Project 3: Saksham Nagpal

[thecapablesnakekeeper]

Using 3 late days
Repo Link

Features Implemented

• Core Features

  • Materials - diffuse, specular (perfect + imperfect)
  • Toggleable stream compaction for ray path termination
  • Toggleable caching of first bounces
  • Toggleable material sorting

• Extra Features

  • Refraction
  • Stochastic Sampled Antialiasing (Toggleable)
  • Physically Based Depth of Field (Toggleable)
  • glTF mesh loading
  • albedo mapping
  • Bounding Volume Hierarchy (Toggleable)
  • Reinhard operator and gamma correction (Toggleable)

• Modifications to CMakeLists.txt

  • Added tiny_gltf.h

• Feedback

  • I and most other students ended up doing all the graphics-related extra credits, which in my opinion is not the aim of the course. It would have been better to give us a basic path tracer and ask us to specifically implement GPU-related optimization techniques such as streams and graphs. It would also have been better to cover those topics in earlier lectures- - they were covered 2 days before the project's due date, and that was not enough time to learn and implement such new concepts.
  • We already go in the depths of a path tracer in CIS5610, so I think Penn courses should not trip on each other and this homework should be switched to something more compute-based to avoid repetition of projects. Also, it is unfair to ask people with no graphics background to learn and implement a path tracer in less than 2 weeks. The learning curve is a lot of overhead for someone who is only interested in 'GPU Programming and Architecture', as the course name says.