Radiation Transfer Computation

for Hewlett-Packard Scalabe Visualization Array

(December, 2008)

The application simulates multiple scattering in participating media in a distributed environment. Radiance transfer is approximated with an illumination network.
> video1, video2, video3

Introduction

We present a fast parallel solution for the radiative transfer equation with an iterative approach. This posed challenges for the parallel implementation because we should attack the sublinear scalability and the communication bottleneck. Our approach is based on two recognitions. Iteration is slow because it requires many .warming up. steps to distribute the power of sources to far regions. Thus, if we can find an easy way to approximate the solution, then iteration should only refine the initial approximation, which could be done in significantly fewer steps. On the other hand, iteration requires the exchange of data from all computing nodes in each step, which leads to a communication bottleneck. We propose an iteration scheme when the data are exchanged less frequently. This slows down the convergence of the iteration, so computing nodes should work longer, but reduces the communication load.

Documentation

File	Description
radtransf-sva-0.1.pdf	Manual of the renderer

Downloads

File	Description
radtransf-0.1-1.x86_64.rpm	Compiled version of the translucent volume renderer for AMD64 HP-XC
radtransf-0.1-1.src.rpm	Source version of the translucent volume renderer for AMD64 HP-XC
glew-1.3.4-1.x86_64.rpm	OpenGL Extension Wrangler (RPM package for AMD64 HP-XC)
glew-devel-1.3.4-1.x86_64.rpm	OpenGL Extension Wrangler headers (RPM package for AMD64 HP-XC)
glui-2.35.zip	GLUI library source (static)