Fast and cross-vendor OpenCL-based implementation for voxelization of triangular mesh models

Abstract

ABSTRACTAs the open standard for parallel programming of heterogeneous systems, OpenCL has been used in this study in the context of a particular and intensive computing task, namely the voxelization of tessellated objects. For this purpose, OpenCL platform has been utilized to develop a parallelized voxelization algorithm that relies on a fast and efficient triangular mesh facet/cube overlapping test. The extensive numerical tests conducted with heterogeneous hardware configurations on geometric objects of varying complexities, mesh/domain sizes, and voxel resolutions suggest that up to 99.6% or 260 times decrease in the computation time can be obtained when GPU- or CPU-based parallelized techniques are used instead of the conventional single-thread CPU approach. Future developments will attempt the integration of the current implementation into a virtual orthopaedic surgery platform.