CUDA ray tracer for GPGPU course
--------------------------------

Hannu Saransaari
hsaransa at cc dot hut dot fi
64647L
26.4.2010

Introduction
------------

User interface is pretty intuitive, you can move camera using arrow keys and
rotate camera by dragging mouse while pressing the left mouse button.

Rendering mode can be changed by pressing numbers 1, 2 and 3. 1 means OpenGL
rendering, 2 means CPU ray tracer and 3 means CUDA ray tracer. Note that CPU
ray tracer is unoptimized and even a single frame takes several seconds to
render.

To ease comparing performance, button 'u' can be used to save the camera and
'p' to load previously saved camera. camera.txt in the package is for
conference room.

Brief guide to code
-------------------

The code includes a lot of stuff that has very little to do with CUDA ray
tracing. This is because the ray tracer was originally implemented in my
another project.

main.cpp
This file includes main(). It also loads 3d object, cuda module, renders stuff
and handles user interface.

bvhrt.cpp and bvhrt.hpp
These files build BVH tree using greedy top-down surface area heuristic.

cudabvh.cpp and cudabvh.hpp
These files are used to convert bvh tree to arrays used by CUDA ray tracer.

cudabvh.cu and cudavec.h
BVH traversal kernel in cuda.

zorder.cpp and zorder.hpp
These files generate Z-order permutation tables.

objloader.cpp and objloader.hpp
These load the standard Autodeks .obj file.

Misc
----

bunny.obj is the famous Stanford Bunny. http://graphics.stanford.edu/data/3Dscanrep/

conference.obj is made by Anat Grynberg and Greg Ward. http://radsite.lbl.gov/mgf/scenes.html

The optimizations in the kernel are based on the paper written by Timo Aila and
Samuli Laine. http://www.tml.tkk.fi/~timo/