A physically based offline CPU pathtracer written in C++20

• Screenshots
• Requirements
• Building
• Features
• Dependencies

Screenshots

San Miguel - Rendered with 1024 SPP, 128 bounces

REQUIREMENTS

• C++20 compiler
• Cmake 3.22+ for building

Build

1. Cloning the project with submodules:

git clone --recurse-submodules https://github.com/marko176/PathTracing.git

2. Building the project:

cd PathTracing
mkdir build
cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
cmake --build ./build --config Release -j N   (set N to thread count)

3. Running the pathtracer do: ./Release/PathTracer

Output images will be stored in: Output/

Features:

• Camera with field of view and thin-lens approximation

• BVH2 using SAH-based binned building and simd during traversal and BVH4 based on BVH4, split into:

  • TLAS (Top-Level Acceleration Structure)
  • BLAS (Bottom-Level Acceleration Structure)

• Model Loading with Assimp

  • Supports obj and gltf formats
  • Supports model serialization for faster reloads in assbin format

• Physically Based Materials

  • Microfacet Dielectric
  • Microfacet Diffuse
  • Specular Dielectric
  • Specular Conductor

• Volumetric Rendering

  • Supports participating media (homogeneous)
  • Handles volumetric scattering and absorption within the path tracer
  • Works with MIS and NEE for efficient light transport through media

• Sampling

  • Cosine-weighted importance sampling for Lambertian BRDF
  • VNDF (visible normal distribution function) importance sampling for GGX specular BRDF
  • Multiple Importance Sampling (MIS) with Next Event Estimation (NEE) weighted with the power heuristic Veach (1997)
  • Russian Roulette path termination

• Texture Mapping

  • Albedo, Normal, Roughness, Metallic, Alpha, Emissive
  • support for HDR textures

Dependencies

• Assimp
• stb_image
• glm
• pcg