View Streamed Recording Download Video Download Video    
1046 Research Summit Fast Forward with Bill Dally NVIDIA embraces emerging technology and those who dedicate their research to growth in visual and parallel Computing.  We believe that these minds lead the future in our industry and we are proud to support the 2009-2010 NVIDIA Fellows.  It is our pleasure to showcase their research during the Fellow Fast Forward.  Research topics include: Computer Vision, Ray Tracing, Quantum Chemistry, VLSI,  Shadow Mapping and much more.  For more information on this years Fellows:  2009-2010 NVIDIA Fellows, please visit Bill Dally General Interest Sorry - this session is not available.        
1047 Computer Vision with Horst Bischof The area of computer vision is entering an exciting period because it is for the first time in its history that we are not limited by the hardware, we have now enough camera resolution, enough memory and computing power and access to virtually an unlimited mount of images via the internet. The challenging question we have to answer now, is how to make use of that.  In this talk I will highlight some of these recent technological trends and how they influence the area of computer vision and more important speculate about their future impact on the field.
Among the topics I touch in the talk are computational photography, interactive vision, internet vision and of course GPU-based visual computing.
Horst Bischof Computational Imaging, Computer Vision   Download Video      
1048 Interactive Ray Tracing with the OptiX ray tracing engine OptiX is the latest and most flexible API for interactive ray tracing, taking full advantage of the power and generality of the CUDA architecture.  This new technology is valuable for anyone wanting to build a high-performance, ray tracing renderer (interactive or off-line), accelerate an existing ray tracing renderer, add raytrace capabilities to raster renderers, or even perform generic ray tracing functions for computation.  By replacing fixed components of a traditional ray tracer with user-specified programs, OptiX is able to implement a rich array of rendering and non-rendering algorithms while still maintaining high performance.  Examine these programmable components, and the flexibility they provide, through a guided tour of this exciting new engine. Steven Parker General Interest, Visualization   Download Video      
1049 Supercomputing Super Session GPU computing is transforming the extreme high-end realms of supercomputing. NVIDIA Tesla GPUs already power several of the world’s sixty fastest supercomputers, and this trend is accelerating.  This three-hour “super session” will feature some of the world’s premiere supercomputing experts, who will discuss their experience building and deploying GPU-based supercomputing clusters, and present case studies of designing and porting codes for “big iron” GPU supercomputers.  Wen-mei Hwu, Satoshi Matsuoka, James Phillips, John Taylor, Jeffrey Vetter High Performance Computing   Download Video      
1050 Quantum Chemistry on the GPU: Accelerating DFT Calculations In this talk we describe a high performance exchange-correlation Fock matrix implementation in DFT accelerated using NVIDIA CUDA toolkit. A BLAS3 kernel based algorithm has been developed to achieve sustained performance on both GPU and CPUs. Benchmark calculation shows more than 30 times speedup using CPU (Quadcore Phenom X4 940) and GPU (Tesla C1060) together compared with original code on single Phenom core. Optimization techniques and heterogeneous computing strategies for acceleration computing will be discussed and illustrated with implementation examples.  Zhengting Gan Quantum Chemistry View Streamed Recording Download Video Download Video   Download PDF
1053 GPU Accelerated Visualization and Analysis in VMD State-of-the-art graphics processing units (GPUs) contain hundreds of processing units  and are able to perform trillions of floating point arithmetic operations per second. The newly  available computational power brought by GPUs is enabling a new generation of scientific and  engineering applications to perform calculations on "personal supercomputers" that previously  required HPC clusters or that were otherwise impractical in everyday use. This talk will present  recent successes in multi-GPU acceleration in VMD, a molecular dynamics visualization and  analysis application in which GPU computing techniques have provided speedups ranging from  10 to over 100 times faster than commodity CPU cores. The talk will describe key challenges and  algorithmic strategies involved in achieving high computational performance on GPUs, discuss  methods for effectively using multiple GPUs in low-latency calculations that drive interactive  visualizations, and will also include some examples of how these performance increases  ultimately enable better science.  John Stone 3D, Visualization, Molecular Dynamics, Life Sciences View Streamed Recording Download Video  Download Video   Download PDF
1055 What Every CUDA Programmer Needs to Know About OpenGL CUDA has attracted thousands of new developers to GPU computing.  Many of these new programmers have pursued CUDA development without any prior exposure to OpenGL or experience with graphics programming, but nonetheless could benefit from the powerful  visualization capabilities GPUs also have to offer.  This talk is intended to provide an introduction to OpenGL graphics programming specifically oriented towards the non-graphics CUDA programmer.  We'll discuss basic OpenGL operations, simple drawing, image/video display and CUDA / OpenGL interoperability concepts for rendering data generated from CUDA programs.   Joe Stam General Interest, Tools View Streamed Recording Download Video  Download Video   Download PDF
1056 Multiparticle Simulation  A diverse array of science, engineering, and computer graphics applications involve simulations of large numbers of particles. These involve computation of interactions between many particles, potentially mediated by a spatial data structure such as a grid. Improvements in computation efficiency can be achieved by sorting particles to determine which particles are involved in interactions or undergo close approaches. Nearest neighbor or collision pair groupings can be used to reduce the total number of computation steps by reducing the number of queries for collisions or can speed up and improve accuracy of simulations via a multiple timestep integrator. Identification of nearest neighbor and collision partner groupings is a task that can be efficiently implemented in parallel on the GPU reducing the number of interactions that must be computed. A broad class of problems known as Particle-In-Cell (PIC) code advect particles through cells of a surrounding grid. During this roundtable we will discuss strategies for increasing the efficiency of multiparticle simulations as a general problem as well as challenges for multiparticle simulation in specific settings such as astrophysics, SPH, PIC, and granular flows. Alice Quillen Astro GPU View Streamed Recording Download Video  Download Video     
1057 NEXUS: A Powerful IDE for GPU Computing on Windows  Unleash your productivity with NEXUS, NVIDIA’s new development environment for GPU Computing and graphics applications that use CUDA C, OpenCL, DirectCompute, Direct3D, or OpenGL. NEXUS introduces native GPU debugging and platform-wide performance analysis tools for both computing and graphics developers, fully integrated into Visual Studio 2008. 
In this session, you will learn to use the powerful NEXUS debugger to set breakpoints and step through your CUDA C kernels or HLSL shader source code. Easily catch hard-to-find bugs using the NEXUS memory inspection tools and set data breakpoints on your running application. Visualize your application’s workloads and performance characteristics across the CPU, GPU, and operating system, and then dig deeper using the NEXUS profilers. NEXUS also includes state of the art graphics debugging features and convenient API state inspection.

This session will include information on how to register for the NEXUS Beta Program. 
Sebastien Domine General Interest, Tools View Streamed Recording Download Video  Download Video   Download PDF
1058 Tridiagonal Solvers on the GPU and Applications to Fluid Simulation This presentation will explore the efficient GPU implementation of direct numerical simulation of turbulent viscous incompressible fluid in 3D domain. We will discuss solving the full system of Navier-Stokes and energy equations using the Alternating Direction Implicit (ADI) numerical method, as well as implementation details of a fast tridiagonal matrix solver on CUDA . Finally we will compare the performance of GPU and CPU on a particular modeling problem in which the GPU outperforms the latest multicore CPUs by an order of magnitude in double precision on the whole solver. Nikolai Sakharnykh Algorithms & Numerical Techniques, Tools, Computational Fluid Dynamics , Astro GPU View Streamed Recording Download Video  Download Video   Download PDF
1059 Fast Tridiagonal Solvers on GPU We study the performance of three parallel algorithms and their hybrid variants for solving tridiagonal linear systems on a GPU: cyclic reduction (CR), parallel cyclic reduction (PCR) and recursive doubling (RD). We develop an approach to measure, analyze, and

optimize the performance of GPU programs in terms of memory access, computation, and control overhead. We find that CR enjoys linear algorithm complexity but suffers from more algorithmic steps and bank conflicts, while PCR and RD have fewer algorithmic steps but do more work each step. To combine the benefits of the basic algorithms, we propose hybrid CR+PCR and CR+RD algorithms, which improve the performance of PCR, RD and CR by 21%, 31% and 61% respectively. Our GPU solvers achieve up to a 28x speedup over a sequential LAPACK solver, and a 12x speedup over a multi-threaded CPU solver.
Yao Zhang Algorithms & Numerical Techniques, Astro GPU, Computational Fluid Dynamics , Tools View Streamed Recording Download Video  Download Video   Download PDF
1060 Graphcuts with CUDA and Applications in Image Processing Graph Cuts is a powerful and popular optimization approach to solve image processing problems such as image segmentation, stereo vision, image restoration and many more. In this talk, we present CUDA implementations of the push-relabel algorithm to compute Graph Cuts. Starting from the basic algorithm we discuss its parallel processing properties. Then different optimization strategies are explored and their strengths and weaknesses are evaluated. We conclude by exploring applications of Graph Cuts to solve image processing problems using GPUs. Timo Stich Computational Imaging, Medical Imaging & Visualization View Streamed Recording Download Video Download Video   Download PDF
1061 Real-time Visual Effects in Film and Video: the GPU Changes Everything Powerful fully programmable GPUs and stream processing technologies such as CUDA  are revolutionizing the field of professional film and video visual effects. Using stream- computing techniques, high quality effects can now be produced in real time, even at high  definition resolutions and above. Traditional OpenGL techniques did not have enough image  quality for professional use, and CPUs are too slow, even in parallel.  Massively parallel  computation changes the game for artists, who will be able to focus on telling the story rather  than waiting for frames to render; studios, whose return on investment will increase  measurably; and for viewers, who will reap the benefits of increased realism and fewer  distractions from the story line. This will have future implications not only for film and video, but  for online pro/consumer videos and games. Gary Oberbrunner 3D, Visualization, Film View Streamed Recording Download Video       
1062 Astrophysical Fluid Simulation Using Adaptive Meshes Adaptive mesh fluid simulations play a crucial role in many areas of astrophysical research including the formation and explosion of stars, jets from black holes, etc. A parallel adaptive mesh multi-physics fluid code, Enzo, has been widely used in astrophysical community in recent years. In this talk I will describe a CUDA implementation of the finite volume fluid solver used in Enzo. The GPU version shows significant speed-up compared to the CPU version. Peng Wang Astro GPU, High Performance Computing View Streamed Recording Download Video Download Video   Download PDF
1065 Diesel-Powered GPU Computing: Enabling a Real-Time Radio Telescope in the Australian Outback The Murchison Widefield Array (MWA) is a next-generation radio telescope currently under construction in the remote Western Australia Outback. The raw data rate is 5 to 20 GiB/sec, precluding offline processing. Since the computing budget for calibration and imaging is 20 TFLOP/sec, a real-time high-performance computer is required on-site. We describe a scalable heterogeneous computing pipeline implementation, exploiting both the high computing density and FLOP-per-watt ratio of modern GPUs. The architecture is highly parallel within and across nodes, with all major processing elements performed by the GPUs. Necessary scatter-gather operations along the pipeline are loosely synchronized and implemented in MPI. Our initial port to NVIDIA hardware shows a typical 10x improvement over the reference CPU implementation, with some portions showing even more substantial gains. The MWA will be a frontier scientific instrument and a demonstrator for planned peta- and exascale facilities. Richard Edgar Astro GPU View Streamed Recording Download Video Download Video    
1066  Astronomy & Astrophysics Astronomy, astrophysics, and cosmology face very large computational challenges. The physical accuracy of simulations depends on computing capacity, often with steep scalings. Moreover, astronomical observing systems that are in design and construction involve data rates of tens and hundreds of gigabits per second. Facilities like the Square Kilometer Array (SKA) telescope, in 2020, will dwarf even these, generating exabytes of data each day. These computational challenges push well beyond the capacity of traditional approaches, and solutions that draw upon high-efficiency parallel implementations and lower power consumption are required. This roundtable will discuss the use of GPU computing in the field of astronomy and astrophysics. This will include current GPU-enabled research areas, as well as potential new areas within the field. Christopher Harris, Alice Quillen Astro GPU Sorry - this session is not available.        
1068 OpenCL Optimization In this session, we will discuss how to optimize OpenCL programs on NVIDIA GPUs. Three main  aspects are discussed: memory, execution configuration, and instruction throughput. On memory optimization, we will discuss how to increase bandwidth by global memory coalescing  and using local memory. Then we will discuss the concept of occupancy and various considerations in specifying the execution configuration of a kernel. Finally, we discuss techniques for improving instruction throughput. Peng Wang Film, Medical Imaging & Visualization, Tools, Visualization View Streamed Recording Download Video Download Video   Download PDF
1069 Zombies on Tegra: A Case Study in Mobile Augmented Reality The NVIDIA Tegra family of ARM-based mobile application processors integrates extensive visual processing acceleration for amazing graphics, video, image processing and audio that can be combined in powerful ways.  This presentation features the ARhrrrr augmented reality zombie game from Georgia Tech to illustrate how Tegra can seamlessly recognize features from a handheld camera image and composite the video with synchronized 3D graphics.  The power of Tegra is accessed through open standard APIs including OpenGL ES, OpenMAX IL and OpenKODE contained in the Tegra software development kit.  This session will provide real-world insights to demonstrate how leading-edge mobile silicon and API standards come together to create a powerful and compelling mobile visual computing platform. Lars Bishop, Blair MacIntyre Computer Vision, Visualization, Embedded & Mobile View Streamed Recording Download Video Download Video    
1070 Face Recognition for Photographs and Video Face Recognition in Video (FRiV) is a challenging task. Successful solution of FRiV problem requires state of the art algorithms for visual analysis as well as their very optimized implementation in order to make the whole system working on off the shelf hardware. Presentation shows how computational performance of Viewdle FRiV technology is improved by harnessing power of GPU based computing using NVidia CUDA technology. Transition from CPU based implementation to GPU implementation is described. Changes in computing model made for efficient execution of computationally intensive technology parts on GPU hardware are presented. Essential improvement of processing speed of Viewdle FRiV technology based on GPU computing is demonstrated. Problems arisen during transition are analyzed. Features needed from GPU to make FRiV technology working even faster are presented.  Yuriy Musatenko Computational Imaging, Computer Vision View Streamed Recording Download Video Download Video    
1071 Face Detection The task of face detection in various types of multimedia content has emerged quickly during the last decade to become essential element in many consumer and specialized applications, hardware, SoC. Traditionally the task of face detection has been executed on CPU; however the algorithms haven't changed significantly during the last few years. The latest trend of moving computationally expensive general purpose calculations to GPU couldn't have no influence on the particular task of Face Detection, which resulted in appearance of specialized Computer Vision libraries written for GPUs. Anton Obukhov Computational Imaging, Computer Vision View Streamed Recording Download Video Download Video    
1073 Programming the Professional Video Pipeline The Quadro Professional Video Pipeline enables the binding of SDI video input and output devices directly to the GPU.  This provides applications with the extraordinary capability to performing capture, processing and then output live to air or other video device completely in video memory on the GPU.  Gone are the days of needing a CPU and lots of system memory for video processing.  Come learn how to capture SDI video directly to GPU memory for subsequent processing as Video Textures, Video Buffer Objects or in CUDA.  Once in the GPU, the video processing possibilities are endless, color grading, transcoding, compositing, restoration, effects etc.  Then, output the ultimate quality imagery directly from GPU memory live to air, the network, presentation system or other video recording or display device.  This session will present the Quadro Profession Video API and solution stack with case studies and usage examples. Thomas True Computational Imaging, Film, Visualization, Tools View Streamed Recording Download Video Download Video    
1074 Computational Fluid Dynamics (CFD) for the GPU The field of computational fluid dynamics (CFD) has far-reaching applications and displays a consistent need for larger and faster simulations. At EM Photonics we have been studying this field and its computational needs for two years. We have identified the GPU as a strong performer in the CFD field and as such have implemented solvers that harness the power of GPUs in the application of CFD formulations. We will present some background on these innovations in this summary discussion. John Humphrey, Daniel Price Astro GPU, Physics Simulation, Computational Fluid Dynamics  View Streamed Recording Download Video Download Video    
1075 Reconstructing the Brain: Extracting Neural Circuitry with CUDA and MPI Determining the detailed connections in brain circuits is a fundamental unsolved problem in neuroscience. Understanding this circuitry will enable brain scientists to confirm or refute existing models, develop new ones, and come closer to an understanding of how the brain works. Prof. Jeff Lichtman and Center for Brain Science (CBS) at Harvard launched the Connectome Project three years ago to determine the complete, detailed wiring diagrams of neural circuits from sequential high-resolution images of the central nervous system using electron microscopy (EM). These high-resolution, large-scale EM datasets pose very challenging computational problems for 3D segmentation and visualization in terms of developing suitable algorithms, coping with the ever-increasing data sizes, and maintaining interactive performance. <p> In this talk we will present our insights and lessons learned in using CUDA to reconstruct neural connections in high-resolution EM data. We will present technical details and non-trivial issues regarding the implementation of NeuroTrace, our system for semi-automatic segmentation and interactive visualization of terabytes of EM image data. The segmentation method is based on a sequence of 2D level set segmentations of cell membranes integrated with an image correspondence energy for robust transition between consecutive slices and a weighted path extrapolation method to trace a 3D centerline of a neural pathway along non-axis aligned slices.  The visualization method employs view-dependent on-demand filtering and evaluation of a local histogram edge metric, as well as on-the-fly interpolation and ray-casting of implicit surfaces for segmented neural structures. We will describe the implementation details of both methods in CUDA for interactive performance. We will also present a novel client-server implementation of NeuroTrace that uses distributed processing with MPI on a CUDA compute cluster with eight sS1070 Tesla and 16 compute nodes.  Won-Ki Jeong Computational Imaging, Medical Imaging & Visualization, Life Sciences, High Performance Computing View Streamed Recording Download Video Download Video    
1076 Optimizing Ion Channel Kinetics Using A Massively Parallel Genetic Algorithm on the GPU Voltage-gated ion channels effect the integration of information in many neurons. Some neurons express over 10 voltage-gated channels that turn information processing into a highly non-linear affair. The only path allowing dissection of the role of each channel in neuronal physiology passes through kinetic modeling of the investigated channels.  The currently popular analysis techniques suffer from various shortcomings that limit the ability of the researcher to rapidly produce physiologically relevant models of voltage-gated ion channels. We have recently shown that combining a stochastic search algorithm with ionic currents measured following multiple voltage-clamp protocols enables to semi-automatically constrain models of voltage-gated ion channels. This approach, while numerically successful, is highly demanding computationally with optimization lasting typically for several days on a 160 CPU Linux cluster. To solve this computational bottleneck we have been converting our optimization algorithm to work on a GPU using CUDA.  We have succeeded to parallelize the process on a GTX 295 giving a speed increase of roughly X100 over that of the CPU. This shortened simulation times to hours instead of days. More importantly, it allows interactive modification of the modeling process by the user.  Once fully automated and expanded to machines containing several GPU cards, this approach may be used for rapid online analysis of voltage-gated currents.  This in turn will allow researchers to greatly shorten the time and costs required for building models of neuronal physiology, which will facilitate our understanding of neuronal physiology. Alon Korngreen Algorithms & Numerical Techniques, Visualization, Physics Simulation View Streamed Recording Download Video Download Video    
1077 Bullet: A Case Study in Optimizing Physics Middleware for the GPU We have been porting and implementing all stages of the collision detection and rigid body dynamics pipeline onto GPU. We describe the changes in data structures and algorithms to parallelize for CUDA and OpenCL. The implementation is available as open source in the Bullet Physics SDK.  Erwin Coumans General Interest View Streamed Recording Download Video Download Video   Download PDF
1078 Adobe Photoshop CS4 and GPUs Adobe Photoshop CS4 and Adobe Bridge CS4 take advantage of the processing power inherent in GPUs. When combined with GPU power, large Photoshop files can be manipulated faster and smoother than ever before. In this session, you'll see how Photoshop excels when used with NVIDIA GeForce and Quadro GPUs.  Jerry Harris, John Nack Computational Imaging, Film, Medical Imaging & Visualization, Visualization Sorry - this session is not available.        
1079 Scalable Multi Agent Simulation on the GPU Multi agent systems have been recently gaining increased attention by game AI developers, mainly in seeking a robust motion planning solution for non player characters. This session addresses some of the principal design considerations for a scalable graphics hardware implementation of collision avoidance simulation, using NVIDIA's CUDA. In particular, it highlights the tremendous performance payoff of the GPU when exploiting nested data parallelism. It also covers the topic of formalizing the integration of game AI and physics tasks on the GPU to form a collaborating and a more intuitive simulation framework to the game developer.  Attendees will acquire a number of highly optimized algorithms in various areas of real-time, multi agent simulation. Avi Bleiweiss Algorithms & Numerical Techniques, Machine Learning & Artificial Intelligence View Streamed Recording Download Video Download Video   Download PDF
1080 The Art of Performance Tuning for the CUDA Architecture This birds of a feather session is an open discussion on the art of performance tuning for the CUDA architecture.   General Interest, Tools View Streamed Recording Download Video Download Video    
1082 Advanced Numeric Computing Accelerating Discontinuous Galerkin Finite Element Methods

