pywgpu

pywgpu is a high-performance Python port of the wgpu-rs project. It provides a modern, safe, and strictly-typed interface for graphics and compute on the GPU, following the WebGPU standard.

🏗 Architecture

The project is an exact structural copy of the original wgpu repository, divided into the following components:

• pywgpu-hal: Hardware Abstraction Layer. Implements low-level backends (Vulkan, Metal, DX12, GLES) via CFFI.
• pywgpu-core: The "brain" of the library. Handles resource management, state tracking, and validation.
• pywgpu: High-level, user-facing Pythonic API (equivalent to the wgpu crate).
• naga: A pure Python port of the Naga shader translator (WGSL/SPIR-V translation and validation).
• pywgpu-types: Shared data structures and descriptors powered by Pydantic V2.

🛠 Tech Stack

• Python 3.12+: Leveraging modern features like typing.Protocol and improved generics.
• Strict Static Typing: 100% type coverage. Verified with pyright in strict mode.
• Pydantic V2: Robust validation for all GPU descriptors and state objects.
• CFFI: High-performance interaction with native graphics APIs.
• NumPy: Efficient zero-copy data handling for buffers and textures.

📄 Documentation

• SPECIFICATION.md: Detailed technical requirements and architecture goals.
• AGENTS.md: Guidelines and standards for AI-driven development.

🚀 Status

The project is in a phase of active logic implementation, having moved significantly beyond the initial skeleton stage.

Key Achievements:

• ✓ Naga WGSL Frontend: Lowerer (AST to IR conversion) is fully complete.
• ✓ Constant Evaluator: Core infrastructure fully complete.
• ✓ Naga GLSL Frontend: Foundation (Types, Offset, Error, Token) complete.

📜 Principles

1. Architectural Identity: Every module mirrors its Rust counterpart.
2. Type Safety First: No Any, no implicit conversions.
3. Performance: Minimal overhead between Python and the GPU.