Title:
|
A HETEROGENEOUS CLUSTER FRAMEWORK FOR COMPUTATIONALLY HEAVY VISUALIZATIONS |
Author(s):
|
Martin Ericsson , Anders Hast , Stefan Seipel |
ISBN:
|
978-972-8924-97-3 |
Editors:
|
Hans Weghorn and Pedro IsaĆas |
Year:
|
2009 |
Edition:
|
V II, 2 |
Keywords:
|
High Performance Computing, Visualization, Ray tracing, Heterogeneous, Cluster |
Type:
|
Poster/Demonstration |
First Page:
|
337 |
Last Page:
|
339 |
Language:
|
English |
Cover:
|
|
Full Contents:
|
click to dowload
|
Paper Abstract:
|
It is desirable to use different kinds of hardware for the same software project but it is in a lot of cases not feasible due to
the programming effort that it takes to get it running. We have developed a framework to gain more experience with
dealing with several different computer architectures in the same project and to have a test suit for different applications.
The hardware we used contains both homogeneous x86 multi-core processors as well as heterogeneous Cell Broadband
Engine processors. We developed an example application in form of a distributed rendering system which demands a lot
of computational power to test it. Also incorporated in this project is a stereoscopic display that put limits on resolution
and frame rate of the system. From the lessons learned in this ongoing work we conclude that it is feasible to build and
use heterogeneous resources for this kind of application even though it will take a bit of effort. |
|
|
|
|