This talk will introduce the high-order discontinuous Galerkin (DGTD) finite  element methods for solving the time dependent Maxwell's equations and  highlight how the methods naturally lend themselves to thread block  GPU implementations. The DGTD methods require block dense  matrix-vector operations on the solution field textures and this can  be naturally handled through a combination of texture and shared  memory. Thae talk will describe several implementation strategies and discuss  how our initial successful CUDA implementation of DGTD led to new  algorithms for handling meshes with curvilinear tetrahedral elements.

 Computational results will be presented to demonstrate the efficacy of  mapping the DG methods onto multiple GPUs, achieving over a teraflop  on a workstation. We will also discuss how the heavyweight nature of  each GPU enables load balancing for multiple GPUs even when the  underlying agglomerated workload graph has a wide spread in work  weights.
Nathan Bell, Timothy Warburton Physics Simulation View Streamed Recording Download Video Download Video    
1083 CUDA Fortran Programming for NVIDIA GPUs This talk provides an introduction to programming NVIDIA GPUs using CUDA Fortran. It is suitable for expert Fortran or CUDA C programmers who need to extract maximum performance from GPUs using an explicit GPU Fortran programming model.  This talk introduces the CUDA Fortran language, and through examples, illustrates how to explicitly program NVIDIA GPUs in native Fortran 95/03 through creation of GPU kernel subroutines, management of host and GPU device memory, definition of CUDA grids and thread blocks, launching kernels on an NVIDIA GPU device, and use of the CUDA Fortran runtime API.  This talk includes a live component with a Linux workstation containing a Tesla card, and the PGI CUDA Fortran compiler. Michael Wolfe High Performance Computing, Tools View Streamed Recording Download Video  Download Video    
1084 The TotalView Debugger for CUDA For applications to take advantage of the performance boost offered by GPU Computing technologies and techniques, developers need to either write or re-write their programs in CUDA or another language supported by the GPU. That development process is much more efficient if developers can rely on a good debugger to give them control over and visibility into the program that they are writing. This talk will discuss a preview version of the TotalView debugger specifically developed to support CUDA development on NVIDIA GPU hardware. Ed Hinkel Tools View Streamed Recording Download Video Download Video    
1085 Debugging Tools for CUDA Explore the ease of debugging heterogeneous applications using the integrated CPU and GPU debugging environment of CUDA-GDB. CUDA-GDB is NVIDIA's very first native GPU debugger that lets you debug CUDA applications on the Linux platforms. In this session you will learn about the capabilities of the CUDA-GDB debugger with a walk through of the simplicities of being able to debug complex parallel applications. CUDA-GDB is also a platform to built custom debuggers so if you plan to build a CUDA debugger this session will provide the details on building your very own CUDA debugger. Satish Salian General Interest, Tools View Streamed Recording Download Video Download Video    
1086 Advanced C for CUDA This presentation covers the major CUDA optimizations. Topics will include: maximizing memory throughput, kernel launch configuration, using shared memory, and improving GPU/CPU interaction. We will also describe how the CUDA Visual Profiler and code instrumentation can be used to assess performance. While C for CUDA is used for illustration, the concepts covered will apply equally to programs written with OpenCL and DirectCompute APIs. Thomas Bradley, Paulius Micikevicius Tools View Streamed Recording Download Video Download Video    
1087 High Dynamic Range Spectral Imagery X-Rite specializes in the construction of devices that acquire highly accurate color data. As part of our advanced development, we have engineered a process for generation of high dynamic range (HDR) spectral images from digital CCD cameras using highly characterized LED illumination sources. Our instrumentation involves multiple cameras and illumination sources arranged to produce a gonio-spectrophotometric set of images. Within this process, standard image processing techniques exacted a high computational cost and initially prevented this process from being attractive and useful. CUDA enabled us to achieve a roughly 140x computational improvement over the original implementation, both by parallelizing the computations via the GPU and by encouraging specific algorithm optimizations.  Marc Ellens Film, Visualization View Streamed Recording Download Video  Download Video    
1088 Medical Image Registration with CUDA This session will discuss the exceptionally fast 3D Medical Image Registrations achieved using CUDA.  Speedups of up to 750 times were obtained as compared to code in daily use at Addenbrookes Hospital and Bio-Medical Campus. Some very recent results are shown in the figures.  This work is of direct application in both research and clinical practice.  A particular application is voxel based MRI morphometry in humans and in animal brains.  Here small changes in particular regions of the brain are studied to monitor disease progression, response to treatment and effects of genetic variation etc. Richard Ansorge Computational Imaging, Medical Imaging & Visualization View Streamed Recording Download Video Download Video    
1090 Spatial Data Structures for Massively Parallel Computing Many applications in physical simulation and rendering require the use of spatial data structures to achieve optimal performance. Spatial data structures can be used to accelerate collision detection, ray tracing and molecular dynamics simulations. This talk will describe how data structures such as uniform grids and bounding volume hierarchies can be efficiently constructed and traversed on the GPU using data parallel primitives such as sort and scan. Richard Tonge Algorithms & Numerical Techniques, Physics Simulation, Tools View Streamed Recording Download Video Download Video    
1091 Clusters with GPUs under Linux and Windows HPC Clusters with GPUs are becoming a popular choice in High Performance Computing. This talk will describe the steps necessary to deploy clusters with GPUs both under Linux and Windows Server HPC,ranging from working with headless compute nodes to the interaction with the queuing system. Calvin Clark, Massimiliano Fatica High Performance Computing View Streamed Recording Download Video Download Video   Download PDF
1100 Speed Limit Recognition System (Optimized for processing on the GPU)  In this study, we investigate the use of a programmable graphics processing unit (GPU) as an embedded processor for real-time recognition of speed limit signs on the road. The input to our system is a video sequence of the road taken from a moving vehicle. We process this video in real-time and determine if there are any speed limit signs present in the scene and, if so, we recognize and output the number indicated by the sign.

The main goal of the recognition system is to operate in real time on a resource-constrained embedded system. Therefore, we first examine the merits and demerits of mapping algorithms often used for speed-limit recognition on to the GPU. Through this process, we find techniques that benefit significantly from the GPU architecture and eliminate algorithms that do not map efficiently on it. We then implement and analyze two sign detection schemes: one feature-based, one template-based. From the results of our experiments, we make several important conclusions about the trade-off between recognition rates and performance. We also make an estimate for the amount of hardware resources needed to perform the recognition in real-time."
Vlad Glavtchev, Pinar Muyan-Ozcelik, Jeff Ota, John Owens Embedded & Mobile, General Interest View Streamed Recording Download Video Download Video    
1101 GPU Computing in Java: Generation and Scheduling of PTX Assembly This talk will discuss an ongoing progress regarding the development of a Java-based library for rapid kernel prototyping in NVIDIA PTX and PTX instruction scheduling. It is aimed at developers seeking total control of emitted PTX, highly parametric emission of, and tunable instruction reordering. It is primarily used for code development at ICHEC but is also hoped that NVIDIA GPU community will also find it beneficial.  Christos Kartsaklis Tools View Streamed Recording Download Video  Download Video    
1102 4D Volume Rendering With advances in image acquisition and numerical simulation techniques, fields ranging from medical imaging to astrophysics are producing data with very large spatial and temporal resolutions. Interactive visualization techniques are crucial to understand and isolate features from the resulting large time dependent 4D volumetric data. This presentation explores the various rendering methods such as texture slicing, raycasting in graphics and cuda as well as hybrid approaches showing their promises and pitfalls. It is common for 4D  data to exceed the graphics memory capabilities and approaches for efficiently streaming data such as PBO’s and CPU/GPU asynchronous modes are explained. We conclude with a discussion on how other related solutions from NVIDIA can be integrated, specifically focusing on 3D Vision stereo and NVScale middleware to harness multiple GPU’s for distributed rendering. Shalini Venkataraman 3D, Visualization, Medical Imaging & Visualization, Film, Energy Exploration View Streamed Recording Download Video     Download PDF
1105 SQL/XML-IMDBg: A GPU In-Memory Database and Query Co-Processor Attendees will learn about the architectural design decisions and implementation strategies chosen in migrating an in-memory database system and query execution engine to execute database workloads concurrently on CPU and GPU hardware. We will also give insights into the optimizer structure and query executer for executing the "split-work" plan simultaneous on CPU and GPU hardware. Harald Frick Databases & Data Mining View Streamed Recording Download Video     Download PDF
1106 Driving on Mars: Simulating Tracked Vehicle Operation on Granular Terrain This work demonstrates the use of high performance GPU computing for the numerical simulation of tracked ground vehicles. The work closes a gap in physics based simulation related to the inability to accurately characterize the 3D mobility of tracked vehicles on granular terrains (sand and/or gravel). The problem of tracked vehicle mobility on granular material is approached using a discrete element method that accounts for the interaction between the track and each discrete particle in the terrain. Presently the approach captures the dynamics of systems with more than 300,000 bodies interacting simultaneously. The method is implemented on the GPU through CUDA. Toby Heyn, Justin Madsen, Hammad Mazhar, Dan Negrut Physics Simulation, Visualization View Streamed Recording Download Video      
1108 Exploiting the GPU in Ultra High-End 4K Video Servers Zaxtar is the highest performance video server that will feed up to 4GB/sec of video data to 4K projectors or 4K displays. This performance is accomplished by utilizing CPU, GPU and synchronization of multiple computers.

The most important feature is mathematically lossless compression, which can compress video or graphics data at the ratio of 3 to 1 without losing any information. Mathematically lossless compression has been achieved by CPU up until now, but we have ported the algorithm to Quadro FX 5800.
Mark Marrin Algorithms & Numerical Techniques, Visualization, Computational Imaging View Streamed Recording Download Video      
1109 Building a Modern Engine for the Visual Simulation Industry Blue Newt Software has been designing visual software for more than 5 years, and in late 2008 won a contract to redesign the visual rendering engine for a series of driving simulators for a large German automotive company. This company has both very high quality demands and firm performance constraints making this a challenging problem. This session will discuss the development process and graphics technologies employed in our modern visual rendering and simulation engine.

To meet these needs we designed a component-based visual engine, a deferred-rendering graphics engine, and a fully-scriptable API. We provide support for full high-dynamic range rendering, a series of post-processing effects, and a data production toolchain for data ingest and usage. We use modern graphics rendering techniques including instancing, shading, buffered objects, and more.

