rayt GEMINI.md

Project Overview

This project is a ray tracer implementation based on Peter Shirley's "Ray Tracing in One Weekend" book series. It's a hybrid Python and Rust application, with the performance-critical ray tracing logic written in Rust and exposed as a Python extension using pyo3 and maturin. The Python part of the project provides a command-line interface (CLI) using the click library and leverages numba for CPU-based performance optimization and numba-cuda for GPU acceleration.

The project is structured to allow rendering on different backends:

• Numba (CPU): Utilizes Numba's JIT compilation to accelerate the Python code on the CPU.
• CUDA (GPU): Offloads the rendering process to NVIDIA GPUs using Numba's CUDA support.
• Rust: The core ray tracing logic, including vector operations, hittable objects, materials, and camera, is implemented in Rust for maximum performance.

Building and Running

The project uses uv as a package manager. The following commands are available to build and run the ray tracer:

Basic Rendering (CPU)

uv run one-weekend --image-width=300 --samples-per-pixel=20 > image.ppm

High-Quality Rendering (CPU)

uv run one-weekend --image-width=1200 --samples-per-pixel=100 > image.ppm

GPU Rendering

To use the GPU-accelerated renderer, you need a CUDA-enabled NVIDIA GPU and the necessary drivers.

uv run one-weekend --image-width=300 --samples-per-pixel=20 --engine=gpu > image.ppm

Testing GPU Availability

You can check if a compatible GPU is available with the following command:

uv run test-gpu

Development Conventions

• Hybrid Python/Rust: The project follows a hybrid development model. The core, performance-sensitive code is in Rust (src/), while the CLI, rendering orchestration, and optimization bindings are in Python (src/rayt/).
• Build System: maturin is used to build the Rust code into a Python extension. The configuration is in pyproject.toml.
• Dependencies: Python dependencies are managed with uv and are listed in pyproject.toml. Rust dependencies are managed with cargo and are listed in Cargo.toml.
• CLI: The command-line interface is built with click and is defined in src/rayt/cli.py.
• Code Style: The Python code follows standard Python conventions. The Rust code follows standard Rust conventions.