Parallel Path Tracing

Parallel Algorithms

David Mayerich

Scalable Tissue Imaging and Modeling (STIM) Laboratory

Department of Electrical and Computer Engineering

Cullen College of Engineering

University of Houston

David Mayerich

STIM Laboratory, University of Houston

Parallel Path Tracing

Find ways to give your path tracer more work

  • Integration to produce better images

  • Deeper sampling (more bounces)

  • Higher resolution

  • More interesting geometry

David Mayerich

STIM Laboratory, University of Houston

Implement Integration

David Mayerich

STIM Laboratory, University of Houston

10 samples

100 samples

1000 samples

10000 samples

Implement Integration

Acquire multiple samples per pixel

  • Write a thread to perform rendering so that you can see the image updated as the thread is integrating

  • Implement a consistent averaging scheme:

    • Be able to calculate an averaged color after every sample

       

    • The image buffer only provides 8-bits! You may want to implement an intermediate buffer to store samples

David Mayerich

STIM Laboratory, University of Houston

10 samples

10000 samples

\mathbf{c}_n=\frac{n-1}{n}\mathbf{c}_{n-1} + \frac{1}{n}\mathbf{s}

Parallelize Evaluation

  • Once your rendering is threaded, parallelize the render loop

  • Use OpenMP to divide up the image into threads for rendering

  • How will you synchronize?

    • Will all of your threads write pixels at the same time?

    • Will the "number of samples" be the same for each thread?

    • How will you divide up the work?

  • Try multiple methods

    • Test round robin, blocking, patches

    • Sample each method and keep a record of differences

  • Be able to easily change the number of threads for testing

David Mayerich

STIM Laboratory, University of Houston