We'll describe the problem domain, our solution and product, and the technologies we employed in our approach. We'll describe how we took the requirements of an industry-leading customer and created a product ready for their future. We'll also detail how we researched and employed best-practices from the game industry and applied them to the tasks and requirements of the visual simulation industry.
Bob Kuehne Embedded & Mobile, Visualization, Physics Simulation View Streamed Recording Download Video Download Video    
1110 Using the GPU for Gradient Reconstruction of Unstructured Meshes Impressive performance gain has been obtained on field calculations for large volume of seismic survey data which is hierarchically represented [1]. Gradient reconstruction, for both scalar and vector unstructured fields, is yet another performance critical task in engineering simulations such as computational fluid dynamics (CFD) and finite element analysis (FEA). The latest GPU hardware has been improved significantly in terms of memory capacity and memory random access efficiency, which makes GPU computing attractive to engineering simulation[2]. Based on requirements drawn from cross disciplinary fields including geophysical modelling, material analysis and manufacturing, this study continues to investigate the double precision performance and its scalability across multiple GPUs. A software framework is designed where algorithms can be conveniently implemented in a heterogeneous computing environment with mixed CPU and GPU configurations. Attention has also been steered toward integration of GPU algorithms in end-user application software Avizo, which will enable the application of the algorithm in industrial aerodynamic simulations where mixed element unstructured mesh dominates. Michael Heck Algorithms & Numerical Techniques, Physics Simulation, Visualization, Computational Fluid Dynamics  View Streamed Recording Download Video      
1111 Industrial Simulation Using Massively Parallel Algorithms The purpose of the talk is to show that there is an exciting industrial market for next generation supercomputing based tools, and what direction this is going in. This talk is a result of a 5 year research within the Oil&Gas sector, identifying new areas where Supercomputing will enhance existing use cases in a game changing manner. The cases are grounded in real needs from operational environments at the world leading Oil&Gas operator StatoilHydro. Thorolf Tonjum Energy Exploration, High Performance Computing, Visualization View Streamed Recording Download Video  Download Video   Download PDF
1112 Visualizing the Universe: Raycasting Astrophysical Simulation Data We use GPU-assisted raycasting to render large, three-dimensional time-dependent astrophysical AMR data sets at interactive frame rates on standard desktop computers. Our approach allows us to embed unstructured point datasets, like stars or galaxy splats, into the rendering of gaseous interstellar or intergalactic material. The approach supports a combined color-mapping of several input data fields and allows for a very flexible adaption to the special requirements of different types of simulations. Its interactivity makes it a useful tool for data analysis as well as for fast generation of high-quality animations from astrophysical datasets. We will show various resulting animations ranging from large scale structure formation in the early universe, to the evolution of the first stellar object and the cosmological reionization era. Finally, we will give an overview about lessons learned and opportunities for future work. Ralf Kaehler Astro GPU View Streamed Recording Download Video      
1113 Programming NVIDIA GPUs using PGI Accelerator Fortran and C Compilers This talk provides an introduction to programming NVIDIA GPUs using the PGI Accelerator Programming Model in C and Fortran. It is suitable for application programmers, in particular those who are not expert GPU programmers. This talk introduces the compute-specific details of the NVIDIA GPU, and through examples, illustrates how to program common computational algorithms on NVIDIA GPUs using portable directive- based C and Fortran 95/03. The material covers programming language features, interpreting compiler feedback, performance analysis, and performance tuning. This talk includes a live component with a Linux workstation containing a Tesla card, and the latest PGI Accelerator compilers and tools. Michael Wolfe Algorithms & Numerical Techniques View Streamed Recording Download Video       
1114 Strategies for GPU Acceleration of Common Visual Effects for Broadcast and Post-Production Since 1995 BorisFX has developed image processing and 3D graphics software for Broadcast and Post-Production, with a particular focus on staple visual effects such as 3D Text, Chroma Key, and Film Look. While achieving award-winning quality, these CPU bound effects have lacked the interactivity today's customers demand. With the advance of GPU hardware from NVIDIA the ability to accelerate these effects has become a reality, and allowed for many previously impractical features to present themselves. In this presentation we will demonstrate how the use of the GPU has benefited our products in terms of performance and features in our creation of GPU-Rendered 3D Text, and discuss the strategies we employed to emphasize the benefits and minimize the drawbacks of the GPU in building the multi-pass Chroma Key and Film Look filters.  Boris Yamnitsky, David Yamnitsky 3D, Film, Visualization View Streamed Recording Download Video       
1122 Maximizing GPU Efficiency in Extreme Throughput Applications NVIDIA GPUs provide the massive data throughput needed to enable real-time processing of high data volumes, such as those encountered in high definition and 4K video applications.  This talk will explore several important design considerations for extreme data rates including proper overlapping of data transfer and compute, multiple GPU approaches, and efficient interoperation with OpenGL. Joe Stam Computational Imaging, Tools View Streamed Recording Download Video      Download PDF
1123 Domain Specific Languages for Programming GPUs Processors and computing systems are becoming very heterogenous.  Imagine writing a program that runs on a cluster, a shared-memory multi-core processor, and a many-core GPU.  In each case the application needs to be completely rewritten.  The cluster will use MPI, the shared-memory system threads and locks, and GPU a data-parallel programming model like CUDA.  The cost of maintaining three different versions of a program is often prohibitive, and limits the adoption of GPUs in many applications.

My thesis is that the only way to program such diverse set of systems is to raise the level of abstraction and build domain-specific programming models that are able to automatically map applications to different types of platforms.  Our goal is to make it much easier
to adopt innovative new platforms.  In this talk, I will present the case for domain-specific programming models, and outline several projects that we are doing at Stanford as part of the Pervasive Parallelism Laboratory.
Pat Hanrahan General Interest   Download Video       
1130 SCVF: A Framework for Visualization of Grid and Particle Based CUDA Simulations The Stone Ridge CUDA Visualization Framework (SCVF) provides a user API that allows developers to easily visualize their simulations on 2D grid, 3D grid and free particle systems. It allows users to rapidly create visual demos of their CUDA enhanced applications and make direct visual comparison between GPU and CPU implementations.
William Brouwer, Ben Karel Visualization View Streamed Recording Download Video       
1134 Using OpenKODE to Create a Seamless Driving Experience The software for an in-car infotainment system requires many different graphical technologies, including 2D widget toolkits, 3D OpenGL ES graphics, navigation engines, and web browsers, all using different APIs and implemented with different coding techniques. Somehow, these technologies must all be brought together to operate as a single, purpose-built system. By leveraging the open standard OpenKODE API, development teams that use Nvidia GPUs can merge the output of these separate technologies into a single GUI, saving development effort while improving the user experience. In this session, attendees learn how the OpenKODE API can help development teams manipulate graphics to create the unified experience that customers demand. Andy Gryc Embedded & Mobile, Visualization View Streamed Recording Download Video      
1138 Rigid Body, Cloth and Fluid Physics for CUDA Since NVIDIA's acquisition of the PhysX technology in 2008, we've been working hard to port Rigid body, Cloth and Fluid PhysX libraries to CUDA. This presentation describes some of the algorithms, their parallelization and optimization for GPUs using CUDA. Richard Tonge Physics Simulation View Streamed Recording Download Video      
1139 CUDA Kernel Performance Analysis Using NVPerfMon Performance optimization is one of the more difficult aspects of kernel development for NVIDIA’s CUDA-enabled GPUs.  It’s often not easy to determine where compute cycles are spent, particularly complex kernels. In this talk I will describe a performance analysis methodology using NVPerfMon, a program annotation-based timing analysis framework that can be used to generate timelines of important events, such as thread start and stop times. NVPerfMon can provide deep visibility into the performance pathologies of a CUDA kernel and can expedite the time to attaining performance goals.  MultiCoreWare has extended Nvidia’s NVPerfMon tool for High-Performance Computing applications and environments, with full 64-bit capability, and full functionality in a multiple GPU environment.  Sanjay Patel High Performance Computing, Tools View Streamed Recording Download Video       
1140 GPU Debugging Made Easy with DDT One of the key components for exploiting the potential of GPUs is the debugging tool - the software through which problems with your GPU code can be resolved quickly and easily.  Allinea Software's DDT is a powerful debugger for multi-threaded and parallel codes, known for its ease of use and scalability on large systems and applications.  We will introduce Allinea DDT for NVIDIA GPUs which brings debugging of CUDA on real hardware - providing intuitive ways to control, visualize and debug CUDA threads. David Lecomber Tools View Streamed Recording Download Video       
1141 Energy Industry Applications: Three Examples of GPU Acceleration in the Oil and Gas Industry Using CUDA The energy industry has traditionally been one of the largest market sectors for High Performance Computing hardware, software, algorithms and related technologies. Many of the critical industry applications map well to GPU hardware and can benefit from the high floating point capability and lower price/performance ratios. This talk will address three representative industry algorithms and their mapping to the GPU i)Kirchhoff time migration for seismic processing ii) Sparse matrix solvers for reservoir simulation and iii) the split-step Fourier transform. An overview of these algorithms will be provided along with a discussion of their GPU implementations, reference performance numbers and a general discussion of price/performance comparisons for the GPU vs. traditional implementations. William Brouwer Algorithms & Numerical Techniques, High Performance Computing, Energy Exploration View Streamed Recording Download Video       
1142 Handheld Augmented Reality  Handheld Augmented Reality (AR) presents an exciting opportunity for mobile interaction and game designers.  By moving experiences off the screen and out into the physical world, handheld AR has the potential to sidestep the limitations of small mobile displays by giving users the illusion that they are looking through a window into a larger 3D space merged with the world.  And by attaching this virtual world to the physical world, AR creates new opportunities for physical and social interaction.  In this talk, I will discuss a range of issues that must be dealt with when creating mobile AR experiences, and illustrate these points with a serious of AR game prototypes, with a focus on social, physical, and tangible interaction. Blair MacIntyre Embedded & Mobile   Download Video      
1144 Use of High Performance GPUs in Military/Aerospace and Defense Applications  GPUs have long been used in Military/Aerospace and Defense applications, but with the advent of GPGPU computing the range of problems that will be solved by GPUs with the CUDA architecture is set to expand dramatically.  This session will look at some of the traditional and new applications for GPUs in the Military/Aerospace arena, together with some of the challenges involved in taking GPU technology into the hostile real-world environments of this market. Jim Berlin General Interest View Streamed Recording Download Video      Download PDF
1145 Integrated Design that Shatters 1U Server Performance Records Supermicro and NVIDIA® recently announced a new class of servers that combines massively parallel GPUs with multi-core CPUs in a single server system. This unique configuration delivers performance at least an order of magnitude better than traditional quad-core CPU-based servers. This breakthrough technology immediately provides users with the ability to implement tasks that were traditionally addressed only with massive supercomputers or that were simply unsolvable.  With GPUs and CPUs on a co-processing 1U server, data centers can achieve greater system manageability and a uniform building block for deployment.

Supermicro will provide an overview of this groundbreaking product in this session. 
Don Clegg General Interest View Streamed Recording Download Video      
1146 Next-Generation Graphics DRAM: Challenges and Opportunities GDDR5 is the most up-to-date graphics memory technology today, offering an impressive bandwidth of between 4 and 7Gbps. However, demand continues for the development of finer and finer graphics to be used in games as well as professional design applications.  So, the memory Industry is considering how it is going to move beyond GDDR5 levels in a couple of years to reach bandwidths of more than 10Gbps. Considerable obstacles await. Samsung research indicates that it will be difficult to overcome a few major barriers to attaining and exceeding data transfer speeds in excess of 10Gbps with the single-ended technology that is used today.

The industry will need to uncover a cost-efficient way to move beyond single-ended memory chip designs.  Memory vendors will have to closely examine a variety of factors including differential technology, and wide I/O schemes. This presentation will review GDDR5 technology and present various ideas on how best to exceed the bellwether threshold of 10Gbps.
Jimmy Chung General Interest View Streamed Recording Download Video       
1147 The High Life for High Performance  Samsung Semiconductor and Lenovo have teamed up to push the envelope on notebook performance and data reliability.  In this presentation, the two leading-edge companies will take a close look at market-ready, state-of-the-art notebook technology for high performance computing applications with case study information that focuses on real-world applications.  They will also discuss the PC industry’s decisive movement to solid state drives, and future technology for product designers. Henry Ramos, Brian Beard General Interest View Streamed Recording Download Video       
1200 Opening Keynote with Jen-Hsun Huang, CEO and Co-Founder, NVIDIA  Jen-Hsun Huang, NVIDIA’s CEO and Co-Founder will open the GPU Technology Conference with this can’t-miss keynote.
Jen-Hsun Huang General Interest View Streamed Recording Download Video (Small - 449MB-mp4 format) Download Video (Medium - 1G mp4 format) Download Video (Large - 2.6G mp4 format)  
1201 Important Trends in Visual Computing (General Session) Throughout the GPU Technology Conference, attendees will be exposed to some of the hottest trends and biggest ideas in the field of visual computing. In this session, moderated by David Luebke of NVIDIA Research, a collection of world-renowned experts will highlight and preview important visual computing trends ranging from computer vision, augmented reality and visual analytics.

Due to overwhelming response and limited capacity if you are interested in attending any of the keynotes or general sessions, be sure to schedule this session into your personal calendar.
Horst Bischof, Pat Hanrahan, Blair MacIntyre Computer Vision, General Interest, Visualization View Streamed Recording Download Video (Small - 297MB mp4 format) Download Video (Medium - 711MB mp4 format) Download Video (Large - 1.7G mp4 format)  
1202 Breakthroughs in High Performance Computing (General Session) GPU computing is one of the biggest topics of discussion in the HPC industry today.  This session, hosted by NVIDIA Chief Scientist Bill Dally, showcases perspectives from industry luminaries on the future of HPC and the role of the GPU.

Due to overwhelming response and limited capacity if you are interested in attending any of the keynotes or general sessions, be sure to schedule this session into your personal calendar.
Shawn Edwards, Steve Scott, Jeffrey Vetter General Interest View Streamed Recording Download Video(Small - 225MB mp4 format) Download Video (Medium - 612MB mp4 format) Download Video(Large - 1.5G mp4 format)  
1203 Using GPUs to Estimate the Value-at-Risk of Portfolios Financial institutions measure market risk by estimating potential losses in the market value of their traded assets. The estimation involves a simulation-based Value-at-Risk (VaR) computation over certain time horizon and provides a confidence interval to guide asset portfolio management decision making. This type of analysis remains one of the most ubiquitous techniques for measuring risk despite its high computational complexity.
In this note we describe the implementation of a VaR Monte-Carlo simulation on GPUs. By grouping random experiments into blocks, we are able to exploit the computational efficiency of the CUDA BLAS3 SGEMM routine to significant effect.  Using a portfolio of approximately 5000 instruments, we benchmark the performance of our CUDA implementation of the VaR simulation on a Nvidia GeForce GTX 280 against a 2.5 GHz quad-core Intel CPU. Simulation time is 6 times faster using the GPU than the CPU, reducing the simulation time to around 15 minutes. Further optimizations and performance diagnostics are presented.
Jike Chong, Matthew Dixon Algorithms & Numerical Techniques, Computational Finance View Streamed Recording Download Video       
1400 GPU Metaprogramming using PyCUDA: Methods & Applications Writing reliable GPU codes that achieve peak performance in the face of changing requirements and hardware platforms can be a challenging task. In this talk, I will introduce the open-source PyCUDA toolkit, which assists in this task in a number of ways: Convenient, high-level interface PyCUDA binds all functionality in Nvidia CUDA to a convenient interface in the high-level scripting language Python. Resource management and error checking are automatic.  Code Templates PyCUDA comes with tuned and debugged code for many common operations, such as vector math and reductions, which saves debugging and coding time.  Metaprogramming PyCUDA allows GPU code to be generated at run-time code, which makes many advanced programming techniques easy–such as empirical optimization, constant folding, and run-time specialization.  Scalability PyCUDA covers ”small-scale” and ”large-scale” uses alike: It allows quick prototyping and experimentation,but it also integrates easily into large-scale computational software.  Having introduced the toolkit, I will show how PyCUDA has supported a number of applications in computational science:
First, we have successfully used PyCUDA in a high-performance discontinuous Galerkin finite element (DG-FEM) solver. The term DG-FEM describes a family of high-order accurate numerical methods for systems of partial differential equations that model real-world processes such as electromagnetic scattering or fluid flow. We found that these methods’ algorithmic structure makes them very suitable for execution on a GPU, often achieving speedup factors on the order of 50 when compared to a single CPU core. PyCUDA and GPU metaprogrammingwere crucial in achieving this level of performance. As an added benefit, the resulting solver turned out to be very versatile with respect to equation types, domain dimensionality, and discretization parameters.  Second, I will discuss a recent effort seeking to automate the writing of high-performance GPU code for a large class of computational kernels that includes many of those needed for the numerical discretization of PDEs.  Again, code generation and empirical optimization as provided by PyCUDA provide the basis for the approach that may make GPU performance possible even in situations where manual development is not economical.
Andreas Kloeckner Tools View Streamed Recording Download Video     Download PDF
1401 Convolution Soup: A Case Study in CUDA Optimization This talk explores various optimization techniques using the simple example of image convolution. We begin with simple naive approaches and explore progressively more advanced optimizations and alternative memory access approaches, eventually arriving at an optimal implementation. Joe Stam Algorithms & Numerical Techniques, Computational Imaging View Streamed Recording Download Video     Download PDF 
1402 Applications of Graphics Processing Units to the Binary Black Hole Evolutions We apply general-purpose computation on GPUs to obtain sizable speedups over a CPU in post-Newtonian evolutions of a binary black hole system. We discuss effective techniques for optimizing our GPU code on the CUDA architecture and present results demonstrating the speedups obtained. We also describe an MPI-based approach for scaling a large number of binary black hole simulations over multiple GPUs. This approach will allow us to complete the largest scientific GPU calculation to date using the NCSA Lincoln cluster. John Silberholz Astro GPU View Streamed Recording Download Video      
1404 Real Time Cloth: 3D Triangular Mesh Algorithms on the GPU Rendering has been the classic challenge of 3D graphics. But today, with complex scenes having millions of polygons and wanting to add physical effects --just moving and deforming those polygons presents a growing challenge especially when trying to achieve real time performance. In some dynamic scenes, tens of thousands of vertexes need to have their

