Raytracer

Figure 1: CPU render showcasing motion blur, BVH (AABB), texture mapping, Perlin noise, quad primitives, area lights, instancing, and constant-density media.

Figure 2: GPU render showcasing ray–sphere intersection, gamma correction, Lambertian, metallic (with fuzz), and dielectric materials (with total internal reflection and Schlick approximation), dynamic camera, and defocus blur.

Performance Comparison

Configuration	Value
Resolution	1200×675
Samples Per Pixel	500
Max Bounce Depth	50

Acceleration	Speedup	Time (hh:mm:ss)
None	1x	00::42::49::688 (2569.69 seconds)
CPU + BVH	5.67×	00:07:32.982 (452.983 seconds)
CUDA (no BVH)	10.44×	00:04:06.148 (246.148 seconds)

Features

This project explores building a Raytracer from scratch, with both CPU and CUDA-accelerated versions.

Feature	CPU Branch	CUDA Branch
Ray–Sphere Intersection	✅	✅
Gamma Correction	✅	✅
Lambertian Material	✅	✅
Metal (with fuzz)	✅	✅
Dielectric (TIR + Schlick)	✅	✅
Dynamic Camera	✅	✅
Defocus Blur	✅	✅
Motion Blur	✅	❌
Bounding Volume Hierarchy (AABB)	✅	❌
Texture Mapping	✅	❌
Perlin Noise	✅	❌
Quadrilateral Primitive (Ray–Plane Intersection)	✅	❌
Area Lights	✅	❌
Instance Translation/Rotation	✅	❌
Constant-Density Media	✅	❌

Technological Stack

C++ • CUDA • CMake

How to Run

• The project uses CMake as the meta build system.
• CUDA version >= 12.6 and MSVC are required for the cuda branch.

Check out the respective branch:

git checkout main    # If CPU
git checkout cuda    # If CUDA

Configure CMake

cmake -S . -B build

Release

cmake --build build --config Release

Debug

cmake --build build --config Debug