This CUDA-Accelerated Path Tracer was written by Tudor Matei Boran originally for CS-GA: 3801 - GPU Programming for the Fall 2018 semester at NYU Courant

It continued on as a personal project where I added many more features and optimizations.

To compile, you need CMake and libraries for OpenGL, GLFW, and glew3 installed. Conditional compilation will be included in a future release where CMake will compile a command-line only version of this program without these dependencies.

Clone this repository, and cd CudaRT

Create a folder called build with mkdir build and cd build

Then run cmake .. and make all

To run this program, the following pattern should be used while still in the CudaRT/build directory ./CudaRT <parameters>

CudaRT comes with a Maya plugin, whose source code can be found at https://github.com/mattboran/CudaRT-Maya-Plugin

Ensure that you have the .env file in the parent directory containing the paths:

MESHES_PATH: the scene to be loaded references <scene>.obj and <scene>.mtl Note that the .mtl files used are different. They will be named .cmtl at some point soon because they contain ray-tracing specific material properties that are exported from Maya

CAMERA_PATH: the directory that contains the <scene>-camera.json file describing the camera, also exported from Maya.

TEXTURES_PATH: the directory containing the textures referenced by <scene>.mtl

Required Parameter: -f <scene>: the scene that should be loaded Optional Parameters: -s <samples per pixel>: number of samples per pixel. Default = 10. reccommendation is to use at least 128, but 512 generates a fairly variance/noise-free image

-w <width (pixels) of output>

-h <height (pixels) of output>

--cpu: flag to use sequential CPU version instead of parallel GPU version

--x: flag to use OpenGL to render to window. Saves to output file on close.