The purpose of these code examples is to show how to use ray tracing for computations and discuss some application patterns that could benefit from ray tracing frameworks. These samples do not perform any graphical rendering. The individual samples are described below.
Prior knowledge of basic ray tracing terminology is required. However, in-depth knowledge and experience with ray tracing is not needed. Please refer to the OptiX ray tracing docs: https://raytracing-docs.nvidia.com/optix7/index.html
The following samples are based on the OptiX 7 API. Old OptiX 6 based samples are in legacy-optix-6 branch in this repository and will no longer be maintained.
CUDA/Optix buffer interop.
This sample shows how to work with CUDA allocated memory buffers and OptiX in order to compute a simple saxpy operation in the ray generation program. There are no rays traced in the code and no geometry is created. This example is useful to understand the OptiX API and code structure.
CAD geometry/cartesian mesh mapping.
This sample shows a 2D projection of a 3D mesh onto 6 planes along the coordinate axes. To compute individual projections, the program shoots rays at a fixed sampling interval (dx,dy) perpendicular to the plane of projection. Projections are also used to find the distance to the 3D mesh object.
Cartesian mesh refinement.
This sample shows how to identify cells that need refinement in a 3D structured mesh. We start with a coarse 3D structured grid overlapping a triangle geometry. Rays are used to identify cells that contain geometry and are tagged for refinement. This process can be repeated for multi stage mesh refinement.
Scattering rays by reflection or refraction.
This sample shows how to generate rays and bounce them across different surfaces. Rays are reflected between different surfaces and the rays keep bouncing until max_bounce is reached. Ray energy can be damped similar to signal propagation, rays can be refracted, and energy can be deposited at different hit points. This is a typical use case in wave propagation.
Scattering rays by reflection or refraction with instancing.
This sample shows how to generate rays and bounce them across different surfaces. Rays are reflected between different surfaces and the rays keep bouncing until max_bounce is reached. Ray energy can be damped similar to signal propagation, rays can be refracted, and energy can be deposited at different hit points. This is a typical use case in wave propagation. Instancing of the geometry is used to replicate the surface.
Scattering rays by reflection, refraction and with absorption.
This sample shows how to generate rays and bounce them across different surfaces. Rays are reflected between different surfaces and the rays keep bouncing until max_bounce is reached. Ray energy can be damped similar to signal propagation, rays can be refracted, and energy can be deposited at different hit points. Rays can also be absorb depending on the material of the object, this is accomplished with different build inputs and their respective hit group. This is a typical use case in wave propagation.
Volume sampling techniques.
This sample reads a simple 3D volume. The purpose here is to collect values when a ray hits a geometry along the direction of the ray. Two approaches are shown in this example. In the first approach, a value is accumulated for every hit point the ray encounters, which can be specific to that triangle or geometry primitive. In the second approach, rays of different lengths are shot from the same origin and then do regular sampling of the 3D volume.
Detecting Particle Collisions with Geometry
This sample reads a simple obj file to create a geometry. It then creates a list of particles with positions inside or around the geomtery and random velocity values. Using particle positions as the center of a ray and direction along the velocity, rays are then shot with infinite length. If a ray hits some geometry, then the distance between the position of the particle and geometry is recorded.
Doing Visibility Test around a Point of Interest in 2D space
This sample create N random triangles in 2D plane between a user provided min and max value. After that it raises the 2D polygons to 3D by adding a small delta in Z dimension. This is required to build Optix BVH. Hence each edge of a triangle in 2D gets converted to 2 triangles in 3D. The program then randomly selects M center point in the above 2D plane and fires 360 rays per each sample point i.e., 1 degree sampling. If a ray hits some geometry the distance between the ray origin and the hit point is store. The user can modify the code to do other calculation upon hits. This program can be extended to do polygon visibility test in 3D as well.
Use CMake (>=3.5) for building. Requires CUDA 8.0 or higher and Optix 7.1 or higher
mkdir build && cd build
cmake ../ -DOPTIX_HOME=<path to OptiX 7>
make
The executables require runtime files, which include pre-compiled PTX files and .obj mesh files. These are copied by the build system into each example's binary directory. For example:
cd optixRayScattering/
# Note the `optixRayScatteringPrograms.ptx` and `sphereAndPlane.obj` files.
cd optixRayScattering/
./rayscattering