|
| 1 | +use lru::LruCache; |
| 2 | +use memmap2::Mmap; |
| 3 | +use ordered_float::OrderedFloat; |
| 4 | +use std::num::NonZero; |
| 5 | +use tokio::fs::File; |
| 6 | +use tokio::sync::Mutex; |
| 7 | +use tokio::try_join; |
| 8 | + |
| 9 | +type NodeId = u32; |
| 10 | + |
| 11 | +#[derive(Debug)] |
| 12 | +pub struct DiskANNIndex { |
| 13 | + /// Memory-mapped vector storage |
| 14 | + vec_file: Mmap, |
| 15 | + |
| 16 | + /// Memory-mapped adjacency lists |
| 17 | + adj_file: Mmap, |
| 18 | + |
| 19 | + /// Cache for hot nodes: NodeId -> Vector + neighbors |
| 20 | + hot_cache: Mutex<LruCache<NodeId, CachedNode>>, |
| 21 | + |
| 22 | + /// Entry points for search |
| 23 | + entry_points: Vec<NodeId>, |
| 24 | + |
| 25 | + /// Number of neighbors per node |
| 26 | + max_neighbors: usize, |
| 27 | + |
| 28 | + /// Dimension of vectors |
| 29 | + dim: usize, |
| 30 | +} |
| 31 | + |
| 32 | +#[derive(Debug, Clone)] |
| 33 | +pub struct CachedNode { |
| 34 | + pub id: NodeId, |
| 35 | + pub vector: Vec<f32>, |
| 36 | + pub neighbors: Vec<NodeId>, |
| 37 | +} |
| 38 | + |
| 39 | +impl DiskANNIndex { |
| 40 | + pub async fn load( |
| 41 | + vec_path: &str, |
| 42 | + adj_path: &str, |
| 43 | + dim: usize, |
| 44 | + max_neighbors: usize, |
| 45 | + ) -> Result<Self, std::io::Error> { |
| 46 | + let (vec_file, adj_file) = try_join!(File::open(vec_path), File::open(adj_path))?; |
| 47 | + let vec_mmap = unsafe { Mmap::map(&vec_file)? }; |
| 48 | + let adj_mmap = unsafe { Mmap::map(&adj_file)? }; |
| 49 | + |
| 50 | + Ok(Self { |
| 51 | + vec_file: vec_mmap, |
| 52 | + adj_file: adj_mmap, |
| 53 | + hot_cache: Mutex::new(LruCache::new(NonZero::new(1024).unwrap())), |
| 54 | + entry_points: vec![0], |
| 55 | + max_neighbors, |
| 56 | + dim, |
| 57 | + }) |
| 58 | + } |
| 59 | + |
| 60 | + async fn load_node(&self, node_id: NodeId) -> CachedNode { |
| 61 | + if let Some(node) = self.hot_cache.lock().await.get(&node_id) { |
| 62 | + return node.clone(); |
| 63 | + } |
| 64 | + |
| 65 | + let start = (node_id as usize) * self.dim * 4; |
| 66 | + let vec_bytes = &self.vec_file[start..start + self.dim * 4]; |
| 67 | + let vector = vec_bytes |
| 68 | + .chunks_exact(4) |
| 69 | + .map(|b| f32::from_le_bytes(b.try_into().unwrap())) |
| 70 | + .collect(); |
| 71 | + |
| 72 | + let neighbors_start = (node_id as usize) * self.max_neighbors * 4; |
| 73 | + let neighbors_bytes = |
| 74 | + &self.adj_file[neighbors_start..neighbors_start + self.max_neighbors * 4]; |
| 75 | + let neighbors: Vec<NodeId> = neighbors_bytes |
| 76 | + .chunks_exact(4) |
| 77 | + .map(|b| u32::from_le_bytes(b.try_into().unwrap())) |
| 78 | + .collect(); |
| 79 | + |
| 80 | + let cached = CachedNode { |
| 81 | + id: node_id, |
| 82 | + vector, |
| 83 | + neighbors, |
