Parallel Transformation of Bounding Volume Hierarchies into Oriented Bounding Box Trees

This repository contains a reference implementation of the method described in:

Parallel Transformation of Bounding Volume Hierarchies into Oriented Bounding Box Trees
N. Vitsas, I. Evangelou, G. Papaioannou, A. Gkaravelis
to appear: Eurographics 2023
[ Paper ]

Building

The project uses the cmake build system.

To configure and generate platform specific build scripts, execute:

cmake -S src -B build

NOTE: If CUDA is not found automatically from cmake, you will need to manually set the CUDA_TOOLKIT_ROOT_DIR variable and reconfigure.

To build the project, execute:

cmake --build build --config Release

Binaries will be created in bin/Release.

Running

The main application is korangar.exe. Integrators and intersectors are compiled as separate, dynamically loaded libraries.

cd bin\Release

This release offers two integrators

pt_gpu and bench

and two intersectors

lbvh_gpu and atrbvh_gpu

The app exposes a very basic CLI.

-s <path/to/scene.gltf>
-intersector lbvh_gpu|atrbvh_gpu
-integrator bench|pt_gpu
-use_obbs
-collapse

• pt_gpu is a basic preview of the provided scene that uses the selected intersector to render the object using ray tracing.

• bench implements a benchmark that calculates the OBB of the point set of the provided scene using the CPU version of DiTO and the GPU-accelerated version of DiTO.

• lbvh_gpu is a custom implementation of the GPU-accelerated LBVH builder by Tero Karras .

• atrbvh_gpu is an adapted implementation of the GPU-accelerated ATRBVH builder by Domingues and Pedrini.

E.g. run:

korangar.exe -s ../../rsrc/gltf/hairball/hairball.gltf -integrator pt_gpu -intersector atrbvh_gpu -use_obbs -collapse

To preview the hairball scene using OBB traversal with ATRBVH and node collapsing.

or

korangar.exe -s ../../rsrc/gltf/hairball/hairball.gltf -integrator pt_gpu -intersector lbvh_gpu -use_obbs -collapse

to use the LBVH structure.

To execute the benchmark that builds the standalone OBB for a scene, run:

korangar.exe -s ../../rsrc/gltf/dragon/dragon.gltf -integrator bench

Implementation

The paper contributes two algorithms:

• A standalone, GPU-accelerated implementation of the DiTO algorithm that calculates a high quality OBB for a point set.

The algorithm is implemented in src\common\algorithm\dito\dito.cu with the function kr_cuda_points_obb

• A parallel, GPU-accelerated transformation of an existing tree based on AABBs to a tree based on OBBs

The algorithm is implemented in src\ads\common\cuda\transforms.cu with the function kr_cuda_bvh_obb_tree

Acknowledgements

This research was funded by the Hellenic Foundation for Research and Innovation (H.F.R.I.) under the “3rd Call for H.F.R.I. Research Projects to support Post-Doctoral Researchers” (Project No: 7310, LOTUS).