Coherent (Faster) Raytracing

[brlcad]

Outline

Utilize cache coherency and parallelization techniques in BRL-CAD's ray tracing pipeline, and demonstrate graphically.

Details

BRL-CAD has one of the oldest and fastest parallel ray tracing implementations around but we don't currently leverage modern techniques that minimize context switching and data access. With implicit geometry and constructive solid geometry (CSG) Boolean operations, we also have a very specific set of validation and performance challenges.

Your project is to help make the pipeline more data-coherent using ray bundling, cache-coherent processing, and OpenCL. Currently, the ray tracer evaluates one ray at a time (in parallel). There is experimental work with OpenCL in a branch.

Your primary objective for this project is to make the ray tracer evaluate groups of rays at a time (e.g., 32x32 "postage stamps") coherently, calculate and demonstrate changes made to performance, and get ray tracing via OpenCL working. A secondary stretch objective is to demonstrate performance graphically using one of several options we've explored in the past (OSPRay, O3DE, and Qt).

Your proposal should be aware of, improve upon, and/or incorporate previous year efforts:

• 2020 OpenCL GPGPU Raytracing
• 2018 Improving Raytracing Performance
• 2017 GPU Boolean Evaluation
• 2016 GPGPU Support for Boolean Evaluation
• 2015 GPU Accelerated Ray Trace Rendering

Expected Outcome

Demonstrate improved ray tracing performance via bundling and OpenCL. Demonstrate performance graphically through a prototype interactive GUI (using OSPRay, O3DE, or Qt).

Project Properties

Skills

• C/C++
• OpenCL
• Intel Threading Building Blocks (TBB) (optional)
• OSPRay (optional)

Difficulty

hard

Additional Information

• Potential mentor(s): Vasco Costa
• Organization website: https://brlcad.org
• Communication channels: https://brlcad.zulipchat.com/

[ayushgupta9906]

I want to work on this issue. Please Assign this to me. Thanks