new position recalculated for each frame and in some cases (like physical simulation) every sub frame.  For this reason, using the GPU to perform these calculations is very tempting. But one of the problems that arise from moving these calculations to the GPU is, triangular mesh topologies don't have a predefined vertex locality, it depends on the arbitrary indexing mechanism used when the topology was created. And even in optimal circumstances,

there are always vertexes which destroy the locality.  At OptiTex, we have been using CUDA since the initial 0.8 Beta. OptiTex specialty is cloth. Using CUDA we are on the way to deliver an implicit cloth engine, which will work in real time for cloth meshes up to 10,000 polygons. In our process we have ported (and are still working) all of our geometry calculations to the GPU. These include normals, smooth skin, collision detection, finite elements and more.  During the development of these parts, different problems occurred relating to vertex index locality. These were solved using a variety of methods and approaches.
Eri Rubin 3D View Streamed Recording Download Video       
1405 Adobe Creative Suite 4 -- Performance Enhancements Featuring GPU Acceleration Plus a sneak peak at the Future of GPU Acceleration!

Many Adobe Creative Suite 4 applications have enhanced features designed to take advantage of nVidia GeForce and Quadro GPU cards to accelerate application performance. In this session, find out how Adobe Photoshop, After Effects and Adobe Premiere Pro get a distinct performance boost when handling large files or creating motion graphics and visual effects. Plus you’ll get a sneak peek at what’s coming in the future around GPU acceleration in Adobe Premiere Pro.
Kevan O'Brien Computational Imaging, Visualization Sorry - this session is not available.        
1406 Embedded Tesla-Using CUDA and Tesla in a Medical Device The TechniScan Whole Breast Ultrasound system images the female breast using a computationally intensive technique known as inverse scattering. Older prototypes used clusters of Pentium M computers to compute the WBU images in about four and one half hours.  By porting the existing Fortran code to C and CUDA, the same algorithm could be run in approximately 16 minutes on 4 first generation Tesla GPUs. I will detail the process porting an MPI-based Fortran algorithm to multiple GPUs and CUDA, including issues and successes, cost, performance, and code design and implementation. Jim Hardwick Computational Imaging, Medical Imaging & Visualization, Visualization View Streamed Recording Download Video       
1407 OpenGL: The Train Has Left the Station (Pre-Conference Tutorial) The train has left the station! OpenGL innovation has greatly accelerated over the past year. Come learn about OpenGL 3 and other new features in NVIDIA’s OpenGL drivers, including how to efficiently mix OpenGL and CUDA or OpenCL in your application.
Michael Gold, Mark Kilgard, Barthold Lichtenbelt Tools View Streamed Recording Download Video Download Video    
1408 DirectX 11 Overview (Pre-Conference Tutorial) This presentation gives an overview of the DirectX 11 pipeline and how it extends previous DirectX versions to enable stunning visual effects in real-time graphics applications.
Cem Cebenoyan Tools View Streamed Recording Download Video Download Video    
1409 OpenCL on the GPU (Pre-Conference Tutorial) This presentation is an introduction to Khronos’ OpenCL programming interface and how it is used to program GPUs. It describes OpenCL’s platform, memory, programming, and execution models and illustrates these concepts with some simple code walkthrough.
Neil Trevett, Cyril Zeller Tools View Streamed Recording Download Video Download Video   Download PDF
1410 C on the GPU (Pre-Conference Tutorial) This presentation teaches the basics of programming GPUs using the C language with CUDA extensions. No prior experience in GPU programming is required. The concepts – data transfers, kernel execution, memory model, synchronization – are introduced progressively and illustrated with step-by-step walkthroughs of code samples.
Timo Stich Tools View Streamed Recording Download Video Download Video    
1411 DirectCompute (Pre-Conference Tutorial) DirectCompute adds the new “Compute Shader” to its API, making possible GPU computing within the familiar programming environment.  Compute shaders provide the programmer with greater flexibility and features such as thread cooperation, shared memory, and unordered data access.  In this talk, we’ll introduce the DirectCompute API by way of building simple examples from the ground up to illustrate the new processing possibilities of DirectCompute.  James Fung Tools View Streamed Recording Download Video Download Video    
1412 Languages, APIs and Development Tools for GPU Computing (Pre-Conference Tutorial) Get a head start on the conference with this first-day introduction to key technologies for GPU Computing.  This tutorial session will cover the key features and differences between the major programming languages, APIs and development tools available today, including case studies of several applications and how they benefit from GPU Computing.  Attendees will also learn several high level design patterns for consumer, professional and HPC applications, with practical programming considerations for each. Will Ramey Tools View Streamed Recording Download Video Download Video   Download PDF
1414 State of the Art in GPU-Accelerated Techniques for Medical Imaging This presentation will provide an overview of world-class medical imaging algorithms exploiting the immense computational power of modern GPUs. The range of techniques where NVIDIA GPUs have proven invaluable range from advanced visualization techniques such as volume rendering of large CT data sets and time-varying volumetric Ultrasound data to GPU-based reconstruction, non-linear filtering, non-rigid registration, segmentation and flow simulation. Siemens Corporate Research conducts research and development for all Siemens Healthcare business units and has a large world-wide network of clinical and academic collaborators. We will present examples of a broad spectrum of clinical applications, both for post-processing as well as real time interventional procedures, which have recently become feasible only thanks to advances in programmable GPUs. Gianluca Paladini Medical Imaging & Visualization View Streamed Recording Download Video Download Video    
1415 Efficient Integration of CUDA & OpenGL Proccessing in XIP This case study presents the results of an effort to implement generic support for CUDA kernels to be executed using the eXtensible Imaging Platform, XIP – an Open Source project funded by the National Cancer Institute. Processing pipelines are constructed using a visual programming paradigm, through graphical representations of pipelines and scenegraphs based on Open Inventor. The developed framework enables efficient integration of CUDA and OpenGL
processing by utilizing memory management exploiting PBO direct transfers between CUDA
and OpenGL. The framework has been used to execute existing CUDA kernels from NVIDIA’s
CUDA SDK as well as a more complex image segmentation algorithm. Performance results and breakdown of the framework will be presented as well as examples using the visual

programming interface.
Patric Ljung Medical Imaging & Visualization View Streamed Recording Download Video       
1417 Using Flash SVG Animations in Enhanced & Virtual Reality This presentation will cover the use of Adobe Flash (SVG) animations in virtual “worlds”. Today’s consumer products have dynamic and interactive multimedia user interfaces. Many businesses would benefit by employing these same sophisticated user interfaces in their traditional graphics-based applications. The integration of Flash animations into these primarily process driven applications increases their appeal, usability and functionality.
During the presentation, we will demonstrate the use of Animated Media’s VGK product platform with an NVIDIA GPU in an industrial control solution that was developed for a client in the fire and safety business. The client’s fire and safety control system was being upgraded to use current technologies and required 2D/3D visualization of real-time activities. In the solution, we virtualized and used multiple VGK engines to render the Flash animations to texture-mapped 3D accelerated OpenGL objects in the floor plans of multi-story buildings. The animations represent various fire system objects (sensor components) in the building such as fire control valves, sprinklers, smoke detectors, speakers, elevators, escape routes, and fire detectors. The building’s fire and safety system, accessed through the building’s fire panel, communicates in real-time using XML over a network. The application monitors the status of every fire system object in the building and commands the ActionScript to display the proper animation to match the object’s status. Integrating standard Adobe Flash animations into a 2D or 3D software application adds attention getting visuals, such as a pulsating speakers and smoke detectors, enabling the building staff to readily identify troubled areas within the building and direct firemen or safety resources to that exact location.
Chris Brady, Bob Starr Computational Imaging View Streamed Recording Download Video Download Video   Download PDF
1418 Heterogeneous Multicore Parallel Programming Modifying legacy codes to take advantage of GPU acceleration can be a time consuming task, and portability concerns must be carefully considered at the same time.  In this presentation we give an overview
of HMPP (Heterogeneous Multicore Parallel Programming) workbench that addresses this challenge by providing:
 * An incremental use of NVIDIA GPUs through the insertion of directives into C & Fortan source code, from which the CUDA code is automatically generated
 * Performance tuning as well as CPU-GPU data transfers optimizations directives
We will also share some experiences from HMPP customers porting their key applications to NVIDIA Tesla systems.
François Bodin High Performance Computing, Tools View Streamed Recording Download Video Download Video   Download PDF
1422 Day 2 Keynote with Hanspeter Pfister, Harvard University High-Throughput Science

How did the universe start? How is the brain wired? How does matter interact at the quantum level? These are some of the great scientific challenges of our times, and answering them requires bigger scientific instruments, increasingly precise imaging equipment and ever-more complex computer simulations. In his keynote address, Harvard professor, researcher and computing visionary Hanspeter Pfister will discuss the computational obstacles scientists face and how commodity high-throughput computing can enable high-throughput science, in which massive data streams are processed and analyzed rapidly -- from the instrument through to the desktop. Finally Professor Pfister will survey several groundbreaking projects at Harvard that leverage GPUs for high- throughput science, ranging from radio astronomy and neuroscience to quantum chemistry and physics.
Hanspeter Pfister General Interest View Streamed Recording Download Video (Small - 271MB mp4 format) Download Video (Medium - 1G mp4 format) Download Video(Large - 1.6G mp4 format) Download PDF
1423 Day 3 Keynote with Richard Kerris, Lucasfilm Games and interactive media have long been the beneficiaries of cutting edge GPU technology and it has not gone unnoticed in the world of feature film production. To date the visual effects industry had been a sideline observer of these advances while awaiting technology to reach maturity. At Lucasfilm, research and development has been on-going for some time and this past summer Industrial Light & Magic employed this technology in two of its summer blockbuster films. Lucasfilm CTO, Richard Kerris, will show a brief history of their computer graphics for film, and will then pull back the curtain on how they are now using GPU technology to advance the state of  the art in visual effects and provide a glimpse of what’s on the horizon for GPU’s in future and how it will impact filmmaking.

Richard Kerris General Interest View Streamed Recording Download Video (Small - 205MB mp4 format) Download Video (Medium - 492MB mp4 format) Download Video (Large - 1.2G mp4 format)  
1425 Future Directions in GPU Computing Join our panel of leading industry experts as they discuss some of the most exciting and disruptive current and future applications to be built on top  of GPUs.  Topics will cover both visual and high performance computing, including rendering, video processing, scientific computing, cloud computing, and more.   Industry analyst Rob Enderle will moderate this marquis panel, which will include Steve Perlman, Founder and CEO- ONLIVE; Simon Hayhurst, Senior Director of Product Management – Adobe; Sean Varah CEO- MotionDSP; and Bill Dally, NVIDIA’s Chief Scientist. Bill Dally, Simon Hayhurst, Steve Perlman, Sean Varah General Interest   Download Video      
1426 Fireside Chat with Jen-Hsun Huang and Jon Peddie For the second year in a row, NVIDIA CEO Jen-Hsun Huang will participate in a freewheeling fireside chat with leading industry analyst Jon Peddie, of Jon Peddie Research.  Topics will include the state of the industry, NVIDIA’s strategy and the future of GPU computing, as well as other issues raised by the audience. Jen-Hsun Huang General Interest Sorry - this session is not available.        
1428 Raising Capital in Difficult Environments The current economic situation poses unique challenges for start-ups seeking to raise critical funds to grow their businesses.  This panel will attempt to assess the current state of both the public and private markets, and will explore various strategies and options for raising capital.  Topics will include equity and debt, as well as other creative forms of financing such as NRE arrangements, etc.  Jeff Herbst, NVIDIA’s Vice President of Business Development will lead this panel discussion, which will include leading industry experts from Sutter Hill Ventures, Citi, Silicon Valley Bank and Deloitte. Heather Gates Massoudi. Jeff Herbst, Peter Kidder. Andrew Sheehan General Interest Sorry - this session is not available.        
1431 GPU Acceleration of Mortgage CMO Option Adjusted Spread Mortgage Backed Security Option Adjusted Spread (OAS) analysis requires computationally intensive Monte Carlo simulation of dynamic cash flows at different interest rate scenarios.  This talk will describe a port of the Bloomberg L.P. MBS OAS legacy code to CUDA, and will compare the performance and cost of running the calculations on a CPU cluster versus a CUDA-enabled GPU cluster.  Significant speedup and power savings are achieved using a CUDA enabled OAS server.   Additionally, the development and rollout process of a CUDA-based application to production will be presented. James Hook, Xusheng Tian Computational Finance View Streamed Recording Download Video Download Video    
1433 Computational Biophysics and Long Range Electrostatics on GPUs This talk will present detailed algorithmic approaches to approximating long-range electrostatics on GPUs and optimizing performance for molecular dynamics, the simulation of molecules.  GPU implementations of these algorithms differ significantly from CPU implementations both in terms of the relative costs of computation versus memory access and in maximizing data parallelization.  To this end, we will describe the Generalized Born method as it was implemented both for Folding@Home and for AMBER, as well as the implementation of the Particle Mesh Ewald (PME) method as implemented for AMBER, focusing on their implementations, and describing the tricks and tradeoffs required to achieve both accuracy and performance.  Finally, we will describe how to extend these algorithms to multiple GPUs. Scott Le Grand Life Sciences, Molecular Dynamics View Streamed Recording Download Video Download Video    
1434 Unlocking Biologically-Inspired Computer Vision: a High-Throughput Approach The study of biological vision and the creation of artificial vision systems are naturally intertwined – exploration of the neuronal substrates of visual processing provides clues and inspiration for artificial systems, and artificial systems, in turn, serve as important generators of new ideas and working hypotheses.  However, while systems neuroscience has provided inspiration for some of the "broad-stroke" properties of the visual system, much is still unknown. Even for those qualitative properties that most biologically-inspired  models share, experimental data currently provide little constraint on their key parameters.  Consequently, it is difficult to truly evaluate a set of computational ideas, since the performance of a model depends strongly on its particular instantiation – the size of the pooling kernels, the number of units per layer, exponents in normalization operations, etc.

