Scalable Visualization Techniques
BME Innovation and Knowledge Centre of Information Technology (BME IT2)
Introduction
Our partner within the BME IT2 project is the High-Performance Computing division at Hewlett Packard (HP), which designs and develops cluster hardware and software technology supporting integrated computation, storage, and visualisation. In a distributed implementation of a visualisation method, the participating computing resources complete the sub-task assigned to them according to the chosen strategy and their results (partial images) must be combined in order to produce the final image. Combining partial images is called image compositing, which is one of our project's main focus. Since nowadays there are no standard software solutions for image compositing, HP, integrating existing efforts, developed a specification and an implementation for image compositing, the Parallel Compositing Library (ParaComp API).
The visualisation methods and applications developed within the knowledge centre are part of HP's demonstration materials for the SVA (Scalable Visualisation Array) and ParaComp products. Developments of the compositing library will be integrated into the product. HP set as an objective for this year that their SVA system should become suitable for general purpose (non-graphical) computing as well, i.e. the graphical processing units (GPU) in a cluster accomplish general purpose computations (GPGPU approach).
The primary goal of the R&D activity is the research of algorithms and rendering methods for visualising large (giga or even terrabyte sized) datasets in real-time and the development of applications based on these results. In order to accomplish these goals, at the BME IT2 we develop solutions that incorporate new algorithmic approaches which have better algorithmic complexity properties than the existing solutions. On the other hand, we also design and implement parallel and distributed versions with superior performance and/or more efficient use of hardware and software resources.
News
RENO, Nev., Nov. 12 -- HPCwire, the leading source for global news and information covering the ecosystem of high productivity computing, revealed the winners of HPCwire's annual Readers' and Editors' Choice Awards at the 2007 International Conference for High Performance Computing, Networking, Storage and Analysis (SC07) in Reno, Nevada.
A Best Price/Performance Visualization Product of Technology with the contribution of BME(IT)2
- Readers' Choice Award Recipient: HP Scalable Visualization Array
- Editors' Choice Award Recipient: HP's Parallel Compositing Library
Recent Projects
The application simulates multiple scattering in participating media in a distributed environment. Radiance transfer is approximated with an illumination network. Read more... |
 |
A translucent volume rendering application designed for Hewlett-Packard Scalabe Visualization Array. This application uses static object-space distribution of the data among the rendering nodes and parallel compositing to get the final results. Read more... |
 |
This is an illustrative volume visualization application designed for Hewlett-Packard Scalabe Visualization Array. Read more... |
 |
The application allows the visualization of volume datasets in a distributed environment by rendering isosurfaces with ambient occlusion shading using screen space distribution.Read more... |
 |
An Eulerian fluid simulator application designed for Hewlett-Packard Scalabe Visualization Array. Our algorithm is based on the Eulerian solution of the Navier-Stokes equations, and runs the simulation on the GPUs of a computational grid. The distributed implementation makes it possible to solve the equations on higher resolution data sets than in case of a single computer application while still preserving interactive frame rates. Read more... |
 |
A Smoothed Particle Hydrodynamics (SPH) fluid simulator application designed for Hewlett-Packard Scalabe Visualization Array. This particle-based application uses volume domain distribution among the nodes. Read more... |
 |
This is an interactive 3D walkthrough of the SDSS DR6 astrophysical data set implemented to run on the Hewlett-Packard Scalabe Visualization Array. This application uses static object-space distribution of the data among the rendering nodes and parallel sort-last compositing to get the final results. Read more... |
 |
A 3D texturing based volume rendering application designed for Hewlett-Packard Scalabe Visualization Array. This application uses static object-space distribution of the data among the rendering nodes and parallel compositing to get the final results. Read more... |
 |
A ray-casting based volume rendering application designed for Hewlett-Packard Visualization Array. The application allows the visualization of volume datasets in a distributed environment by rendering isosurfaces using screen space or object space decomposition approaches. Read more... |
 |
ParCompMark (Parallel Compositing Benchmark) framework allows simulating various CPU, GPU exercises, rendering and compositing modes for different software scenarios. It uses the PC (Parallel Compositing) API and generates XML output. Read more... |
 |
Visualization Data Server
The upper projects do not contain any sample data. The data sets can be downloaded from http://visdata.ik.bme.hu/.
Contributors
 |
 |
 |