🧊 CUDA VolPT

Abstract

CUDA VolPT is the web edition of my Master’s Thesis. It examines optimization techniques for volumetric path tracing on GPU architectures, with particular attention to efficient participating-media rendering.

Web edition. This text was converted from the original LaTeX thesis. Equations, figures, and tables are retained in the individual chapters. You can also:

• Read the complete PDF version here
• View the source code on GitHub

Table of Contents

1. 1️⃣ Introduction Motivation, participating-media rendering, applications, and objectives.
2. 2️⃣ Problem Statement Radiance, extinction, participating media, and the volumetric rendering equation.
3. 3️⃣ Related Work Stochastic volume rendering and prior GPU path-tracing implementations.
4. 4️⃣ Background Monte Carlo estimation, volumetric path tracing, and CUDA execution and memory models.
5. 5️⃣ Optimization Methodology Kernel organization, path regeneration, compaction, memory locality, and sorting.
6. 6️⃣ Implementation Details Generic CUDA programming, scene assembly, and renderer data transfer.
7. 7️⃣ Discussion Measured performance, comparison with Mitsuba, and scene-dependent behavior.
8. 8️⃣ Conclusion Findings, limitations, and directions for further work.

Additional Materials

1. A List of Abbreviations Rendering, radiometry, and GPU terminology used throughout the thesis.
2. B Attachments Digital contents, build requirements, and test-system specifications.