Matching Socks CUDA path tracer by Samuel Lapere, 2016 https://raytracey.blogspot.com
Demo of high performance CUDA accelerated path tracing based on the GPU ray tracing framework of Timo Aila, Samuli Laine and Tero Karras (Nvidia Research)
Screenshots and more info about this tutorial can be found at http://raytracey.blogspot.co.nz/2016/09/gpu-path-tracing-tutorial-4-optimised.html
Downloadable demo available at https://github.com/straaljager/GPU-path-tracing-tutorial-4/releases
Source code for original framework:
- https://code.google.com/archive/p/understanding-the-efficiency-of-ray-traversal-on-gpus/
- https://research.nvidia.com/publication/understanding-efficiency-ray-traversal-gpus-kepler-and-fermi-addendum
- https://mediatech.aalto.fi/~timo/HPG2009/
Features:
- fast interactive path tracing
- progressive rendering
- interactive camera and depth-of-field
- real-time user interaction
- BVH acceleration structure built with surface area heuristic (SAH) and spatial splits
- Woop ray/triangle intersection
- highly optimised hand tuned ray traversal and intersection kernels
- HDRI environment mapping
- anti-aliasing
- basic OBJ loader
- basic material system with support for diffuse, clearcoat, refractive (glass), specular(mirror) and metal material
- support for spheres and triangle meshes
Instructions for compiling with Visual Studio 2013/2015:
- download and install the CUDA 7.5 toolkit and choose integration with Visual Studio
- open VS2013/2015 (Express or any other version such as the free Community version)
- click New Project...
- select Visual C++, then General, then Empty Project
- right click on the project, select Build Dependies > Build Customizations, check CUDA 7.5, click OK
- add all the files to the project
- in the solution explorer, right click on "render_kernel.cu", select Properties, change the Item type to CUDA C/C++
- right click on the project name and select Properties
- under Linker > Input > Additional Dependencies, add "cudart.lib" and "glew32.lib" (glew32.lib should be automatically found when the CUDA toolkit is installed)
- when the compiler can't find glew32.lib, add the library path to Linker > General > Additional Library Directories, the path is something like "%NVSDKCOMPUTE_ROOT%\C\common\lib")
- disable SAFESEH by selecting NO in Linker > Advanced > Image Has Safe Exception Handlers
- select Build > Rebuild Solution and run the program
- IMPORTANT: build the code in Release mode (building in Debug mode will result in extremely slow render performance, in some cases more than 100x slower due to hundreds of CUDA processors logging state with extra debug information)
Issues:
- the OBJ loader is quite basic at the moment and only loads obj models that are in a particular format, hence some but not all OBJ files will work
- at the moment there is no CUDA error checking when copying data to the GPU. When CUDA tries to access data on the GPU that is missing/has not been copied to the GPU, the program may render a black screen instead.