My favorites | Sign in
Project Home Downloads Wiki Issues Source
Search
for
GSoC2011_ClusterGL  
Information about my Summer of Code project, ClusterGL
gsoc
Updated Oct 17, 2011 by kiel.endpoint

Introduction

ClusterGL is a system to transparently intercept OpenGL API calls and spread them over multiple machines. This means that it isn't necessary to write applications specifically to run over multiple machines.

I developed it originally to run on the Symphony Cluster display wall at my university, and it's now being adapted to run on Liquid Galaxy hardware.

Above: Running unmodified OpenArena on our wall under ClusterGL. There's five display nodes, one for each column of screens. Total resolution is 8400x4200, although it's run at a virtual 8880x4560 to allow for bezels.

Comparison to Chromium

Conceptually CGLs aims are similar to the Chromium parallel rendering system.

The differences:

  • Efficiency. We have gone to some length to optimise the process - on the above OpenArena test, CGL is approximately 3x faster than Chromium on the same hardware.
  • Modernity. Unfortunately, Chromium development has slowed. CGL implements a larger amount of newer graphical features, letting more advanced applications run.

Check out the paper linked below for details.

Technical Information

We presented ClusterGL at Eurographics 2011.

Current status

Check out the Google Code page: http://code.google.com/p/clustergl2/

Snapshots of ClusterGL running at UWS - from Andrew

We have a prototype Liquid Galaxy rig at UWS, called the Wonderama, which we've been using to run up & test Paul's code. Below are some snaps of the progression over the course of the project (Andrew, Paul's GSoC Mentor).

16 July 2011

29 July 2011

13 August 2011

Powered by Google Project Hosting