arcweight 0.3.0

A high-performance, modular library for weighted finite state transducers with comprehensive examples and benchmarks
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

use arcweight::prelude::*;
use criterion::{criterion_group, criterion_main, Criterion};
#[cfg(feature = "parallel")]
use rayon::prelude::*;
use std::hint::black_box;
use std::time::Duration;

fn create_parallel_test_fst(n: usize) -> VectorFst<TropicalWeight> {
    let mut fst = VectorFst::new();
    let mut states = Vec::new();

    for _ in 0..=n {
        states.push(fst.add_state());
    }

    fst.set_start(states[0]);
    fst.set_final(states[n], TropicalWeight::one());

    // Create a complex FST with many parallel paths
    for i in 0..n {
        // Add multiple parallel arcs
        for j in 1..=5 {
            fst.add_arc(
                states[i],
                Arc::new(
                    j as u32,
                    j as u32,
                    TropicalWeight::new(j as f32),
                    states[i + 1],
                ),
            );
        }
    }

    fst
}

pub fn bench_parallel_composition(c: &mut Criterion) {
    let mut group = c.benchmark_group("parallel_composition");
    group.measurement_time(Duration::from_secs(10));

    let small_fst = create_parallel_test_fst(100);
    let medium_fst = create_parallel_test_fst(500);
    let large_fst = create_parallel_test_fst(1000);

    group.bench_function("small_fst_parallel_compose", |b| {
        b.iter(|| {
            let fst1 = black_box(&small_fst);
            let fst2 = black_box(&small_fst);
            let result: VectorFst<TropicalWeight> = compose_default(fst1, fst2).unwrap();
            black_box(result);
        })
    });

    group.bench_function("medium_fst_parallel_compose", |b| {
        b.iter(|| {
            let fst1 = black_box(&medium_fst);
            let fst2 = black_box(&medium_fst);
            let result: VectorFst<TropicalWeight> = compose_default(fst1, fst2).unwrap();
            black_box(result);
        })
    });

    group.bench_function("large_fst_parallel_compose", |b| {
        b.iter(|| {
            let fst1 = black_box(&large_fst);
            let fst2 = black_box(&large_fst);
            let result: VectorFst<TropicalWeight> = compose_default(fst1, fst2).unwrap();
            black_box(result);
        })
    });

    group.finish();
}

pub fn bench_parallel_shortest_path(c: &mut Criterion) {
    let mut group = c.benchmark_group("parallel_shortest_path");
    group.measurement_time(Duration::from_secs(10));

    let small_fst = create_parallel_test_fst(100);
    let medium_fst = create_parallel_test_fst(500);
    let large_fst = create_parallel_test_fst(1000);

    group.bench_function("small_fst_parallel_shortest_path", |b| {
        b.iter(|| {
            let fst = black_box(&small_fst);
            let result: VectorFst<TropicalWeight> = shortest_path_single(fst).unwrap();
            black_box(result);
        })
    });

    group.bench_function("medium_fst_parallel_shortest_path", |b| {
        b.iter(|| {
            let fst = black_box(&medium_fst);
            let result: VectorFst<TropicalWeight> = shortest_path_single(fst).unwrap();
            black_box(result);
        })
    });

    group.bench_function("large_fst_parallel_shortest_path", |b| {
        b.iter(|| {
            let fst = black_box(&large_fst);
            let result: VectorFst<TropicalWeight> = shortest_path_single(fst).unwrap();
            black_box(result);
        })
    });

    group.finish();
}

pub fn bench_parallel_arc_processing(c: &mut Criterion) {
    let mut group = c.benchmark_group("parallel_arc_processing");
    group.measurement_time(Duration::from_secs(10));

    let small_fst = create_parallel_test_fst(100);
    let medium_fst = create_parallel_test_fst(500);
    let large_fst = create_parallel_test_fst(1000);

    #[cfg(feature = "parallel")]
    group.bench_function("small_fst_parallel_arc_sum", |b| {
        b.iter(|| {
            let fst = black_box(&small_fst);
            let states: Vec<_> = fst.states().collect();
            let sum: f32 = states
                .par_iter()
                .flat_map(|state| fst.arcs(*state).collect::<Vec<_>>())
                .map(|arc| *arc.weight.value())
                .sum();
            black_box(sum);
        })
    });
    #[cfg(not(feature = "parallel"))]
    group.bench_function("small_fst_parallel_arc_sum", |b| {
        b.iter(|| {
            let fst = black_box(&small_fst);
            let states: Vec<_> = fst.states().collect();
            let sum: f32 = states
                .iter()
                .flat_map(|state| fst.arcs(*state).collect::<Vec<_>>())
                .map(|arc| *arc.weight.value())
                .sum();
            black_box(sum);
        })
    });

    #[cfg(feature = "parallel")]
    group.bench_function("medium_fst_parallel_arc_sum", |b| {
        b.iter(|| {
            let fst = black_box(&medium_fst);
            let states: Vec<_> = fst.states().collect();
            let sum: f32 = states
                .par_iter()
                .flat_map(|state| fst.arcs(*state).collect::<Vec<_>>())
                .map(|arc| *arc.weight.value())
                .sum();
            black_box(sum);
        })
    });
    #[cfg(not(feature = "parallel"))]
    group.bench_function("medium_fst_parallel_arc_sum", |b| {
        b.iter(|| {
            let fst = black_box(&medium_fst);
            let states: Vec<_> = fst.states().collect();
            let sum: f32 = states
                .iter()
                .flat_map(|state| fst.arcs(*state).collect::<Vec<_>>())
                .map(|arc| *arc.weight.value())
                .sum();
            black_box(sum);
        })
    });

    #[cfg(feature = "parallel")]
    group.bench_function("large_fst_parallel_arc_sum", |b| {
        b.iter(|| {
            let fst = black_box(&large_fst);
            let states: Vec<_> = fst.states().collect();
            let sum: f32 = states
                .par_iter()
                .flat_map(|state| fst.arcs(*state).collect::<Vec<_>>())
                .map(|arc| *arc.weight.value())
                .sum();
            black_box(sum);
        })
    });
    #[cfg(not(feature = "parallel"))]
    group.bench_function("large_fst_parallel_arc_sum", |b| {
        b.iter(|| {
            let fst = black_box(&large_fst);
            let states: Vec<_> = fst.states().collect();
            let sum: f32 = states
                .iter()
                .flat_map(|state| fst.arcs(*state).collect::<Vec<_>>())
                .map(|arc| *arc.weight.value())
                .sum();
            black_box(sum);
        })
    });

    group.finish();
}

criterion_group!(
    benches,
    bench_parallel_composition,
    bench_parallel_shortest_path,
    bench_parallel_arc_processing
);
criterion_main!(benches);