The heterogeneous computing research group works on projects mainly in four fields: simulation, visualization, geometry and seismic processing. After a brief introduction to these areas, scroll further down to find examples of both previous and ongoing activities in projects we run. (For more information grouped according to formal project definitions, please see the "about us" tab.)
Geometry, especially spline technology, has traditionally been one of the most important research areas at the Department of Applied Mathematics. Our research group also has strong ties to the internationally renowned spline community at the University of Oslo.
We have coupled these traditional relations and skills with heterogeneous computing, speeding up existing algorithms, finding new uses of existing algorithms, and even design completely new algorithms specifically for heterogeneous architectures.
Our work includes view-dependent subdivision of silhouettes on triangle meshes, ray-casting of implicitly defined surfaces, decimation of finite element meshes and parallelized intersection of spline surfaces.
Visualization is often both computationally demanding, and requires handling of huge datasets. Accelerating visualization techniques using heterogeneous computing can enable interactive handling of data not previously possible.
Visualization is a growing activity in the Heterogeneous Computing group. It is simply not possible to interactively handle many datasets without accelerated computing. This becomes a greater concern with the ever increasing size of datasets. One example of scientific visualization is the Voluviz viewer created for the heart project at simula.research laboratory, where the datasets were too large to handle interactively by conventional methods.
Simulation of physical phenomena modeled by partial differential equations was one of the first areas we started researching. Some applications we have accelerated have been more than an order of magnitude faster than the CPU.
We have shown orders of magnitude of performance gains when using the GPU to accelerate applications. These applications include the shallow water equations, the Euler equations, among others.
Using GPUs for simulation allows for interactive simulation of large scale water basins with interactive visualization. The effects of floods and tsunamis are devastating and realtime performance can ultimately save lives.
The need for visualization of multidimensional, multivariate vector fields arises in different contexts. Since there is no such thing as photorealistic rendering of vector fields, a major challenge is to develop visualization techniques providing useful information to the user. Currently, our research efforts aim at providing a flexible framework for interactive visualization of out-of-core vector fields for a wide range of sources, including meteorological data stored in regular grids and seismic data stored in corner-point grids. In the Parallel3D project, we have an activity on this topic.
For more information, contact Jon Hjelmervik.
HPMC is a GPL OpenGL/C/C++-library that extracts iso-surfaces of volumetric data directly on the GPU.
HPMC is by far the fastest marching cubes implementation available!
The library analyzes a lattice of scalar values describing a scalar field that is either stored in a Texture3D or can be accessed through an application-provided snippet of shader code. The output is a sequence of vertex positions and normals that form a triangulation of the iso-surface. HPMC provides traversal code to be included in an application vertex shader, which allows direct extraction in the vertex shader. Using the OpenGL transform feedback mechanism, the triangulation can be stored directly into a buffer object.
Read more about the library at the project page.
The Tecwel2 activity is a project for the company TecWel, in which the goal is to develop a software application to accompany one of their measuring tools. In this project, the GPU is currently used for standard visualization tasks. In a next phase, multiple CPU cores and GPUs will be used to speed up computations. See for example another of our activities, HPMC, for a GPGPU-exploitation very suitable for an application such as this.
You see here a very short session, in which a particular iso-surface of a cylindrical scalar field is rendered. The application offers numerous tools to manipulate the scalar field, transfer functions and iso-surface rendering parameters, all in order to facilitate the identification of certain geometric features of the model being scanned. Only a small subset of the possibilities offered is shown in this short sequence.
To get in touch with the project leader, you may click here:
Published December 3, 2009
