Abstract

Ray tracing (RT) is a graphics rendering technique that simulates the physical behavior of light as it interacts with objects in a scene to generate high-fidelity images, often at the expense of high computational and memory costs when determining ray-object intersections. To mitigate the computational costs, a specialized RT accelerator is used to yield high-quality 3D images rapidly on modern GPUs. However, conventional RT accelerators can increase the computational speed but cannot effectively reduce the memory traffic produced during RT. The memory traffic between the host and GPU device memory often contributes to the significant energy consumption of RT processing. In this work, I will present a new RT accelerator that utilizes our proposed multi-level quantization method, enabling RT to directly utilize low-bit arithmetic hardware units without requiring additional format conversion. The quantized RT processing is composed of bounding boxes encoded in formats with varying bit widths, which minimizes unnecessary bounding volume hierarchy (BVH) tree traversal overhead due to the use of low-bit arithmetic. Additionally, it decreases the hardware area needed for intersection calculations with rays by utilizing simple integer hardware units in the RT accelerator. As a result, our proposed RT accelerator enhances the energy efficiency and performance of RT processing compared to modern GPU RT accelerators.

Bio.

Dr. Tsung Tai Yeh received his Ph.D degree from the Department of Electrical and Computer Engineering at Purdue University in 2020. He is currently an Associate Professor in the Computer Science (CS) Department at National Yang Ming Chiao Tung University (NYCU), Taiwan. His research interests span computer architecture, computer systems, and programming languages, with a primary focus on the GPU, Domain-Specific Accelerators, and AI compiler systems. Dr. Yeh has also been recognized as the recipient of several awards. He received a Fellowship of the HEA from Advance HE in 2023, as well as the Excellent Mentor Award from NYCU in 2021 and 2025. He won the 2025 Qualcomm Innovation Fellowship. He is also a recipient of the 2030 Cross-Generation Young Scholars Program (NSTC), Taiwan, 2025. His compiler research work was nominated for the Best Paper Award at the PPoPP conference and his research work was published in multiple top-ranking conference proceedings (ISCA, ASPLOS, HPCA, PPoPP, NeuraIPS).

All faculty and students are welcome to join.