Struct GpuGeometricNetwork 

pub struct GpuGeometricNetwork { /* private fields */ }

GPU-accelerated geometric network analysis

Implementations

impl GpuGeometricNetwork

pub async fn new() -> GpuNetworkResult<Self>

Initialize GPU context for network analysis

pub async fn compute_all_pairwise_distances<const P: usize, const Q: usize, const R: usize>( &self, network: &GeometricNetwork<P, Q, R>, ) -> GpuNetworkResult<Vec<Vec<f64>>>

Compute all pairwise Euclidean distances using GPU acceleration.

This GPU path supports vector-only Cl(P,0,0) embeddings with P <= 3. Use AdaptiveNetworkCompute for automatic CPU fallback on unsupported signatures or non-vector multivectors.

pub async fn compute_geometric_centrality<const P: usize, const Q: usize, const R: usize>( &self, network: &GeometricNetwork<P, Q, R>, ) -> GpuNetworkResult<Vec<f64>>

Compute geometric centrality using GPU distances plus CPU reduction.

pub async fn geometric_clustering<const P: usize, const Q: usize, const R: usize>( &self, network: &GeometricNetwork<P, Q, R>, k: usize, max_iterations: usize, ) -> GpuNetworkResult<Vec<Community<P, Q, R>>>

GPU-distance-assisted k-medoids clustering for community detection.

Pairwise distances are computed by the GPU path; assignment, medoid updates, and cohesion scoring are CPU-side for 0.20.0.

pub fn should_use_gpu(num_nodes: usize) -> bool

Determine if GPU acceleration should be used based on network size.

pub fn supports_gpu_distance<const P: usize, const Q: usize, const R: usize>() -> bool

Return whether this signature can use the current GPU distance kernel.