edgevec 0.9.0

High-performance embedded vector database for Browser, Node, and Edge
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247

//! SIMD vs Scalar Comparison Benchmark (Week 30 Day 2)
//!
//! Run with: `cargo bench --bench simd_comparison`
//!
//! This benchmark validates the 2-3x speedup target for SIMD operations.
//!
//! # Target Metrics
//!
//! | Metric | Scalar (Baseline) | SIMD Target | Improvement |
//! |:-------|:------------------|:------------|:------------|
//! | Dot Product (768-dim) | ~500ns | <200ns | 2.5x |
//! | L2 Distance (768-dim) | ~600ns | <250ns | 2.4x |
//! | Search (10k, k=10) | ~5ms | ~2ms | 2.5x |
//! | Hamming Distance (1024-bit) | ~100ns | <40ns | 2.5x |

use criterion::{criterion_group, criterion_main, BenchmarkId, Criterion, Throughput};
use edgevec::hnsw::{HnswConfig, HnswIndex, SearchContext};
use edgevec::metric::{DotProduct, L2Squared, Metric};
use edgevec::quantization::binary::QuantizedVector;
use edgevec::storage::VectorStorage;
use rand::{Rng, SeedableRng};
use rand_chacha::ChaCha8Rng;
use std::hint::black_box;

// ============================================================================
// HELPER FUNCTIONS
// ============================================================================

fn generate_random_vector(dim: usize, seed: u64) -> Vec<f32> {
    let mut rng = ChaCha8Rng::seed_from_u64(seed);
    (0..dim).map(|_| rng.gen_range(-1.0..1.0)).collect()
}

fn generate_vectors(count: usize, dim: usize, seed: u64) -> Vec<Vec<f32>> {
    (0..count)
        .map(|i| generate_random_vector(dim, seed + i as u64))
        .collect()
}

fn generate_quantized_vector(seed: u64) -> QuantizedVector {
    let mut rng = ChaCha8Rng::seed_from_u64(seed);
    let mut data = [0u8; 96];
    for byte in &mut data {
        *byte = rng.gen();
    }
    QuantizedVector::from_bytes(data)
}

// ============================================================================
// DOT PRODUCT BENCHMARK (Multiple Dimensions)
// ============================================================================

fn bench_dot_product(c: &mut Criterion) {
    let mut group = c.benchmark_group("dot_product");

    for dim in [128, 256, 384, 512, 768, 1024, 1536] {
        let a = generate_random_vector(dim, 42);
        let b = generate_random_vector(dim, 43);

        group.throughput(Throughput::Elements(dim as u64));

        group.bench_with_input(BenchmarkId::new("simd", dim), &dim, |bench, _| {
            bench.iter(|| DotProduct::distance(black_box(&a), black_box(&b)))
        });
    }
    group.finish();
}

// ============================================================================
// L2 DISTANCE BENCHMARK (Multiple Dimensions)
// ============================================================================

fn bench_l2_squared(c: &mut Criterion) {
    let mut group = c.benchmark_group("l2_squared");

    for dim in [128, 256, 384, 512, 768, 1024, 1536] {
        let a = generate_random_vector(dim, 42);
        let b = generate_random_vector(dim, 43);

        group.throughput(Throughput::Elements(dim as u64));

        group.bench_with_input(BenchmarkId::new("simd", dim), &dim, |bench, _| {
            bench.iter(|| L2Squared::distance(black_box(&a), black_box(&b)))
        });
    }
    group.finish();
}

// ============================================================================
// COSINE SIMILARITY BENCHMARK (Multiple Dimensions)
// ============================================================================

fn bench_cosine(c: &mut Criterion) {
    let mut group = c.benchmark_group("cosine_similarity");

    for dim in [128, 256, 384, 512, 768, 1024, 1536] {
        let a = generate_random_vector(dim, 42);
        let b = generate_random_vector(dim, 43);

