Cluster:Realtime Parallel Visualization

From Earlham CS Department
Jump to navigation Jump to search

Realtime Parallel Visualization

Graphical Methods vs Visualization of a Problem Space

There are two paths we could take with this project. The main goal of the first would be to create parallel versions of previously serial graphical methods. This would envolved a lot of learning of the theory behind graphics and would be a serious challenge. As this ground has been fairly well explored, we would have to be either esoteric in the methods we choose to parallelize or we would have to choose a known parallelized method and improve upon it. It seems this would take a great deal of learning, theory, and engineering to pull off.

Alternatively, the second path, visualizing an existing problem space, consists of visualizing the data of an existing problem. A very large data set would be needed to take advantage of the parallel aspect of this project. The advantages to this path include the easy (and almost necessary) inclusion of a computational science project. Starting with the visualization of GROMACS results would be very reasonable. Another perspective on the second path would be the corrolation and pre-processing of a large data set in parallel (i.e. compiling checkpoints of an MD simulation for a temporal view of the problem) for realtime visualization on a single machine. This could result in visulization tools that are much more meaningful than what is currently available (I'm thinking realtime, free perspective visualation of proteins folding - not only do you have complete control of camera movement, but you have complete control over time as well).

A side note: If we chose a multidimensional data set to work with, it would involved much more computation and would be look very impressive. We also have a head start with the ndopengl project.