To pave a way forward, we have developed a high-throughput approach to more expansively explore the possible range of biologically-inspired models, including models of larger, more realistic scale, leveraging recent advances in commodity stream processing hardware - particularly, high-end NVIDIA GPUs.  In analogy to high-throughput screening approaches in molecular biology and genetics, we generated and trained thousands of potential network architectures and parameter instantiations, and "screened" the visual representations produced by these models using an object recognition task.  From these candidate models, the most promising were selected for further analysis. We have shown that this approach can yield significant, reproducible gains in performance across an array of basic object recognition tasks, consistently outperforming a variety of state-of-the-art purpose-built vision systems from the literature, and that it can offer insight into which computational ideas are most important for achieving this performance.

In this talk, I'll also highlight how the application of flexible programming tools, such as high-level scripting and template metaprogramming, can enable large performance gains, while managing complexity for the developer.  As the scale of available computational power continues to expand, our approach holds great potential both for accelerating progress in artificial vision, and for generating new, experimentally-testable hypotheses for the study of biological vision.
Nicolas Pinto Computer Vision View Streamed Recording Download Video      Download PDF
1437 Directing Experiments in the International Space Station With GPU-Assisted Image Analysis We implement image correlation, a fundamental component of many real-time imaging and tracking systems, on a graphics processing unit (GPU) using NVIDIAs CUDA. We use our code to analyze images of liquid-gas phase separation in a model colloid-polymer system, photographed in the absence of gravity aboard the International Space Station (ISS). Our GPU code is 4000 times faster than simple MATLAB code performing the same calculation on a central processing unit (CPU), 130 times faster than simple C code, and 30 times faster than optimized C++ code using single-instruction, multiple data (SIMD) extensions.

The speed increases from these parallel algorithms enable us to analyze images downlinked from the ISS in a rapid fashion and send feedback to astronauts on orbit while the experiments are still being run.
Peter Lu Astro GPU View Streamed Recording Download Video       
1440 Computational Based Modelling and Simulation  The impact of computational based Modelling and Simulation in providing industry with a competitive edge delivering higher performance and cost effective products is well documented and cannot be under estimated. This has, in the past been enabled by the year on year floating point performance improvements delivered by the processor and system manufacturers and has been the primary end user experience of Moore’s law. The development of MultiCore technologies in all its wide variety of forms challenges this end user effect of Moore’s law requiring a radical rethink of algorithms and their implementations just to maintain the same progress seen over the previous decades. The exploitation of GPU Computing promises to deliver a radical and affordable step change in simulation capability and this will be illustrated with details of a Computational Fluid Dynamics (CFD) solver (3D Finite Volume Unstructured) that demonstrates 20x performance benefits over the latest Intel Nehalem cores. Jamil Appa Computational Fluid Dynamics  View Streamed Recording Download Video       
1441 Binary Black Holes using GPUs We perform ensemble studies of binary black hole inspirals. The binary black hole problem is of great interest for the cosmological community (merger of galaxies with BHs at the center) as well as the gravitational wave community (where the merger of BHs is the most important signal source). The full binary black hole merger problem is computationally very demanding and even with advanced numerical techniques ensemble studies are currently not possible. Using a standard approximate solution to Einstein's equation (the post-Newtonian equations) one can accurately model the inspiral until shortly before merger when the approximation techniques break down. Utilizing this approximation technique we study the 7-dimensional parameter space of the BH merger problem using a Monte-Carlo approach, which extends very naturally to GPUs.
Frank Herrmann Astro GPU View Streamed Recording Download Video       
1442 Numerical Cosmology Powered by GPUs By definition, cosmology cannot rely on lab experiments to reproduce the phenomenons observed in the sky and test its theories. For this very reason, the use of numerical simulations is widely spread within this community in order to understand the formation of the astrophysical objects and to put constrains on the physical ingredients that lead to the Universe as it is currently observed. Since 2001, I have been personnaly involved in trying to understand these questions through the intensive use of numerical simulations that reproduce the evolution of the Universe from the Big-Bang to our epoch. During the last two years, I have been investigating the new possibilities offered by GPUs to boost these numerical calculations, mostly using CUDA. At the current stage, three applications benefited from these studies and using 8800 GTX and Tesla C1060 devices, we found that accelerations range from factors 20 to 80 compared to CPU versions : - a cosmological N-Body integrator CUDAPM. It follows the evolution of millions of particles that interact through gravitation in an expanding Universe, modelling the rise of large scale structures. - a non-linear full multigrid solver, for the Poisson equation of modified Newtonian gravity (CUDAMOND). - a cosmological radiative transfer code CUDATON. It models the propagation of ionising radiation and its effect on the gas that filled the Early Universe. This application is multi-gpu and currently runs on 192 devices on the CCRT supercomputing centre.

Most of the techniques used in these applications are fairly standard and are not specific to astrophysics and cosmology. Therefore describing my own experience of porting these applications to GPUs as a physicist is likely to benefit to a large public of numerical scientists.

Dominique Aubert Astro GPU View Streamed Recording Download Video      
1443 Thrust: A Parallel Template Library for CUDA Thrust is a parallel template library for developing CUDA applications.  Modeled after the C++ Standard Template Library (STL), Thrust brings a familiar abstraction layer to the realm of GPU computing.  Thrust provides host and device variants of the STL vector container to simplify memory management and facilitate data transfers.  These containers are complemented with a large collection of generic data-parallel algorithms and a suite of useful iterator adaptors. Together, these features form a flexible high-level interface for GPU programming that greatly enhances developer productivity.

In this tutorial we'll walk through Thrust's features and demonstrate how Thrust is used to develop larger-scale libraries and applications. We'll also discuss how Thrust enables "kernel fusion" and circumvents the "array of structs" vs. "structure of arrays" problem.  Lastly, we'll present evidence that Thrust implementations are fast, while remaining concise and readable.
Nathan Bell Tools View Streamed Recording Download Video Download Video    
1444 GPU-based Decompression for Medical Imaging Applications Medical imaging applications such as computed tomography (CT), ultrasound, and magnetic resonance imaging (MRI) capture an exponentially increasing amount of real-world sensor data.  Real-time compression of this sensor data reduces bandwidth and storage costs of medical imaging equipment by 5% to 25%.  This paper illustrates how real-time compression in field-programmable gate arrays (FPGAs) is being combined with GPU-based decompression and image reconstruction to enable new high-performance, low-cost imaging systems.  We compare GPU-based decompression with CPU-based compression and identify ways that future GPUs could further reduce the decompression overhead during image reconstruction.  More generally, transferring compressed packets across the GPU-CPU interface solves signal processing I/O problems in CUDA-based compute applications. Albert Wegener Medical Imaging & Visualization, Visualization View Streamed Recording Download Video     Download PDF
1445 GPU-Accelerated Imaging on Mobile X-ray C-arm  Digital X-ray on a mobile C-arm device is an imaging tool for assisting the surgeon during image-guided interventions. The current generation of C-arm systems comprises various imaging modes: traditional, real-time high-resolution fluoroscopy, and two new, near-time C-arm CT and tomosynthesis. To tune fluoroscopic IP chain, improve effectiveness of observer studies, and support research efforts, we discuss HW and SW implementation and review the early results associated with GE Healthcare-Surgery Interactive Imaging Chain Simulator. The Simulator has image processing functionality close to one that can be observed on industrial C-arms equipment. Using numerical X-ray CT Exerciser and experimental data from the C-arm prototype capable of rotational scan, we model cone-beam CT performance on a mobile C-arm, benchmark GPU-accelerated image reconstruction, and demonstrate the imaging results. Our methodology is to create a friendly scalable framework utilizing a common workstation and running surgical applications on a cost-effective heterogeneous computing and visualization platform.   Arvi Cheryauka Medical Imaging & Visualization, Visualization Sorry - this session is not available.        
1446 Emerging Companies Summit Opening Address The Emerging Companies Summit is a unique forum for startup companies to showcase innovative applications that leverage the GPU to solve visual and high-performance computing
problems. The Opening Address includes selected demos from a few presenting companies. The ECS is a great opportunity to discover new players in the visual and high performance
computing ecosystems, find great investments, explore partnership opportunities, network/build relationships, and discuss the future of an industry that is reshaping computing.
Jeff Herbst General Interest Sorry - this session is not available.        
1447 The GPU Revolution in Film Production: Report from the Battlefield Come join our panel of renowned industry experts working at major US film studios including Disney Feature Animation, Sony Pictures Imageworks and Digital Domain as they discuss and present their latest work from feature films and how the GPU is changing the way they make films by solving complex technical challenges and enhancing production pipeline worldwide. Rob Bredrow, Arthur Shek, Chris Horvath Film Sorry - this session is not available.        
1449 Accelerating Quantum Chemistry Research Using GPUs-Two Electron Repulsion Integrals in GAMESS We present an implementation of Rys quadrature algorithms for Electron Repulsion Integrals (ERI) of up to g functions on Graphical Processing Units (GPUs). We outline the general GPU programming model, challenges associated with implementing the Rys quadrature on highly parallel emerging architectures, and our approach to implementing the quadrature. The performance of the implementation is evaluated for single and double precision on two GPU devices. The performance obtained is on par with the matrix-vector routine from the CUDA Basic Linear Algebra Subroutines (CUBLAS) library.  Veerendra Allada Algorithms & Numerical Techniques View Streamed Recording Download Video  Download Video   Download PDF
1451 iray®: Interactive, Zero Tuning, Consistent Rendering in mental ray, RealityServer and other Applications iray® is a interactive, global illumination ray-traced rendering technology that generates photo real imagery without introducing rendering algorithm specific artifacts, and without requiring the use of renderer specific parameterizations. This “push-button” rendering capability can dramatically improve artist productivity in a wide variety of use cases, including architectural visualization, automotive styling and marketing, product styling, among others.

iray takes full advantage of the CUDA programming model, allowing interactive performance on single NVIDIA GPUs, and scaling almost linearly on multi-GPU platforms. The iray rendering mode will be available in the next versions of mental ray (3.8) and RealityServer (2.4).
We will demonstrate the key features of iray for end users and application developers, including the progressive refinement of images until maximum fine detail is reached, providing a single process which smoothly combines interactive pre-visualization and final frame rendering.
Michael Kaplan General Interest View Streamed Recording Download Video      
1452 Visual Innovation in the Classroom 3D models are transforming primary education. Research has proven that only 35% of students can think in 3D, so visualization has become an important tool in the classroom for improving engagement, comprehension and boosting test scores. Sean O’Brien, president of NavTech, will discuss the innovative technologies being leveraged in K – 12 classrooms, and provide a development overview of the technology used to drive these applications. Sean will share some of the models that secondary and college educators have built to improve learning in science, technology, engineering and math (STEM). Attendees will don 3D glasses and be given a demonstration of scientific concepts being taught in today’s classrooms.  Sean OBrien General Interest View Streamed Recording Download Video Download Video    
1454 GPU Ventures, for the Entrepreneur During this panel, we will discuss how Viewdle and NVIDIA have been working together and bring a product to market.  Viewdle first came to NVIDIA with a technology that required massive amounts of raw compute.  Working with NVIDIA's technical team, Viewdle was able to harness the power of the GPU, and bring advanced photo and video analysis to the consumer desktop.  NVIDIA not only provided the technical expertise, but also the business and marketing support.  Working together with NVIDIA, Viewdle was able to formulate a consumer strategy and secure distribution partners.  This discussion will be from the point of view of the entrepreneur who would like to work with NVIDIA and will include real examples and demos of Viewdle's soon to launch software. Laurent Gil, Jeff Herbst General Interest View Streamed Recording Download Video Download Video    
1455 Median Filtering: A Case Study in CUDA Optimization In this tutorial we present a new approach to integrating CUDA code into MATLAB leveraging Jacket’s Developer SDK.

Jacket’s Developer SDK makes integration of custom CUDA code into Jacket’s runtime very easy.  With a few simple jkt functions (which mimic standard MEX API functions), you can integrate custom CUDA kernels into Jacket.  This enables your CUDA code to inherit Jacket’s optimized memory management and kernel execution runtime system.

In this tutorial, we share examples of using the Developer SDK to write a median filter in CUDA and integrate it into Jacket.  We will start with the naďve approach, then show how to optimize by using shared memory, and finally show the impact of using texture memory on this problem.  In each case, we will integrate the code into Jacket & MATLAB via the Developer SDK and do benchmarking in MATLAB.

A Q/A session will enable audience members to ask specific questions about the Developer SDK, CUDA programming, and MATLAB-based GPU computing.
James Malcolm Tools   Download Video Download Video    
1456 Accelerating Finite Difference Time Domain (FDTD) Simulation with CUDA Advances in a variety of fields ranging from cellular phone design to MRI coil design necessitate the use of high accuracy electromagnetic computational methods such as Finite Difference Time Domain (FDTD) to analyze complex new products. High fidelity techniques such as FDTD generally yield highly accurate results at the cost of increased computation time. Using the processing power of NVIDIA’s CUDA architecture allows this time to be reduced by over two orders of magnitude.

This session presents an implementation of the FDTD method using NVIDIA’s CUDA technology. The core algorithm is presented in C code and translated to a CUDA kernel. Branching statements are eliminated in order to improve parallelism. Shared memory and an optimized looping method are employed to avoid memory communication bottlenecks. Timing comparisons against an optimized commercial FDTD code are given.

James Stack General Interest View Streamed Recording Download Video Download Video    
1457 Examples of CUDA-based Realizations in Artificial Vision and Soft Computing We present CUDA-based realizations of computationally expensive algorithms in artificial vision as well as soft computing, running at ultra-fast rates. One example addresses the problem of agglomerative clustering, exploiting parallel realizations to achieve very high performance gains. Another example highlights a new, biologically-inspired approach to the implementation of deep artificial neural networks in CUDA.  To illustrate this approach, visualizations of deeper layer receptive fields will be presented, as well as a discussion regarding future implementations to improve both network accuracy, as well as its overall information capacity. Network training in CUDA is also presented. Benchmarking of these realizations against those with OpenCV/Intel platforms will also be done, as well as highlighting both GPU occupancy and performance with the Nexus profiler.  Tarek El Dokor Computer Vision View Streamed Recording Download Video Download Video    
1458 Raster Processing Using Multiple GPUs and CPUs With the ever-increasing availability of satellite data streams, affordable storage space and communication bandwidth, there is a strong demand for processing increasingly large raster data sets in GIS and Remote Sensing applications. Although available computing power also increases day to day, these increases come mostly in the form of additional processing units.

Utilizing these additional units effectively requires significant modifications to data processing algorithms. This talk describes an approach to creating a generalized pipeline for processing raster data that scales with the number of GPUs as well as the number of CPUs, and
is suitable for implementing most of the classic raster algorithms.
Dimitri Rotow Computational Imaging, Visualization View Streamed Recording        
1459 Multi-Frame Analysis: The Future of Video Processing  MotionDSP makes software that significantly improves video from a wide range of sources -- from mobile phones to surveillance cameras. Its patented, multi-frame video enhancement technology is available in Ikena, a video forensic solution, and in vReveal, a Windows application for consumers.

High-fidelity motion estimation plays a critical role in almost any video processing technology. The open-loop nature of MotionDSP's multi-frame video reconstruction framework requires a departure from quantized block-based motion estimation concept, commonly used in video encoding, as super-resolution temporal filtering can not rely on a prediction error to compensate for motion estimation inaccuracies. The need to use floating-point accurate motion fields, coupled with a multi-frame nature of the algorithm, makes GPGPU a natural choice when addressing this massive computational problem.

