Richmond, Katie and Asher, Hancock and Shervin, Sammak and Barry, Matthew
(2021)
Numerically resolved Radiation View Factors via Multi-GPU Accelerated Ray Tracing.
In: UNSPECIFIED.
Abstract
A robust computational framework is presented to directly solve for the radiation view factors (Fij) of participatory surfaces within complex three-dimensional geometries. This framework exploits the embarrassingly-parallel nature of the formulation and solution of Fij through a multiple graphics processing units (GPU)-accelerated ray tracing scheme. The presented computational methodology was developed in Java and incorporated Aparapi for OpenCL compatibility. The surfaces of the geometries of interest are constructed via the creation of stereolithography (STL) files, which represent surfaces as tessellations. The shadow effect, where cast rays are obstructed by non-participatory surfaces, is handled via the Möller-Trumbore (MT) ray-triangle intersection algorithm. To ensure generality and robustness, a self-intersection algorithm is implemented for both planar and non-planar surfaces via the MT algorithm with back-face culling enabled. Validation of the algorithm was performed for a variety of three-dimensional geometries. The proposed multi-GPU framework was benchmarked to a conventional computer processing unit (CPU)-based version of the code and exhibited substantial decreases in computational time. Results indicate that near-linear speed-up is achievable with increasing numbers of GPUs. Additionally, a converging solution is obtained with increasing tessellation and GPU count, indicating no perceivable discrepancy in solutions in comparison to CPU and single-GPU based solutions.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[img]](http://d-scholarship.pitt.edu/style/images/fileicons/text_plain.png)
![[feed]](/images/feed-icon-32x32.png){kind=icon}