| 84 | + }; |
| 85 | + self.hot_cache.lock().await.put(node_id, cached.clone()); |
| 86 | + cached |
| 87 | + } |
| 88 | + |
| 89 | + fn normalize(v: &[f32]) -> Vec<f32> { |
| 90 | + let norm = v.iter().map(|x| x * x).sum::<f32>().sqrt(); |
| 91 | + if norm == 0.0 { |
| 92 | + v.to_vec() // avoid division by zero |
| 93 | + } else { |
| 94 | + v.iter().map(|x| x / norm).collect() |
| 95 | + } |
| 96 | + } |
| 97 | + |
| 98 | + pub fn distance(a: &[f32], b: &[f32]) -> f32 { |
| 99 | + assert_eq!(a.len(), b.len(), "Vectors must have the same length"); |
| 100 | + |
| 101 | + let a_norm = Self::normalize(a); |
| 102 | + let b_norm = Self::normalize(b); |
| 103 | + |
| 104 | + a_norm |
| 105 | + .iter() |
| 106 | + .zip(b_norm.iter()) |
| 107 | + .map(|(&x, &y)| (x - y).powi(2)) |
| 108 | + .sum::<f32>() |
| 109 | + .sqrt() |
| 110 | + } |
| 111 | + |
| 112 | + pub async fn search( |
| 113 | + &self, |
| 114 | + query: &[f32], |
| 115 | + k: usize, |
| 116 | + beam_width: usize, |
| 117 | + ) -> Vec<(NodeId, OrderedFloat<f32>)> { |
| 118 | + use std::cmp::Reverse; |
| 119 | + use std::collections::{BinaryHeap, HashSet}; |
| 120 | + |
| 121 | + let mut candidates: BinaryHeap<Reverse<(OrderedFloat<f32>, NodeId)>> = BinaryHeap::new(); |
| 122 | + let mut visited: HashSet<NodeId> = HashSet::new(); |
| 123 | + let mut top_k: BinaryHeap<Reverse<(OrderedFloat<f32>, NodeId)>> = BinaryHeap::new(); |
| 124 | + |
| 125 | + for &ep in &self.entry_points { |
| 126 | + let node = self.load_node(ep).await; |
| 127 | + let dist = Self::distance(query, &node.vector); |
| 128 | + candidates.push(Reverse((OrderedFloat(dist), ep))); |
| 129 | + visited.insert(ep); |
| 130 | + } |
| 131 | + |
| 132 | + while let Some(Reverse((dist, node_id))) = candidates.pop() { |
| 133 | + let node = self.load_node(node_id).await; |
| 134 | + |
| 135 | + if top_k.len() < k { |
| 136 | + top_k.push(Reverse((dist, node_id))); |
| 137 | + } else if dist < top_k.peek().unwrap().0.0 { |
| 138 | + top_k.pop(); |
| 139 | + top_k.push(Reverse((dist, node_id))); |
| 140 | + } |
| 141 | + |
| 142 | + for &nbr in &node.neighbors { |
| 143 | + if !visited.contains(&nbr) { |
| 144 | + visited.insert(nbr); |
| 145 | + let nbr_node = self.load_node(nbr).await; |
| 146 | + let nbr_dist = Self::distance(query, &nbr_node.vector); |
| 147 | + candidates.push(Reverse((OrderedFloat(nbr_dist), nbr))); |
| 148 | + if candidates.len() > beam_width { |
| 149 | + candidates.pop(); |
| 150 | + } |
| 151 | + } |
| 152 | + } |
| 153 | + } |
| 154 | + |
| 155 | + let mut result: Vec<_> = top_k.into_iter().map(|Reverse((d, id))| (id, d)).collect(); |
| 156 | + result.sort_by(|a, b| a.1.partial_cmp(&b.1).unwrap()); |
| 157 | + result |
| 158 | + } |
| 159 | +} |
0 commit comments