M.Sc. Christoph Neuhauser


Picture of Christoph Neuhauser

M.Sc. Christoph Neuhauser

Postal address

Postal:
Boltzmannstr. 3
85748 Garching b. München

Research

I am a PhD student at the group of Prof. Dr. Rüdiger Westermann. My major fields in research comprise scientific visualization and real-time rendering. Furthermore, I investigate the application of deep learning techniques in these two fields.

Topics for Bachelor and Master Theses

Evaluation of the Performance of Raster Order Views (ROVs) on AMD GPUs (Master)

Order Independent Transparency (OIT) is a class of techniques for correctly rendering transparent geometry with alpha blending. Raster Order Views (called Fragment Shader Interlock in Vulkan and OpenGL) are often used by different Order Independent Transparency (OIT) techniques like Multi-Layer Alpha Blending (MLAB) or Moment-based Order Independent Transparency (MBOIT). In our paper "A Comparison of Rendering Techniques for 3D Line Sets with Transparency", we have compared different OIT techniques, many utilizing Fragment Shader Interlock.

On AMD hardware, Fragment Shader Interlock is unfortunately so far not available to developers using OpenGL or Vulkan. Developers at AMD previously have raised concerns with respect to the performance aspect of Fragment Shader Interlock and its limited use in video games. Thus, this hardware feature can so far only be used on DirectX.

On NVIDIA and Intel hardware, we were able to achieve performance benefits when utilizing Fragment Shader Interlock in an OpenGL program, and would like to investigate how this behaves on AMD GPUs.

Your tasks:
(1) Write a Direct3D 12 test program in which you can render line data without any transparency effects.
(2) Port the most important OIT techniques (like MLAB and MBOIT) from our OpenGL program (https://github.com/chrismile/LineVis) to your Direct3D 12 program.
(3) Write code for measuring the GPU frame time needed for each of these algorithms in an automated manner.
(4) Compare the algorithms with respect to their performance.

NOTE: You do NOT need to own an AMD GPU in order to do this thesis. If you are interested in this topic and have a GPU from another vendor, it should be possible that the chair buys a GPU that you can do testing on via remote access.

Supervised Theses

Topic Student Type of Thesis
2021    
Evaluation of Adaptive Sampling Techniques for Ray Tracing (ongoing) Tim Kaiser Master
Extension of Precomputed Real-time Global Illumination by Specular Reflections (ongoing) Berk Saribas Master
Terrain Rendering using RT Cores in Vulkan (ongoing) Florian Hamberger Bachelor

 

Publications

A Comparison of Rendering Techniques for 3D Line Sets with Transparency
Michael Kern, Christoph Neuhauser, Torben Maack, Mengjiao Han, Will Usher, and Rüdiger Westermann
IEEE Transactions on Visualization and Computer Graphics, 2020
[doi] [Bibtex]