ruvector_mincut/optimization/
mod.rs

1//! Performance Optimizations for j-Tree + BMSSP Implementation
2//!
3//! This module implements the SOTA optimizations from ADR-002-addendum-sota-optimizations.md:
4//!
5//! 1. **Degree-based presparse (DSpar)**: 5.9x speedup via effective resistance approximation
6//! 2. **LRU Cache**: Path distance caching with prefetch optimization
7//! 3. **SIMD Operations**: Vectorized distance array computations
8//! 4. **Pool Allocators**: Memory-efficient allocations with lazy deallocation
9//! 5. **Parallel Updates**: Rayon-based parallel level updates with work-stealing
10//! 6. **WASM Optimization**: Batch operations and TypedArray transfers
11//!
12//! Target: Combined 10x speedup over naive implementation.
13
14pub mod benchmark;
15pub mod cache;
16pub mod dspar;
17pub mod parallel;
18pub mod pool;
19pub mod simd_distance;
20pub mod wasm_batch;
21
22// Re-exports
23pub use benchmark::{BenchmarkResult, BenchmarkSuite, OptimizationBenchmark};
24pub use cache::{CacheConfig, CacheStats, PathDistanceCache, PrefetchHint};
25pub use dspar::{DegreePresparse, PresparseConfig, PresparseResult, PresparseStats};
26pub use parallel::{ParallelConfig, ParallelLevelUpdater, WorkStealingScheduler};
27pub use pool::{LazyLevel, LevelPool, PoolConfig, PoolStats};
28pub use simd_distance::{DistanceArray, SimdDistanceOps};
29pub use wasm_batch::{BatchConfig, TypedArrayTransfer, WasmBatchOps};
ruvector_mincut/optimization/mod.rs

ruvector_mincut/optimization/
mod.rs
