innr 0.2.2

SIMD-accelerated vector similarity primitives with binary, ternary, and scalar quantization
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

#![allow(missing_docs)]
//! Benchmarks for batch vector operations (PDX-style columnar layout).

use criterion::{black_box, criterion_group, criterion_main, BenchmarkId, Criterion, Throughput};
use innr::batch::{
    batch_cosine, batch_dot, batch_knn, batch_knn_cosine, batch_l2_squared, batch_norms,
    VerticalBatch,
};
use rand::prelude::*;

fn random_vecs(n: usize, dim: usize) -> Vec<Vec<f32>> {
    let mut rng = StdRng::seed_from_u64(42);
    (0..n)
        .map(|_| (0..dim).map(|_| rng.random_range(-1.0..1.0)).collect())
        .collect()
}

fn random_vec(dim: usize) -> Vec<f32> {
    let mut rng = StdRng::seed_from_u64(99);
    (0..dim).map(|_| rng.random_range(-1.0..1.0)).collect()
}

fn bench_batch_l2(c: &mut Criterion) {
    let mut group = c.benchmark_group("batch_l2");

    for (n, dim) in [(1000, 128), (1000, 768), (10000, 128), (10000, 768)] {
        let vectors = random_vecs(n, dim);
        let batch = VerticalBatch::from_rows(&vectors);
        let query = random_vec(dim);

        group.throughput(Throughput::Elements((n * dim) as u64));
        group.bench_with_input(
            BenchmarkId::new("l2_squared", format!("{n}x{dim}")),
            &(),
            |bench, _| bench.iter(|| batch_l2_squared(black_box(&query), black_box(&batch))),
        );
    }

    group.finish();
}

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

    for (n, dim) in [(1000, 768), (10000, 768)] {
        let vectors = random_vecs(n, dim);
        let batch = VerticalBatch::from_rows(&vectors);
        let query = random_vec(dim);

        group.throughput(Throughput::Elements((n * dim) as u64));
        group.bench_with_input(
            BenchmarkId::new("dot", format!("{n}x{dim}")),
            &(),
            |bench, _| bench.iter(|| batch_dot(black_box(&query), black_box(&batch))),
        );
    }

    group.finish();
}

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

    for (n, dim) in [(1000, 768), (10000, 768)] {
        let vectors = random_vecs(n, dim);
        let batch = VerticalBatch::from_rows(&vectors);
        let norms = batch_norms(&batch);
        let query = random_vec(dim);

        group.throughput(Throughput::Elements((n * dim) as u64));
        group.bench_with_input(
            BenchmarkId::new("cosine", format!("{n}x{dim}")),
            &(),
            |bench, _| {
                bench.iter(|| batch_cosine(black_box(&query), black_box(&batch), black_box(&norms)))
            },
        );
    }

    group.finish();
}

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

    let n = 10000;
    let dim = 128;
    let vectors = random_vecs(n, dim);
    let batch = VerticalBatch::from_rows(&vectors);
    let query = random_vec(dim);

    for k in [1, 10, 100] {
        group.throughput(Throughput::Elements((n * dim) as u64));
        group.bench_with_input(BenchmarkId::new("l2", format!("k={k}")), &k, |bench, &k| {
            bench.iter(|| batch_knn(black_box(&query), black_box(&batch), k))
        });
    }

    group.finish();
}

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

    let n = 10000;
    let dim = 128;
    let vectors = random_vecs(n, dim);
    let batch = VerticalBatch::from_rows(&vectors);
    let query = random_vec(dim);

    for k in [1, 10, 100] {
        group.throughput(Throughput::Elements((n * dim) as u64));
        group.bench_with_input(
            BenchmarkId::new("cosine", format!("k={k}")),
            &k,
            |bench, &k| bench.iter(|| batch_knn_cosine(black_box(&query), black_box(&batch), k)),
        );
    }

    group.finish();
}

criterion_group!(
    benches,
    bench_batch_l2,
    bench_batch_dot,
    bench_batch_cosine,
    bench_batch_knn,
    bench_batch_knn_cosine,
);
criterion_main!(benches);