Expand description
Multi-Backend Abstraction
Provides a unified interface for different GPU backends:
- CUDA (NVIDIA) - Primary, uses PTX
- WGPU (WebGPU) - Cross-platform, uses WGSL (Vulkan/Metal/DX12/WebGPU)
- Metal (Apple) - Native Apple GPU compute via manzana crate
- Vulkan (cross-platform, future)
Structs§
- Cuda
Backend - CUDA backend (NVIDIA GPUs)
- Metal
Backend - Metal backend (Apple GPUs)
- Vulkan
Backend - Vulkan backend (cross-platform) - placeholder
- Wgpu
Backend - WGPU backend (WebGPU - cross-platform via wgpu crate)
Traits§
- Backend
- Backend trait for GPU operations
Functions§
- detect_
backend - Detect best available backend