February 16, 2024
Bounding Volume Hierarchy (BVH), a spatial tree, has been increasingly used to perform Collision Detection. However, BVH traversal is an irregular operation and is not well-suited to be accelerated on GPUs. We present Mochi, a fast collision detection engine that accelerates BVH construction and traversal by taking advantage of the capabilities of Ray Tracing cores of GPUs. We introduce multiple new reductions to perform collision detection on three types of objects: simple spherical particles, objects described by mathematical equations, and complex objects composed of a triangle mesh. Mochi achieves several orders of magnitude speedups on synthetic datasets and 5x-28x speedups on real-world triangle mesh datasets.
About Durga Mandarapu
Durga is a CS PhD student advised by Prof. Milind Kulkarni. Her research interests are in High-Performance Computing and Databases. Her work mainly focuses on using GPUs to accelerate irregular tree-based applications.