Struct BvhGpuTraverser 

pub struct BvhGpuTraverser { /* private fields */ }

GPU-accelerated BVH ray traversal.

Encodes a flat BVH into GPU-resident buffers and dispatches the bvh_traversal.wgsl kernel. Falls back to CPU traversal when no GPU adapter is available.

Usage

use oxiphysics_gpu::bvh::{Aabb, Bvh, BvhPrimitive, BvhGpuTraverser, GpuRay};

let prims = vec![
    BvhPrimitive::new(Aabb::new([0.0, 0.0, 0.0], [1.0, 1.0, 1.0]), 0),
];
let bvh = Bvh::build(prims);
let traverser = BvhGpuTraverser::new(&bvh);
let rays = vec![GpuRay::new([0.5, 0.5, -1.0], [0.0, 0.0, 1.0], 100.0)];
let hits = traverser.traverse_rays(&rays);
assert_eq!(hits.len(), 1);

Implementations

impl BvhGpuTraverser

pub fn new(bvh: &Bvh) -> Self

Create a traverser from a BVH.

Flattens the BVH and uploads the node + primitive buffers to the GPU. Falls back to CPU traversal if no GPU adapter is found.

pub fn new_cpu(bvh: &Bvh) -> Self

Create a CPU-only traverser (useful for testing without a GPU).

pub fn is_gpu(&self) -> bool

Returns true if using a real GPU backend.

pub fn dispatch_count(&self) -> u64

Returns the current dispatch count (GPU variant only; always 0 for CPU).

pub fn creation_id(&self) -> Option<u64>

Returns the creation_id of the underlying GPU state (for reuse tests).

pub fn traverse_rays(&self, rays: &[GpuRay]) -> Vec<i32>

Traverse the BVH for each ray.

Returns a Vec<i32> of length rays.len(). Each element is either:

• the object_id of the first hit leaf’s primitive, or
• -1 if no intersection was found.