Module index_builder 

GPU-accelerated HNSW index construction

This module implements GPU-based HNSW graph construction using CUDA kernels (feature-gated behind the cuda feature). When CUDA is not available, it falls back to an efficient CPU implementation.

Architecture

The GPU index builder works in phases:

1. Vector Upload: Transfer vectors to GPU memory in batches
2. Distance Matrix Computation: Compute all-pairs distances via CUDA kernels
3. Neighbor Selection: Apply heuristic neighbor selection on GPU
4. Graph Assembly: Assemble the HNSW graph structure from GPU results

Pure Rust Policy

All CUDA code is gated with #[cfg(feature = "cuda")] so the default build is 100% Pure Rust.

Module Layout

• crate::gpu::index_builder_types: Config types, distance metrics, graph structures
• crate::gpu::index_builder_phases: Builder implementations and pipeline stages

Re-exports

pub use crate::gpu::index_builder_phases::BatchSizeCalculator;

pub use crate::gpu::index_builder_phases::ComputedBatch;

pub use crate::gpu::index_builder_phases::GpuBatchDistanceComputer;

pub use crate::gpu::index_builder_phases::GpuHnswIndexBuilder;

pub use crate::gpu::index_builder_phases::GpuIndexOptimizer;

pub use crate::gpu::index_builder_phases::GpuMemoryBudget;

pub use crate::gpu::index_builder_phases::IncrementalGpuIndexBuilder;

pub use crate::gpu::index_builder_phases::IndexedBatch;

pub use crate::gpu::index_builder_phases::PipelinedIndexBuilder;

pub use crate::gpu::index_builder_phases::PreparedBatch;

pub use crate::gpu::index_builder_types::ComputationCache;

pub use crate::gpu::index_builder_types::GpuDistanceMetric;

pub use crate::gpu::index_builder_types::GpuIndexBuildStats;

pub use crate::gpu::index_builder_types::GpuIndexBuilderConfig;

pub use crate::gpu::index_builder_types::HnswGraph;

pub use crate::gpu::index_builder_types::HnswNode;