This talk will provide an overview of MotionDSP's video technology and products, as well as details of the design and performance challenges of MotionDSP's multi-frame video pipeline, one of the first commercial video processing solutions fully ported to GPU architecture.
Nikola Bozinovic Computational Imaging View Streamed Recording Download Video Download Video    
1460 Supercomputing Research Roundtable This moderated Research Roundtable will follow up Thursday’s “GPU Supercomputing Supersession” with a discussion among the presenters and audience members about the issues, challenges, and opportunities in GPU-powered “big iron” supercomputing.  The roundtable will bring together researchers that have recently deployed or are interested in deploying GPU clusters to enable scientific research. Attendees and presenters will share their experiences in deploying GPU technology and discuss the future of this technology in supporting research to tackle the world's most challenging scientific problems. Topics will include
•what it takes to build a GPU cluster
•current issues affecting GPU cluster computing
•how to best utilize a GPU cluster for scientific research
•software tools for scientific computing on GPU clusters
•integrating compute, data management and visualisation on GPU clusters
John Taylor High Performance Computing View Streamed Recording Download Video      
1461 A Large Scale Simulation of Lattice QCD with a GPU Cluster Quantum Chromodynamics (QCD) is the quantum field theory of the strong interaction, describing the interactions of the quarks and gluons making up hadrons (e.g., proton, neutron, and pion). Most importantly, it accounts for the nuclear energy inside an atom, as well as plays an important role in the evolution of the early universe. To solve QCD is a grand challenge among all sciences. Now the most promising approach to solve QCD nonperturbatively is to discretize the continuum space-time into a 4 dimensional lattice (i.e., lattice QCD), and to compute physical observables with Monte Carlo simulation. For lattice QCD with exact chiral symmetry, it often requires supercomputers (e.g., 10 racks of IBM BlueGene) to perform the simulations. The TWQCD Collaboration in Taiwan is the first lattice QCD group around the world to use a GPU cluster (with 120 GPUs) to perform large-scale unquenched simulations of lattice QCD with the optimal domain-wall fermions, attaining 14 Teraflops (sustained) at a price of $200,000. This has significant impacts to the lattice QCD, as well as the physics of the strong interaction.
Ting-Wai Chiu General Interest, High Performance Computing View Streamed Recording Download Video Download Video    
2001 Visioglobe Startup Presentation 3D real time Visualization Engine available on
-Mobile Device
To provide a real 3D navigation and local search, enhanced with real time information coming from
-Social networking
-Local based Advertising
Eric Bernard 3D, Embedded & Mobile, Visualization View Streamed Recording Download Video      
2002 Viewdle Inc. Startup Presentation By unleashing Viewdle technology, content producers and distributors can effectively leverage their video assets to open new revenue streams and attract, engage and retain customers/visitors. Initially focused on providing internal and private-label indexing and search capabilities to news and entertainment media, we will soon launch targeted advertising solutions, developer tools and a consumer-facing search site. Viewdle was founded in 2006 to revolutionize video search by solving the inefficiencies inherent in using text-based metadata and other existing tools for indexing moving media. The Viewdle engine fuses pioneering visual analysis technology, continuously developed over the past 35 years, along with other leading search techniques to go "inside the clip"and identify contextually relevant appearances based on actual video assets. Laurent Gil Computational Imaging, Visualization View Streamed Recording Download Video Download Video    
2003 VERTICE Startup Presentation Founded in 2002, Vertice is located in Toulouse. Vertice is a Young Innovating Company (JEI) which was elected in 2003 1st Innovating Company. Vertice develops and distributes the Nova sotware range dedicated to 3D real time and accompanies its customers in the customization of applications. Vertice is a sister company of Bewise, one of the rare French software firms exclusively specialized on Microsoft development technologies. David Catuhe 3D View Streamed Recording Download Video Download Video    
2004 TechniScan, Inc. Startup Presentation The Company's Whole Breast Ultrasound (WBU) system produces three separate images; two images from sound transmission properties (one for the speed of sound and one for the attenuation of sound) and a more standard reflection image, is generated from data collected from three reflection transducers with different focal depths.  All three images are produced as a series of planar slices of the breast that are reconstructed into a coherent three-dimensional CT image. Images produced by the WBU system possess important advantages over traditional ultrasound methods including: standardized, high-quality images that are generated independent of operator skill;
-images that are accurately registered in 3-D; -images that are of high spatial resolution and low noise that allow for accurate localization of lesions and calculation of mass size; and

- images that show either coronal slice of the total breast or a complete 3-D representation of the section imaged, rather than a small "flashlight view" of the breast as shown by traditional ultrasound.  Importantly, the WBU system offers a painless and efficient method, without ionizing radiation, to screen and diagnose the breast and provide unique information about bulk tissue properties that may aid in the effective diagnosis of breast cancer.  
David Robinson Medical Imaging & Visualization View Streamed Recording Download Video Download Video    
2005 Stormfjord Startup Presentation Stormfjord is a leader in interactive visualization of very large scale industrial data, specialising in Gpu-centric simulation. We provide next generation tools to drive novel industrial use-cases enhancing complex planning, work process modeling, robotic prototyping, training, control, and monitoring. Our tools are best in their class, derived from the bleeding edge of technology, and open to 3rd party developers. Thorolf Tonjum Visualization View Streamed Recording Download Video Download Video    
2006 SOFTKINETIC S.A. Startup Presentation Softkinetic develops 3D gesture recognition middleware to build natural interfaces for the interactive digital entertainment and consumer markets. Michel Tombroff 3D View Streamed Recording Download Video Download Video    
2007 Realtime Technology AG Startup Presentation RTTprovides 3D real-time visualization technology and services for innovative workflows in the auto, aeronautical and consumer-goods industries, as well as in the fields of interior design and architecture.  It supports customers in coming up with new ways how to design and to market their products.  The company supports its customers not only during the design and developing phases with regard to the virtual display of different product concepts (Virtual Prototyping), but also during the marketing and sales stages in presenting the products in all variants and with all additional options in the Showroom, on the Internet, at fairs or at dealerships (Virtual Marketing).  Ludwig A. Fuchs 3D, Visualization View Streamed Recording Download Video Download Video    
2008 Right Hemisphere Startup Presentation Right Hemisphere is the leading provider of enterprise visual product communication, collaboration and publishing solutions that optimize mission critical business processes such as sourcing, manufacturing, marketing and customer support for leading Global 1000 manufacturers. Right Hemisphere integrates disparate product data readily available in CAD, PLM, and ERP systems and automates the delivery of intelligent product information as common documents and files or directly into business applications to downstream stakeholders on a global basis.  Founded in 1997, Right Hemisphere is a privately held, venture-funded corporation based in Silicon Valley and Auckland, New Zealand. For more information please visit Michael Lynch General Interest, Visualization View Streamed Recording Download Video Download Video    
2009 OptiTex Startup Presentation OptiTex is the leading developer of 2D & 3D CAD solutions for virtually all sewn-products industries.  OptiTex technologies allows designers to create, correct and adjust compelling designs before the first piece of fabric is cut, giving a new dimension to the motto, "Virtual is Real" and proving what OptiTex has long maintained - that corporate responsibility and an environmentally friendly approach are not mutually exclusive but in fact are intimately entwined. OptiTex presents a complete content creation solution, to create and visualize customized garments, to simulate fitting and draping of garments on fully parametric virtual models and to create movie clips in an immediate and direct manner. OptiTex system consists of three main components: cloth content creation system with our PDS software, 3D Runway Designer, a virtual try-on-system, which includes both cloth simulation and accurate 3D parametric mannequins; motion animation engine which enables the generation of motion sequences with interactive cloth.  OptiTex brings a wealth of virtual textile experience to the gaming, feature animation and digital effects industries. OptiTex's products are second only to real life in depicting fabric movement and dynamics.  The combination of OptiTex's rich graphics and data intensive computation and NVIDIA's CUDA-enabled GPUs provides a high-performance environment for animators and digital artists via the integration of OptiTex cloth engine technology. Eri Rubin 3D View Streamed Recording Download Video Download Video    
2010 Nurien Software Startup Presentation Nurien Software, established in 2004, is a leading developer of next-generation 3D social networking platform and gaming services geared to deliver unprecedented experience to the end user. The Nurien platform includes user-centric social networking features a fully customizable 3D avatar system, 3D objects and user-created content support as well as game and other social applications including dance, fashion and quiz game applications. Users can create one-of-a-kind avatars to their exact liking, and use avatars to participate in games, form communities, and interact with other players. With stunning visuals and innovative game and entertainment functions, the Company aims to revolutionize next-generation 3D social network platform market.  Taehoon Kim 3D View Streamed Recording Download Video Download Video    
2011 Metaio, Inc. Startup Presentation metaio designs, develops and markets solutions based on augmented reality. This innovative technology allows virtual 3D information to be superimposed seamlessly into the real environment at real-time. metaio was founded in 2003 and is fully owned by CEO Dr. Thomas Alt and CTO Peter Meier. Thomas Alt 3D View Streamed Recording Download Video Download Video    
2012 Innovative Converged Devices, Ltd Startup Presentation Innovative Converged Devices Ltd (ICD) is a small, highly efficient company that designs consumer electronic products such as mobile phones, personal computers and new types of Internet devices. ICD achieves differentiation from the very large Far Eastern companies by creating products that are optimised for purpose with aesthetic design that appeals to Western tastes combined with the latest hardware & software technology available.

ICD's business model is to conceive, design, prototype and prepare products for manufacture. ICD does not actually perform the manufacturing but arranges introductions between our customers and high-grade contract manufacturers who form their own commercial relationship. ICD facilitates this arrangement but does not participate in the financial transactions of manufacture. ICD describe this model as a "Fabless Original Design Manufacturer" or Fabless ODM.  
David Hayes Computer Vision, Embedded & Mobile View Streamed Recording Download Video Download Video    
2013 Forterra Systems, Inc Startup Presentation Forterra provides enterprise grade virtual world software that enables our customers and partners to deploy their own secure, scalable 3D Internet solutions for the corporate, healthcare, government, and education industries. Our software and services enable organizations to collaborate, train, practice, and plan in compelling 3-D virtual worlds that integrate with existing media and business systems. Since 1998, the company has used its unique combination of experience in enterprise software, social networks, video game production, modeling and simulation, real-time graphics, and artificial intelligence to deliver innovative solutions around collaboration, training, and operations applications.

Forterra's flagship product OLIVE (On-Line Interactive Virtual Environment) is a powerful client-server based virtual world platform that connects users in real-time over any network. Within the virtual world, users are represented by realistic 3D avatars that create natural human interaction and discussion. Independent of scale, solutions built on the OLIVE platform can be integrated with existing business systems and media, 3-D content, and data sets to create powerful collaborative environments. OLIVE's architecture is mature and proven, which is an important consideration for customers and partners making long term investment decisions. Forterra is a trusted business partner aligned with key industry leaders like IBM, Accenture, Nvidia, and Autodesk.
David Rolston, PhD 3D, Visualization View Streamed Recording Download Video Download Video    
2014 Elemental Technologies, Inc. Startup Presentation Elemental Technologies is the leading provider of massively parallel processing solutions for video applications. Using off-the-shelf, programmable graphics processing units (GPUs) for compute-intensive video processing and conversion tasks, Elemental's products are ideally suited for a variety of video applications including consumer media processing, professional video editing and Internet video processing. Founded in 2006, Elemental is headquartered in Portland, Oregon.  Sam Blackman Computer Vision, General Interest View Streamed Recording Download Video Download Video    
2015 Edge 3 Technologies Startup Presentation Edge 3 Technologies is an offshoot of the Embry-Riddle Machine Vision Lab, the premier research facility in 3D gesture recognition with GPU technology. We specialize in the development of GPU middleware with advanced learning, coupled with machine vision for sophisticated Fortune 500 companies. Our technology is up to 100x faster and more accurate than conventional CPU-based technology. The company's goal is to help redefine the relationship between humans and machines. Our methodology involves:
1) CUDA/openCL/GPU to mitigate computationally challenging aspects and improve real-time object segmentation and tracking performance, and reduce error rates
2) CUDA/GPU for machine learning/training on massive amounts of data to improve classification and identification of various user gestures and postures
3) Machine learning of users' behavior and habits with hundreds of features
4) A new AI capable of identifying facial gestures as well as hand gestures, within the same framework
5) Differentiating users with various biometrics and depths
6) Usability testing on thousands of subjects
7) A portfolio of patent applications in the field
Tarek El Dokor 3D, Computer Vision View Streamed Recording Download Video Download Video    
2016 Cooliris Startup Presentation Cooliris, Inc. was founded in January 2006 with a simple mantra: "Think beyond the browser". We focus on creating products that make discovering and enjoying the Web more exciting, efficient, and personal.

Our core products include Cooliris (formerly PicLens), which transforms your browser into an interactive, full-screen "cinematic" experience for web media, and CoolPreviews, which lets you preview links instantly.  Cooliris has reached 12 million installs of the product, with tens of thousands downloading everyday. The latest version is available for the PC, Mac, and Linux platforms, and can be downloaded at Headquartered in Palo Alto, California, the Cooliris team consists of seasoned executives, engineers and entrepreneurs from companies like Apple, eBay, Electronic Arts, Google, Microsoft, PayPal and YouTube. For more information, please visit
Austin Shoemaker General Interest View Streamed Recording Download Video      
2017 Canesta Startup Presentation Canesta is the inventor of revolutionary, low-cost electronic perception technology that is the foundation for the “other side of 3-D” – true 3-D perception as input to everyday devices, rather than the widely-understood 3-D representational technologies as output. Canesta’s 3-D input technology, based upon tiny, CMOS 3-D imaging chips or “sensors”, enables fine-grained, 3-dimensional depth-perception in virtually any kind of consumer device such as PC’s, TVs, game consoles, and mobile phones, as well as automotive, industrial, and other products. Such products can then react on sight to the actions or motions of individuals and objects in their field of view, gaining levels of functionality and ease of use that were simply not possible in an era when such devices were blind.
Numerous applications are under active development by Canesta’s OEM customers and partners, including consumer electronics, PC, TV, building automation, security, robotics, automotive, and others. Canesta is located in Sunnyvale, CA. The company has filed in excess of fifty patents, 40 of which have been granted so far. All trademarks and registered trademarks are those of their respective companies. The omission of a trademark or registered trademark symbol is inadvertent and is not intended to convey the status of any mark or contravene any claim.
Jim Spare 3D, General Interest, Visualization View Streamed Recording Download Video Download Video    
2018 C3 Technologies Startup Presentation World's most realistic 3D city models

Swedish C3 Technologies fully automatically creates huge 3D city models from images only. The flagship product C3 Maps enables online mapping providers a much richer map display and the possibility for the user to fly around in 3D to find and explore locations. While current map solutions offer 2D maps and ortho photos C3 offers unique realistic high resolution 3D imagery where everyone can dive down to see their own back yard in 3D. The possibility of free navigation (fly overs) is today available for the web and desktop solutions and high-end mobile units and navigation devices.Using a highly efficient algorithm from the Swedish defence industry C3 computes a textured elevation model from thousands of images in a few hours time. A 3D model can be delivered within weeks from the photography, containing whole large cities, not only city centers. With tremendous amounts of 3D data streamed over the Internet, C3 carefully  benefits from the full ability of 3D hardware, challenging old 3D city model concepts with more elaborated level of detail technology. The algorithms are currently used for aerial and satellite imagery and during fall of 2009 C3 plans a release of unique street level products using similar technology.  

