oxiphysics-gpu 0.1.0

GPU acceleration backends for the OxiPhysics engine
Documentation

oxiphysics-gpu

Status: [Partial] — v0.1.0 (2026-04-06)

Tests docs.rs

GPU-accelerated compute abstractions for the OxiPhysics engine.

Note: v0.1.0 ships a CPU backend only. No wgpu or CUDA dependencies are present in this release. GPU dispatch (wgpu/CUDA) is planned for v0.2.0.

Features

  • Compute abstraction: ComputeBackend trait + CpuBackend implementation
  • Kernel dispatch: ComputeKernel, BufferHandle, DispatchTimer; dispatch_count, aligned_size, linear_index_3d utilities
  • Particle system: ParticleSystem — position/velocity buffers, neighbor queries
  • Spatial acceleration: BVH (bounding volume hierarchy), cell-list neighbor search, SDF compute (sdf_compute)
  • Parallel primitives: parallel sort (parallel_sort), grid reduction (grid_reduce), flux compute (flux_compute)
  • Sparse GPU: sparse matrix/vector operations on CPU (sparse_gpu)
  • Pipeline: compute pipeline management (compute_pipeline, pipeline), shader registry (shader_registry, shaders)
  • Neural compute: neural network forward-pass compute kernels (neural_compute)
  • 2,748 public API items, 2,740 tests — 0 stubs

Key Exports

use oxiphysics_gpu::{
    ComputeBackend, CpuBackend, ComputeKernel, BufferHandle,
    dispatch_count, aligned_size, linear_index_3d,
    DispatchTimer, ParticleSystem,
};

Roadmap

Version Feature
0.1.0 CPU backend (current)
0.2.0 wgpu backend (planned)
0.3.0 CUDA backend via cudarc (planned)

License

Apache-2.0 — Part of the OxiPhysics project by COOLJAPAN OU (Team Kitasan)