CIS565-Fall-2016 · krupkad · Oct 11, 2016 · Oct 13, 2016 · Oct 13, 2016 · Oct 13, 2016
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -36,6 +36,7 @@ set(CORELIBS
     "${GLFW_LIBRARY}"
     "${OPENGL_LIBRARY}"
     "${GLEW_LIBRARY}"
+    pthread
     )
 
 # Enable C++11 for host code
@@ -68,17 +69,16 @@ if(${CMAKE_SYSTEM_NAME} MATCHES "Darwin")
 endif()
 
 include_directories(.)
-#add_subdirectory(stream_compaction)  # TODO: uncomment if using your stream compaction
+add_subdirectory(stream_compaction)  # TODO: uncomment if using your stream compaction
 add_subdirectory(src)
 
 cuda_add_executable(${CMAKE_PROJECT_NAME}
-    "src/main.h"
     "src/main.cpp"
     )
 
 target_link_libraries(${CMAKE_PROJECT_NAME}
     src
-    #stream_compaction  # TODO: uncomment if using your stream compaction
+    stream_compaction  # TODO: uncomment if using your stream compaction
     ${CORELIBS}
     )
 

diff --git a/README.md b/README.md
@@ -1,13 +1,54 @@
-CUDA Path Tracer
-================
+CUDA Stream Compaction
+======================
 
-**University of Pennsylvania, CIS 565: GPU Programming and Architecture, Project 3**
+**University of Pennsylvania, CIS 565: GPU Programming and Architecture, Project 2**
 
-* (TODO) YOUR NAME HERE
-* Tested on: (TODO) Windows 22, i7-2222 @ 2.22GHz 22GB, GTX 222 222MB (Moore 2222 Lab)
+* Daniel Krupka
+* Tested on: Debian testing (stretch), Intel(R) Core(TM) i7-4710HQ CPU @ 2.50GHz 8GB, GTX 850M
 
-### (TODO: Your README)
+# Project 3 - GPU Path Tracing
+My goal for this project was to extend my existing path tracer to use CUDA. All code is hand-written or
+adapted from given CIS 565 code -- no external libraries used.
 
-*DO NOT* leave the README to the last minute! It is a crucial part of the
-project, and we will not be able to grade you without a good README.
+# Features
+The primary performance features implemented are
+1. Stream Compaction - Rays that are no longer alive (e.g. have hit lights, escaped the scene)
+are culled.
+2. Bounce Caching - Rays' first intersection are constant between iterations, and thus can be cached for some time savings.
+3. Sort-by-material - As objects with similar materials can be expected to take similar code paths during shading, it may be
+advantageous to sort rays by the material they've hit prior to shading.
+
+Additionally, I implemented some extra features and improvements.
+1. Fresnel scattering - The exact Fresnel equations for reflection/refraction are solved.
+2. Depth of field - Physically accurate depth of field.
+3. Arbitrary meshes - .obj files can be loaded, with optional AABB culling.
+
+# Results
+
+![Cornell Box](img/cornell.png "Cornell Box")
+
+## Base features
+The CUDA-accelerated path tracer performed much faster than the CPU version, as expected. Though I did not run CPU tests,
+similar scenes generally took minutes where the GPU version takes seconds.
+
+
+![First-Bounce caching comparison](img/cache.png "First-Bounce caching comparison")
+
+As expected, caching the first ray intersection did produce a performance gain, but this diminished as the maximum
+number of bounces increased. These tests were conducted on the Cornell box seen above.
+
+## Extra features
+![DOF](img/dof.png "Depth of Field")
+
+Depth of field is physically realistic, and achieved by a "DENSITY N" and "APERTURE R" parameters in the CAMERA section, which
+causes the ray origins to be jittered N^2 times within an aperture of radius R. This leads to a slow-down equivalent to
+having N^2 times as many rays.
+
+![Cow](img/cow.png "Arbitrary meshes")
+![Teapot](img/teapot.png "Arbitrary meshes")
+
+The tracer supports loading arbitrary meshes from a standard .obj file, by setting "SHAPE mesh"
+and FILE file.obj" in an OBJECT section. AABB culling can be enabled/disabled by "BBCULL 1/0".
+For the teapot, the non-culled render took 118 seconds, the culled 102 seconds The cow, a more complex model,
+took 635 seconds non-culled, but only 593 seconds culled.
 
diff --git a/img/REFERENCE_cornell.5000samp.png b/img/REFERENCE_cornell.5000samp.png
diff --git a/img/cache.png b/img/cache.png
diff --git a/img/cornell.png b/img/cornell.png
diff --git a/img/cow.png b/img/cow.png
diff --git a/img/dof.png b/img/dof.png
diff --git a/img/glitch1.png b/img/glitch1.png
diff --git a/img/glitch2.png b/img/glitch2.png
diff --git a/img/glitch3.png b/img/glitch3.png
diff --git a/img/stacked_bar_graph.png b/img/stacked_bar_graph.png
diff --git a/img/teapot.png b/img/teapot.png
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
@@ -1,23 +1,38 @@
 set(SOURCE_FILES
-    "stb.cpp"
-    "image.cpp"
-    "image.h"
-    "interactions.h"
-    "intersections.h"
-    "glslUtility.hpp"
-    "glslUtility.cpp"
-    "pathtrace.cu"
-    "pathtrace.h"
+    "camera.cpp"
+    "camera.hpp"
+    "scene.hpp"
     "scene.cpp"
-    "scene.h"
-    "sceneStructs.h"
-    "preview.h"
-    "preview.cpp"
-    "utilities.cpp"
-    "utilities.h"
+    "conf.hpp"
+    "conf.cpp"
+    "util.hpp"
+    "kdtree.cpp"
+    "kdtree.hpp"
+    "light.hpp"
+    "node.cpp"
+    "node.hpp"
+    "shader.cpp"
+    "shader.hpp"
+    "texture.hpp"
+    "texture.cpp"
+    "mesh/intersection.hpp"
+    "mesh/intersection.cu"
+    "mesh/object.cpp"
+    "mesh/object.hpp"
+    "mesh/cube.cpp"
+    "mesh/cube.hpp"
+    "mesh/sphere.cpp"
+    "mesh/sphere.hpp"
+    "mesh/cylinder.cpp"
+    "mesh/cylinder.hpp"
+    "mesh/mesh.cpp"
+    "mesh/mesh.hpp"
+    "cuda_camera.cu"
+    "util_cuda.cu"
+    "stb.cpp"
     )
 
 cuda_add_library(src
     ${SOURCE_FILES}
-    OPTIONS -arch=sm_20
+    OPTIONS -arch=sm_50 -std=c++11
     )