ClusterGL - Home page

	ClusterGL is : . a cluster-aware real-time ray-tracing library that provides an OpenGL-like subset of functions. . designed having in mind a Beowulf cluster made by a set of economic eterogeneous Linux boxes on-line on an high speed LAN (Myrinet or Giga Ethernet). . designed to use MPICH for its the cluster driven part and standard X Window for it's graphics aware part.
Downloads : . CGL Distribution v1.00 Skinny distribution with source code under GNU GPL	The ClusterGL goals are : . Interactive results (15 fps) . Mantain a compatibility with OpenGL permitting rapid and little work migration from OpenGL source -> ClusterGL source obtaining ray-tracing quality and mantaining the interactivity. . Being cluster aware but cluster independent. This is reached using standard MPI(CH) implementations. . Creating a complete set of libraries like GLU, GLUT, GLAUX to provides Platonian Solids and X Window events driver.
Links : . OpenGL.org . MPI Standard	The ClusterGL backgrounds are : . Beowulf Linux Cluster made by 6 nodes . High speed, high performance 1.28Gbit/s Myrinet LAN with a backup 100Mb/s Ethernet LAN . Particular attentions to standard to try not to fake exploiting hardware. Our results a 100% cluster-CPU power, no hardware accellerations. This makes our results widely exportable.
Authors' email : Rosario De Chiara Ugo Erra Project email : ClusterGL	The ClusterGL current status of development is : . v0.70 - Fully functional ray-tracing engine providing shades, shinyness, directional lights, point lights. Not yet cluster aware. 3~4fps on a singles node. . Working on : Reflections and penumbra. . Next stage : Parallelization of the ray-tracing algorithms. . v1.00 - Fully functional ray-tracing engine providing shades, shinyness, directional lights, point lights. Reflection and refraction fully functional >Cluster aware<. 13~15fps on a 5 nodes cluster. . Working on : Load balancing
The ClusterGL snapshots :
.Snapshot1 : lights sample .Snapshot2 : refraction .Snapshot3 : shadow and shinyness
Best viewed in a 800x600 browser.