        // Cosine = 1 - (dot / (|a| * |b|))
        // We'll just measure dot product since cosine uses it internally
        let a_norm: f32 = a.iter().map(|x| x * x).sum::<f32>().sqrt();
        let b_norm: f32 = b.iter().map(|x| x * x).sum::<f32>().sqrt();

        group.throughput(Throughput::Elements(dim as u64));

        group.bench_with_input(BenchmarkId::new("simd", dim), &dim, |bench, _| {
            bench.iter(|| {
                let dot = DotProduct::distance(black_box(&a), black_box(&b));
                // Note: DotProduct returns negative dot product for HNSW,
                // so cosine similarity = 1 + dot/(a_norm*b_norm)
                let cosine = 1.0 + dot / (a_norm * b_norm);
                black_box(cosine)
            })
        });
    }
    group.finish();
}

// ============================================================================
// HAMMING DISTANCE BENCHMARK (Binary Quantization)
// ============================================================================

fn bench_hamming(c: &mut Criterion) {
    let mut group = c.benchmark_group("hamming_distance");

    // 768-bit = 96 bytes (standard QuantizedVector size)
    let q1 = generate_quantized_vector(42);
    let q2 = generate_quantized_vector(43);

    group.throughput(Throughput::Bytes(96 * 2)); // 2 × 96-byte inputs

    group.bench_function("768bit", |b| {
        b.iter(|| black_box(&q1).hamming_distance(black_box(&q2)))
    });

    group.finish();
}

// ============================================================================
// SEARCH BENCHMARK (Multiple Collection Sizes)
// ============================================================================

fn bench_search(c: &mut Criterion) {
    let mut group = c.benchmark_group("search");
    group.sample_size(50); // Fewer samples for slower benchmarks

    for count in [1_000, 10_000] {
        let dims = 768;
        let seed = 42u64;

        // Build index
        let vectors = generate_vectors(count, dims, seed);
        let config = HnswConfig::new(dims as u32);
        let mut storage = VectorStorage::new(&config, None);
        let mut index = HnswIndex::new(config, &storage).unwrap();

        for v in &vectors {
            index.insert(v, &mut storage).unwrap();
        }

        let query = generate_random_vector(dims, 999);

        group.throughput(Throughput::Elements(1));

        group.bench_with_input(BenchmarkId::new("k=10", count), &count, |bench, _| {
            let mut ctx = SearchContext::new();
            bench.iter(|| {
                index
                    .search_with_context(black_box(&query), 10, &storage, &mut ctx)
                    .unwrap()
            })
        });
    }
    group.finish();
}

// ============================================================================
// DISTANCE BATCH BENCHMARK (10k distance calculations)
// ============================================================================

fn bench_distance_batch(c: &mut Criterion) {
    let mut group = c.benchmark_group("distance_batch_10k");

    for dim in [128, 384, 768] {
        let query = generate_random_vector(dim, 42);
        let vectors = generate_vectors(10_000, dim, 100);

        group.throughput(Throughput::Elements(10_000));

        // L2 Squared
        group.bench_with_input(BenchmarkId::new("l2", dim), &dim, |bench, _| {
            bench.iter(|| {
                for v in &vectors {
                    black_box(L2Squared::distance(black_box(&query), black_box(v)));
                }
            })
        });

        // Dot Product
        group.bench_with_input(BenchmarkId::new("dot", dim), &dim, |bench, _| {
            bench.iter(|| {
                for v in &vectors {
                    black_box(DotProduct::distance(black_box(&query), black_box(v)));
                }
            })
        });
    }
    group.finish();
}

// ============================================================================
// HAMMING BATCH BENCHMARK (10k distance calculations)
// ============================================================================

fn bench_hamming_batch(c: &mut Criterion) {
    let mut group = c.benchmark_group("hamming_batch_10k");

    let query = generate_quantized_vector(42);
    let vectors: Vec<QuantizedVector> = (0..10_000)
        .map(|i| generate_quantized_vector(100 + i))
        .collect();

    group.throughput(Throughput::Elements(10_000));

    group.bench_function("768bit", |b| {
        b.iter(|| {
            for v in &vectors {
                black_box(black_box(&query).hamming_distance(black_box(v)));
            }
        })
    });

    group.finish();
}

criterion_group!(
    benches,
    bench_dot_product,
    bench_l2_squared,
    bench_cosine,
    bench_hamming,
    bench_search,
    bench_distance_batch,
    bench_hamming_batch
);
criterion_main!(benches);