C3 and its products has been noticed by slashdot and furthermore the models are claimed by the press to beat Google Earth and Microsoft Virtual Earth content. 
Mattias Astrom 3D, Embedded & Mobile View Streamed Recording Download Video Download Video    
2019 Aqumin LLC Startup Presentation Aqumin, founded in 2008, builds financial market interpretation software called AlphaVision.  Using 3D visualization, sophisiticated data-aggregation, and high-end analysis algorithms, AlphaVision enables market paticipants to identify activity in financial markets that would otherwise be difficult or impossible to detect.   Aqumin's technology platform provides seamless integration of large data streams from first tier providers such as ThomsonReuters, Bloomberg, NYSE, Nasdaq, ActiveFinancial and others; covering securities from all over the world and combining them into an interactive 3D visual and computational environment.  Customers benefit by being able to bring diverse information together to answer questions about securites markets and navigate the results in global context.  Much as Google Earth has enabled new classes of GIS applications through the use of XML APIs and mashups, Aqumin's AlphaVision is enabling new paradigms in the way professionals look at equity and equity deriviatives trading, risk management, order execution & buy-right, as well fundamental research.  AlphaVision is sold as a monthly subscription service or as an add-on to the Bloomberg Terminal (see REG<GO>). Headquartered in Houston, Texas with a presence in New York and Chicago, Aqumin is currently generating revenue and is presently raising captial to fund expanded marketing and sales opportunities in the United States and select markets globally.  Sean Spicer 3D, Computational Finance View Streamed Recording Download Video      
2020 Acceleware Startup Presentation Acceleware provides software solutions to harness the parallel processing capabilities of multi-core GPUs/CPUs for the Electronic Design and Oil & Gas industries. Our Acceleration platform seamlessly integrates with applications from industry leading vendors in the Electromagnetic Simulation, Seismic Data Processing, Reservoir Modeling industries and Linear Algebra solvers. Acceleware solutions enable users of single-threaded applications to access multi-core processing hardware and achieve dramatic compute speed-ups.  With Acceleware solutions installed, run times for data processing and simulation applications are reduced by more than 50 times.

Our customers share a common and urgent need - the need for powerful and timely computer modeling and testing. They are pushing the boundaries of innovation and demand faster product-development cycles for more complex products. They want more effective tools to interpret vast amounts of data. 
Michal Okoniewski General Interest View Streamed Recording Download Video Download Video    
2021 Immersive Media Startup Presentaiton Immersive Media Corp. (IMC) is the world leader in the technology and production of full motion, high-resolution, 360-degree video experiences for websites. Well known as the developer of Google Street View visual mapping, IMC's fully immersive spherical videos allow viewers complete control to explore environments with a simple click and scroll of the mouse. Viewers can plunge themselves into live 360-degree streaming video on location or full motion prerecorded video/webisodes.
Among its many applications, IMC's immersive video is utilized by major consumer product brands to increase customer loyalty, website interest, and viral campaigns. Tourism and travel industries capitalize on immersive video's unique ability to put viewers directly into a scene. TV and documentary production companies enhance their storytelling and viewer experiences far beyond the traditional narrow viewpoint of a single camera. Noted IMC clients include Mercedes-Benz, Nike, National Geographic, NBC Sports, the NBA, Red Bull, Adidas, Armani and the Emmys telecast.
IMC's immersive videos provide unparalleled visual intelligence in security applications for the military, fire and police departments and state and local government agencies. Whether it's on a school bus, in an office building, down an alley, or deep behind the lines in a military theater, immersive videos protect people, manage assets and ensure safety.
Myles McGovern General Interest, Visualization View Streamed Recording Download Video Download Video    
2022 NextIO Inc. Startup Presentation NextIO, Inc. is the leader in next-generation I/O virtualization solutions for today's dynamic data center in a variety of industries including enterprise, telecommunications, high performance computing, digital media and financial services. With its innovative Express Connect platform, NextIO offers the unique ability to virtualize I/O technology on any server, operating system, hypervisor and storage architecture. Leveraging PCI Express, NextIO offers true I/O virtualization for any end-point technology. Express Connect delivers unprecedented rack-level scalability, with I/O and server resources that can be scaled independently for 50-70% savings in capital, power, and cooling. NextIO's any-to-any I/O connectivity boosts performance and reliability while streamlining IT deployment, simplifying administration and reducing costs. Kenton C Murphy General Interest, Visualization View Streamed Recording Download Video      
2023 Motus Games Startup Presentation Motus Games is a leader in Human-Computer Interaction, founded in 2001 by MIT graduates and experienced entreprenuers with the intial goal of creating a motion-based light saber game, clearly an obvious first mission for a bunch of gamers.
While the prototype was successful and one of the first if not the first controller of its kind, the world didn't seem ready for motion-based control. So the prototypes went home with gamers and our focus turned to golf, where the team created the industry-leading iClub (, a wireless peripheral for the golf club that analyzes the golf swing and shows a 3D "X-Ray" of the swing.
In 2007, Motus Games returned to its roots with a mission to create and enable the best motion-based user experience for gamers. From the next-generation Darwin hardware to artificial intelligence and analysis middleware (codenamed Graffiti), Motus Games aims to drive the innovation behind how motion-based games are played and give developers the tools needed for fueling the next generation of motion-based interfaces for Visual Computing.
Satayan Mahajan 3D, General Interest View Streamed Recording Download Video      
2024 GASS Company for Advanced Supercomputing Solutions LTD Startup Presentation *Founded in 2008, after years of experience in Supercomputing and visualizations
*Privately owned
*Developing Hoopoe - a cloud infrastructure and service for GPU computing
*Providing libraries to assist developers with using GPU technologies from .NET and Java applications (CUDA.NET, jCUDA, OpenCL.NET)
*Develops GPU based solutions for the Israeli industry targeting general computing aspects and visualization
Mordechai Butrashvily General Interest, Visualization View Streamed Recording Download Video Download Video    
2025 Visualization Sciences Group, Inc The Visualization Sciences Group ("VSG"),  is the leading provider of a scalable application framework for 3D visualizing and understanding scientific and industrial complex and calculated datasets. VSG delivers 3D visualization solutions and services to developers through its Open Inventor toolkit product line and its Avizo end-user application to manage and visualize complex information, large quantities of data, and multi-modality datasets. Open Inventor provides cross-market solutions for the development of 3D interactive professional and commercial applications with state of the art data management and visualization. Avizo leverages all of the Open Inventor suite IP and delivers high-end visualization solutions to end-users in a number of vertical markets. VSG technology improves time to market, integrates and processes data from diverse sources, and presents visual analytics that enable users to understand their data. The technology continues to be applied to new applications across markets as the trend expands to acquire, integrate, visualize, process, analyze and present datasets in 3D. Jean Bernard Cazeaux 3D View Streamed Recording Download Video Download Video    
2026 Foster Findlay Associates Startup Presentation ffA originates and provides world leading 3D seismic analysis software and services to the oil and gas industry. Application of ffA Technology accelerates 3D seismic interpretaton, improves understanding of  exploration risk and enables optimal hydrocarbon exploitation strategies to be defined. Steve Purves 3D, General Interest View Streamed Recording Download Video Download Video    
2027 Multiverse Startup Presentation Multiverse is a pioneering interactive entertainment company. In addition to providing a revolutionary development platform to independent creative teams, Multiverse also works closely with leading brands to translate their ideas into engaging online consumer experiences. Corey Bridges, Rafhael Cedeno General Interest View Streamed Recording Download Video      
2028 Teradici Corporation Startup Presentation Teradici Corporation has developed PCoIP (PC-over-IP) technology, a unique remote display protocol, which makes network delivered computing a viable corporate computing reality. Founded in 2004 and headquartered in Burnaby, BC, Canada, Teradici for the first time enables an exceptional end user experience for datacenter-based computing. Through a combination of unique graphics algorithms, high-performance silicon processing, and workstation/server add-in cards, the company is changing how personal computers are used, deployed and managed. More information is available at  Dan Cordingley 3D View Streamed Recording Download Video      
2029 MotionDSP Inc. Startup Presentation MotionDSP makes revolutionary video software that dramatically improves video from a wide range of sources -- from mobile phones to surveillance cameras. Its patented, multi-frame video enhancement technology is available in Ikena, a "CSI-style" video forensic application used by intelligence agencies and law enforcement, and in vReveal, a Windows application for consumers, launched in partnership with NVIDIA in March 2009, which offers a one-click fix for the issues that plague consumer-generated video: shake, noise artifacts, pixelation, poor lighting, and low resolution.

MotionDSP's investors include In-Q-Tel, the independent investment firm that invests to support the mission of the CIA and US Intelligence Agencies, and NVIDIA, the world leader in visual computing.

MotionDSP is based in San Mateo with a development office in Nis, Serbia. 
Sean Varah Embedded & Mobile, Visualization View Streamed Recording Download Video Download Video    
2030 MirriAd Startup Presentation MirriAd is the only end to end digital marketing solution that creates a new way to reach audiences by embedding brands into video content. MirriAd offers content owners and advertisers a complete marketplace with MirriAd's online services platform, a full on-line management system. Content owners can upload their content securely and automatically to the highly secure, scalable platform and offer the inventory for sale. Advertisers can view a catalog of video content files and search for the video or set of videos that best meets their campaign needs. Using MirriAd an advertiser may select their preferred zones, insert their brand images and preview a low quality stream of the embedded advertising video before they commit and execute a buy.Videos can be ingested individually or in volume with various degrees of automation. By managing engagements and campaigns online, content owners can exert as much editorial control over the process as they need or as required by advertising standards or regulatory guidelines. Brand images and video can be embedded in professionally produced content such as DVD, TV and online and even user generated content. MirriAd's unique campaign management platform manages multiple stakeholders securely, allowing each stakeholder the appropriate level of visibility of the process: before the placement, and at final sign-off stage. The platform also delivers sophisticated metrics detailing the effectiveness of any particular campaign on any platform.  Mark Popkiewicz Computational Imaging, General Interest View Streamed Recording Download Video Download Video    
2031 Mirics Semiconductor Inc. Startup Presentation Mirics provides total system solutions for converged wireless connectivity on portable consumer electronic devices such as Notebook PCs, Media Players and Cellphones. The initial focus of the company is the delivery of free-to-air broadcast services and content to portable computing and communications devices.  The Mirics FlexiTV platform leverages re-configurable RF tuner hardware combined with a host processor based software demodulator.
The solution is capable of covering all broadcast standards in all terrestrial broadcast bands from simple AM radio at LW to DVB-T in UHF Band V. Currently focusing primarily on providing global TV and radio reception on PC platforms, FlexiTV offers substantial benefits in terms of system flexibility and solution cost and footprint when compared to conventional hardware based solutions.  Current solution offerings with the FlexiTV platform include: AM/FM Radio, DAB Radio along with ISDB-T, DVB-T and T-DMB for TV applications.
Mirics has been collaborating closely with NVIDIA to migrate increasing amounts of the FlexiTV software demodulator from the x86 CPU platform to the CUDA GPU platform. In doing so, Mirics and NVIDIA are demonstrating the long term viability of software based modems whilst at the same time illustrating the efficiency of the CUDA GPU as a signal processing platform for non-graphics applications. 
Simon Atkinson Computer Vision, General Interest, Embedded & Mobile View Streamed Recording Download Video Download Video    
2032 Milabra Startup Presentation Milabra is a visual recognition company that focuses on improving online advertising performance through the analysis and active utilization of commercially relevant visual data. Milabra uses advanced  neural networks to identify the commercially valuable visual aspects of images, video and graphics, uses predictive analysis to synthesize the data collected, and provides partners with actionable data to advertising performance online.

With over half of the Internet graphical in nature, and with over 8 billion photos and 500 million videos uploaded per month to social media in the United States alone, visual media is the most important and fastest growing component of the internet. The high cost of serving and storing this media cannot be offset with advertising dollars unless the contents, impact and meaning of that content is known. Milabra decodes that meaning and turns these visual liabilities into visible revenue.

Milabra's Image Intelligence platform and commercial decisioning framework is fast, accurate and easy to use. Our sophisticated neural network technology has the ability to add value to every image and visual element we see. Founded in 2008, the company was formed by a team of experienced entrepreneurs, industry veterans and scientists from the best visual recognition programs in the country. The company fuses neuroscience with proven advertising, psychology and grapic theory. 
Samuel Cox General Interest, Visualization View Streamed Recording Download Video Download Video    
2033 ICIDO GmbH Startup Presentation ICIDO, pioneer and global leader of Virtual Reality and Virtual Engineering Solutions support companies to make their product development process sustainable, competitive and environmentally friendly. ICIDO's virtual engineering solutions minimize risks, reduce failure costs and decrease the use of resources. They also allow for faster and easier global collaboration and communication, ensure better visibility and visualization of early stage products. At the end they integrate customers, partners and other departments  in the product lifecycle early.
ICIDO's Virtual Reality solutions are used throughout the entire product development process from design to manufacturing. 
Ralf Heimberg General Interest, Visualization View Streamed Recording Download Video Download Video    
2034 Hanweck Associates, LLC Startup Presentation As an established leader in high-performance quantitative financial computing, Hanweck Associates pioneered commercial GPU-based computing solutions for the finance industry, focusing tomorrow's technology on today's risk management, research and trading problems.  Our customers include top-tier banks, broker/dealers, asset managers and hedge funds around the world.  Hanweck Associates team brings over 40 years of hands-on expertise in financial markets and technology services.  Our experience as traders, strategists, researchers and  technologists at major financial institutions spans all major asset classes and products. Gerald Hanweck General Interest View Streamed Recording Download Video      
2035 DUX Soft Pvt. Ltd. Startup Presentation DUX Soft Pvt. Ltd. has been founded to design, integrate and deliver an innovative mix of productivity and performance based products and solutions, for enterprises working with floating point computations or rich media technologies, and in the process establish itself as a leader in the High Productivity Computing, Digital Media Solutions, Artificial Intelligence based Storage & Workflow integration market.

DUX serves its niche market by bringing together innovative ideas, scalable technologies, and a top-of-the-line team -- with over 50 years of combined experience setting up major animation studios in India, designing and implementing successful pipelines, and managing scalable infrastructures. DUX strives to provide exceptional value and peace of mind to our customers, minimize their costs, maximize resource utilization, and allow them to concentrate freely on their core competencies.

At the forefront of DUX's solutions is a unique digital workflow tool called SWIVEL which manages pipelines, rendering and assets and extends to provide digital mining capabilities for new media enterprises. DUX has also acquired a cutting edge Artificial Intelligence & Predictive Engine algorithm which it is using to create a self-learning and self-healing intelligent cache based Storage environment.
Alex Santo, Amit Srivastava General Interest View Streamed Recording Download Video Download Video    
2036 Anark Corporation Startup Presentation Anark Corporation provides powerful enterprise software and solutions to manufacturing organizations that need to cost-effectively utilize their 3D CAD data and other graphics assets for visualization, simulation, CAE, and supply-chain data exchange.

Anark helps market leaders such as Boeing, Cessna, Pratt & Whitney, Lockheed Martin, and NASA unlock the potential of their 3D data assets to save money, accelerate product development, and to collaborate and communicate more effectively and securely.
Stephen Collins 3D, Visualization View Streamed Recording Download Video Download Video    
2037 Allegorithmic Startup Presentation Allegorithmic is the first company to propose a professional middleware for the authoring and on-the-fly rendering of textures. Substance allows content developers to produce texture assets twice as fast as usual, while Substance description files are typically 500-1000 times smaller than regular textures. Allegorithmic's new product "Substance" is poised to redefine the development and distribution of rich content for the next generation of online games. Allegorithmic is currently partnering with the most innovative companies of the field to provide industry innovators with the most advanced packages for content production.  Sebastien Deguy Algorithms & Numerical Techniques View Streamed Recording Download Video      
2038 Affine Systems Startup Presentation Affine Systems is an online video advertising company that can target advertisements against specific videos across all the top online video web sites.  Affine uses its proprietary computer vision technology to scan all the videos on these web sites to find the best advertising opportunities for its clients.  Affine's core vision technology is able to recognize people, brands and products in the visual stream of a video, as well as detect inappropriate content and extract important information about the quality and content of the videos that it scans.  By using this metadata to target video advertising campaigns, Affine is able to achieve better ROI for advertisers and more revenue for publishers.  Michael Sullivan General Interest View Streamed Recording Download Video      
2039 HPC PROJECT Startup Presentation HPC Project, under the Wild Systems brand, provides application-in-a-box solutions for demanding users who require intensive computational power. The company brings to the market the cumulative experience of experts in simulation, parallel computing and computer architecture. The company is dedicated to package turnkey products, providing outstanding computing performance in a desktop form factor for interactive use of complex dataset.

