Thesis (Selection of subject)Thesis (Selection of subject)(version: 393)
Thesis details
   Login via CAS
An Efficient Load-balancing Image Sampler for Path Tracing
Thesis title in Czech: Efektivní vzorkovač obrazu s podporou rozložení zátěže pro sledování paprsku
Thesis title in English: An Efficient Load-balancing Image Sampler for Path Tracing
Key words: sledování paprsku|vzorkování obrazu|paralení programování|počítačová grafika|renderování
English key words: path tracing|image sampling|parallel programming|computer graphics|rendering
Academic year of topic announcement: 2021/2022
Thesis type: Bachelor's thesis
Thesis language: angličtina
Department: Department of Software and Computer Science Education (32-KSVI)
Supervisor: prof. Dr. techn. Alexander Wilkie
Author: Bc. Tomáš Kubíček - assigned and confirmed by the Study Dept.
Date of registration: 20.10.2021
Date of assignment: 20.10.2021
Confirmed by Study dept. on: 26.04.2022
Date and time of defence: 12.09.2022 09:00
Date of electronic submission:20.07.2022
Date of submission of printed version:22.07.2022
Date of proceeded defence: 12.09.2022
Opponents: Dipl.-Ing. Thomas Klaus Nindel, Ph.D.
 
 
 
Guidelines
Goal of the thesis is to replace the existing but primitive parallel image sampler in ART (https://cgg.mff.cuni.cz/ART/), the rendering research toolkit developed and maintained by the Computer Graphics Group of Charles University. The existing component of the system was written as a functional placeholder until something modern could be developed: properly load-balancing and lock-free tiling image samplers that scale to many CPU cores are not trivial, so putting a simple placeholder class there made sense while the toolkit was not being used for large projects. However, this placeholder runs into problems on machines with many cores, and scales poorly: which is a significant obstacle for e.g. the computation of reference images, as it ties up too many resources on our large compute servers.

Sub-goals of the thesis include identification of a suitable tiled technique that load balances well while allowing for non-trivial sample splatting kernes, analysis of the resulting performance, and extensive testing of the implementation on the compute servers of our group, as well as comparisons with other Open Source rendering research software like Mitsuba.
References
Matt Pharr, Wenzel Jakob, and Greg Humphreys. 2016.
Physically Based Rendering: From Theory to Implementation (3rd ed.).
Morgan Kaufmann Publishers Inc., San Francisco, CA, USA.
 
Charles University | Information system of Charles University | http://www.cuni.cz/UKEN-329.html