This repository shows the experimental code used to obtain the OptiX-related measurements in the paper:
"Spatial Partitioning Strategies for Memory-Efficient Ray Tracing of Particles", LDAV 2020, to appear
This application has been used on Windows 10.
This application requires CUDA 10.1+, OptiX 7+, and TBB.
It depends on OWL. Checkout tag ldav_2020_sub from the fork https://github.com/UniStuttgart-VISUS/owl.
Original repo https://github.com/owl-project/owl.
The folder structure has to be:
___ some folder name
|___ owl
|___ rtxpkd
for rtxpkd to find OWL and to use it in subsequent build steps.
Run rtxParticleViewer [options] <filename>
The application determines from the file extension which loader to use.
Accepted file types are: p4, xyz, random, pkd, and mmpld.
For mmpld files, the USE_MMPLD option has to be ON in the CMake configuration.
PKD files are dumps of datasets after pkd construction.
Random is a type indicating that a random dataset has to be created by the application.
The syntax is: <num>_<radius>.random.
Num determines the number of particles to generate.
The value of radius is currently overriden by the default value of the --radius argument.
Subset of arguments accepted by the application:
| Option | Description |
|---|---|
| --allpkd | renders scene with a PkD tree |
| --nopkd | renders scene with a BVH |
| --treelets | renders scene with the nested hierarchy using PkD treelets (combine this with max-size) |
| --treelets-brute | renders scene with the nested hierarchy using flat lists (combine this with max-size) |
| --max-size <size> | determine the maximal size of treelets in the nested hierarchy |
| --rec-depth <depth> | recursion depth of the ray tracer |
| --radius <radius> | set radius of the rendered spheres |