Wild Systems Solutions are supported by best-of-breed technologies representing more than 25 years of R&D in the supercomputing field. This technology is based upon a formal analysis of source codes enabling correct automated transformations. These techniques are used to detect inherent parallelism in applications that could then be optimized to take advantage of the latest chipset architecture.

Such a technology happens to be particularly handy when dealing with specialized architectures such as GPU. These circuits require a particular attention in the organization of data in memory and of communication between on-board memory and main memory. Hand-coding this type of structures could be very tedious and error-prone. Our compiler framework provides in this context the necessary alignment of data and the pipelining of their communication for a high-performance execution.
Pierre Fiorini General Interest View Streamed Recording Download Video      
2040 Wild Pockets Startup Presentation Wild Pockets is the next generation solution for creating and monetizing interactive 3D content embedded right within your web-browser.  Whether you are an independent developer working on games in your garage or a large company looking for the fastest, most powerful development environment to create stunning immersive 3D web content – Wild Pockets is your solution. 
The Wild Pockets platform is disrupting the notion that you have to license and pay big dollars for development software before your project has even begun.  Instead, Wild Pockets provides FREE and equal access to all developers, professional, hobbyist and student alike, because you never know where the most innovative new ideas will emerge from!  Wild Pockets is designed as a full end to end solution and surrounding eco-system because it is takes more than just creating content to be successful.  All Wild Pockets content can be embedded and distributed instantly on any website anywhere online similar to a YouTube video, and the built-in transaction system and marketplace allows developers to monetize their content quickly and easily.  Check out the demos and sign up for your free account today at
Shanna Tellerman 3D, Embedded & Mobile View Streamed Recording Download Video Download Video    
2041 Luminova Startup Presentation Luminova utilises core NVIDIA-mental images 64-bit visualisation technologies in uniquely providing large scale construction/engineering project management systems through a web based, real-time interactive, highly detailed and photographic 3D visual information portal, utilising standard files from the existing project CAD system, where all of this data is communicated through our proprietary on-line interface application, Project Hub.
Luminova’s technology system uniquely handles data sets hundreds of times larger than any other existing technology. Our system revolutionises the speed, scale, accuracy and communication of the complete design and construction detail; resulting in significantly shortened design/engineering programs by providing  a faster and a far more assured decision making process.
The entire ‘as built’ construction detail can be readily assembled into a single environment that can all be viewed, measured, analysed, tested and shared in real time, by any number of concurrent users, anywhere in the world, through a standard web browser.
This system has already been proven on many very large projects including Heathrow Airport - Terminal 5, Greenwich Peninsula, Wembly Precinct and London 2012 Olympic Games Facilities and many others.
Christopher Blewitt 3D, Visualization View Streamed Recording Download Video Download Video    
2042 Emergent Game Technologies Startup Presentation Through our state-of-the-art software platforms, Gamebryo LightSpeed and Gamebryo, Emergent Game Technologies helps developers fulfill their maximum creative potential by giving them the power to endlessly – and efficiently – iterate development.  From prototype to final product, Emergent fuels creativity while reducing risk. Geoff Selzer General Interest View Streamed Recording Download Video Download Video    
2043 Empulse GmbH Startup Presentation Empulse GmbH is a Cologne/Germany based IT company focussing on innovative technology and Web applications.

The Empulse Indexing Technology (EIT)  is a framework for analyzing and searching large data sets on GPGPUs. This framework was developed for the implementation of a tourism web portal, accessing up to 5 billion travel offerings. Upcoming implementations of the  EIT    include OLAP and Data Warehouse style applications for telco and financial services industries.
Joerg Bienert General Interest View Streamed Recording Download Video Download Video    
2044 ENODO Startup Presentation By adapting state of the art video-game technology to industrial project's needs, ENODO became a leading 3D visualization company, providing cutting-edge solutions for real time & interactive manipulation of  industrial data. Laleh-Chloe Sahrai 3D, General Interest View Streamed Recording Download Video Download Video    
2045 NaturalMotion Limited Startup Presentation NaturalMotion Ltd. is a leading entertainment software company with offices in Oxford (England) and San Francisco (California). The company produces the widely-adopted animation technologies euphoria,morpheme and endorphin, used across the game and movie industries by companies such as Rockstar Games, LucasArts, Disney and Bioware as well as in the development of Backbreaker, the company’s first in-house game. Torsten Reil General Interest View Streamed Recording Download Video Download Video    
2046 BumpTop Startup Presentation Bump Technologies Inc. is a Toronto, Ontario, Canada-based company that has created and brought to market BumpTop, an innovative 3D user interface enabling natural organization and sharing. The technology was demonstrated at TED, the annual, invitation-only "Technology, Entertainment, Design" conference where the world's leading thinkers and doers gather to find inspiration; based on its success there, it received angel and venture capital funding and incorporated in 2007. Investors include GrowthWorks Capital, Extreme Ventures and angel investors Andy Hertzfeld, Austin Hill and G. R. Heffernan. Bump Technologies also provides OEMs and third party developers with a multi-touch and 3D rendering framework to enhance the interactivity and productivity of their applications. Anand Agarawala 3D, General Interest View Streamed Recording Download Video Download Video    
2047 IQ Engines Startup Presentation IQ Engines was founded as a collaboration of computer neuroscientists at UC Berkeley and UC Davis. We are inspired by the brain's vision system - how the brain represents images and how the brain recognizes objects. Our team has mathematical expertise in sparse coding and neural networks. It is our goal to bring advances in biological vision models to practical image and video search, using algorithms that are hierarchical and massively parallel (fast and accurate). We deliver these advances in image and video search on a web server platform that can be used by our customers to solve image and video search applications. Our team is the first in the world to deliver any-image labelling with 100% success and near-100% accuracy. A demonstration of the image identification technology can be seen at
Gerry Pesavento General Interest, Visualization Sorry - this session is not available.        
2048 Mersive Technologies Startup Presentation Mersive is a software company that enables the creation of large seamless displays with resolutions exceeding high-definition. 

As a leader in visual technology, Mersive software simplifies the way high-resolution displays are designed, configured, and maintained to provide a cost-effective, true auto-alignment solution capable of merging any number of projectors into a seamless display of almost any size, shape, and resolution.

Technology once considered practical only for select capital-rich industries is now made available by Mersive for a range of applications from simulation and training to museum displays and digital signage. Mersive’s products serve  corporate, government, and academic organizations.

Mersive mSeries Displays are unique in that they maintain their visual accuracy over time using auto-calibration software and occupy a very small footprint – 36 inches deep – compared to traditional rear-projection, seamless displays.  This small footprint enables use within existing spaces; it is not necessary to make any facility modifications.   Design, engineering & marketing teams can view all the details of their work in lifelike brilliance on one seamless, high pixel count display.  Because the display is easily installed and configured, maintenance costs are significantly reduced thereby opening new markets for large-scale, high resolution displays.
Randall Stevens Visualization View Streamed Recording Download Video      
2049 SceneCaster Startup Presentation SceneCaster is a leading provider of 3D social media applications that enrich the online consumer experience and enable highly immersive and contextual advertising opportunities. SceneCaster’s applications reach millions of users around the world and extend the reach and accessibility of 3D applications to a mass market of social networking and social media users. SceneCaster’s mass market appeal and immersive 3D Web technology offers advertisers a unique, interactive and entirely performance-based advertising solution that reaches a global audience of highly engaged users. Alain Chesnais 3D View Streamed Recording Download Video      
2050 Total Immersion Startup Presentation Founded in 1999 by Bruno Uzzan and Valentin Lefevre, Total Immersion is based in France and maintains a U.S. presence in Los Angeles. Its solutions are also available through a network of resellers worldwide.

Total Immersion designs and produces innovative, interactive exhibits, serving diverse business segments on a global basis, including theme parks and attractions, special events, retail, and other digital marketing environments. A versatile “platform for discovery” with an undeniable wow factor, Total Immersion offers consumers a compelling way to interact with brands in their own environment. With augmented reality, the brand temporarily “resides” in the viewer’s space. Imagine a favorite animated character sitting in the next chair, or a static product suddenly “come to life”– that’s Total Immersion’s augmented reality.

Total Immersion brings to market sophisticated yet practical applications of industry protocols and standards, including the most advanced proprietary recognition, tracking and rendering available from any source. D’Fusion is the leading marketable technology that injects a brand or product representation into a live video stream in real time; the interactive experience is not confined by the physical space of the screen or the display. Operating on standard PC devices and compatible with HD, D’Fusion is highly interactive, practical, educational and entertaining.
Bruno Uzzan General Interest, Visualization View Streamed Recording Download Video Download Video    
2051 Unity Technologies Startup Presentation Unity enables console-quality games for the desktop, the Web, the iPhone and the Wii console. The Unity tools have been hailed as revolutionary for enabling high quality results in an easily accessible development environment. Unity is the fastest growing platform with at least one new Unity game appearing for iPhone on a daily basis.  David Helgason General Interest Sorry - this session is not available.        
2052 Visuvi, Inc. Startup Presentation Visuvi develops targeted visual search engine solutions for a wide range of vertical applications in medicine, ecommerce and general-purpose visual search and maintains an index of images on the Internet. Visuvi has patent pending technology fort it’s search engine that examines the content and patterns within an image, categorizes that information via mathematical indexing and delivers search results based on the image itself – no text or meta-tags required.

Visuvi Inc. is a privately held company based in Redwood City, CA and is managed by a seasoned executive team consisting of Christopher Boone, President and CEO, Alexander Valenica, Chief Scientist and Co-Founder, Florian Brody, VP Marketing and Yuri Drozd, VP Product Management.
Christopher Boone General Interest, Visualization View Streamed Recording Download Video Download Video    
2053 Xulu Entertainment Startup Presentation Xulu is developing a new type of interactive experience based on massively distributed physics simulation that delivers unprecedented realism in a virtual world.  Capable of supporting up close and personal social interactions, alongside full high-end gaming and sporting experiences, the system provides open API’s and a suite of tools that allow both professional and novice users alike to contribute a broad range of highest quality content. The Xulu platform scales gracefully for a large number of simultaneous users with minimal reliance on centralized servers, and harnesses the power of leading-edge GPUs and CPUs to deliver ultimate online end-user experiences.

Compelling content is a key element of a successful launch, and is needed to attract a large initial audience that will drive demand for Xulu’s platform and generate early revenue. Xulu’s content will showcase the beginnings of an expanding universe with both real-world and fantasy-world elements. It will include a range of seed activities from extreme sports and gaming to gentle self-paced sightseeing excursions, episodic adventures, face-to-face socializing and play, and support for vibrant and viable communities. Third party and user-created entertainment, businesses, goods, and services will follow.

Xulu is leveraging more than 100 man-years effort in a prior related project that included the development of the first generation of the technology, virtual destinations, and multi-player games.
Joe Santos, Nanci Solomon General Interest, Visualization View Streamed Recording Download Video Download Video    
2054 Evolved Machines Startup Presentation We are pioneering the reverse engineering of brain circuitry to build intelligent machines. Paul Rhodes General Interest View Streamed Recording Download Video      
2055 Universal Robotics, Inc. Startup Presentation The company’s first product, Neocortex, to be sold in 2010, is the control system for a robotic box mover in the materials handling industry, a multi-billion dollar market .  Neocortex enables robots to stack and unload pallets with mixed size boxes, a capability  prospective customers have said is the “holy grail of automated material handling”.  Large distribution centers such as The Gap's, Gallatin, TN facility (a Neocortex alpha test site) will see significant reductions in material handling costs due to improvements in the speed of operation, a reduction in human errors, and increased safety and efficiency in warehouse operations, resulting in lowered green house gas emissions.
Furthermore, because Neocortex is only the software, it will radically expand the use of automation into new hardware applications such as forklifts, mining equipment, and railroad locomotives. Until now, these applications have been unable to capitalize on automation efficiencies due to the current limitations of Artificial Intelligence.
Alan Peters, David Peters Machine Learning & Artificial Intelligence View Streamed Recording Download Video      
2056 Kerosene and a Match Startup Presentation Based in Orange County, CA, Kerosene and a Match (KaaM) is a software developer building tools that leverage the massively parallel, low cost computing power of commodity graphics processors to build ultra-high performance cloud computing platforms.

The company's first product is a new multimedia discovery and indexing platform that combines GPU and cloud computing to give standard commercial search systems the ability to “discover” the content of “unsearchable” (aka, un-indexed and/or untagged by people) images, audio and video, and return the results just as it would text. Because upwards of 80% of all non-text content on the Internet is unsearchable, the KaaM technology offers organizations the ability to generate income from huge volumes of formerly unmonetized media.  KaaM licenses its technology to enterprises, government agencies and other organizations with large scale networks housing enormous amounts of rich media. The company also plans to deliver a software as a service (SaaS) offering for smaller entities late in 2010.
Cleve Adams General Interest View Streamed Recording Download Video      
2057 OnLive Startup Presentation OnLive is launching the world’s highest performance Games On Demand service, instantly delivering the latest high-end titles over home broadband Internet to the TV and entry-level PCs and Macs.

Founded by noted technology entrepreneur Steve Perlman (WebTV, QuickTime) and incubated within the Rearden media and technology incubator, OnLive spent seven years in stealth development before officially unveiling in March 2009.  OnLive, together with its Mova subsidiary, lies directly at the nexus of several key trends, all of which are reshaping the way we think about and use digital media:
    * The shift to cloud computing, displacing the limitations, cost and complexity of local computing;  * An explosion of consumer broadband connectivity, bringing fast bandwidth to the home; * Unprecedented innovation, creativity and expansion within the video game market.

Pioneering the delivery of rich interactive media to the home, OnLive will change the way that entertainment applications are created, delivered and consumed.
Steve Perlman General Interest Sorry - this session is not available.        
2058 Perceptive Pixel Startup Presentation “Perceptive Pixel is dedicated to the research, development and deployment of multi-touch interfaces for the knowledge worker. The company’s products enable both novice and expert users to manipulate complex datasets through a new class of intuitive yet powerful and visually rich interface techniques.
The company, founded in 2006 by Jeff Han, has applied these novel interfaces to applications in such areas as business intelligence, medical imaging, energy exploration, and defense. Recently, Perceptive Pixel became renowned for completely transforming the way television broadcasters including CNN, Fox and ABC covered the historic 2008 US Presidential elections. Most recently, Perceptive Pixel was awarded the 2009 National Design Award from the Smithsonian in the inaugurual category of Interaction Design.
Perceptive Pixel is privately held and based in NYC with offices in DC and Portland.
Jeff Han General Interest, Visualization View Streamed Recording        
2059 Ubitus, Inc. Startup Presentation Ubitus Inc. , a leading software platform provider for fixed-mobile convergence applications, develops a advanced cloud computing solution for telecom operators and service providers to drive consumer adoption of premium services relevant to sharing live video among friends as well as accessing digital media anytime, anywhere, without being restricted by the devices in use. Wesley Kuo Embedded & Mobile, General Interest View Streamed Recording Download Video      
2060 Zebra Imaging Startup Presentation Founded in 1996 and headquartered in Austin, Texas, Zebra Imaging, Inc. designs, develops and manufactures state-of-the-art 3-D display images for a variety of visualization applications in both government and commercial markets. Zebra Imaging is the only company that offers digital display systems that are autostereoscopic (no special eyewear required), full-parallax (viewing of the image from any view-point) and full-color. These innovations dramatically improve situational awareness, collaboration and decision-making in business-critical arenas, and with the introduction of new technologies, Zebra Imaging continues to improve the speed, quality and cost of creating compelling 3-D images. Zebra Imaging displays are portable, scalable, intuitively-understood and easily viewable to help customers turn design data into reality. Customers include Federal and Department of Defense agencies; leading architecture, design and planning firms; civilian topographic agencies and various groups who need to clearly and effortlessly demonstrate sophisticated 3-D data in a comprehensive, durable and portable medium. Michael Klug 3D, Visualization View Streamed Recording Download Video