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
|
#![allow(deprecated)]
use criterion::{criterion_group, criterion_main, Criterion};
use itertools::{cloned, Itertools};
trait IterEx: Iterator {
// Another efficient implementation against which to compare,
// but needs `std` so is less desirable.
fn tree_fold1_vec<F>(self, mut f: F) -> Option<Self::Item>
where
F: FnMut(Self::Item, Self::Item) -> Self::Item,
Self: Sized,
{
let hint = self.size_hint().0;
let cap = std::mem::size_of::<usize>() * 8 - hint.leading_zeros() as usize;
let mut stack = Vec::with_capacity(cap);
self.enumerate().for_each(|(mut i, mut x)| {
while (i & 1) != 0 {
x = f(stack.pop().unwrap(), x);
i >>= 1;
}
stack.push(x);
});
stack.into_iter().fold1(f)
}
}
impl<T: Iterator> IterEx for T {}
macro_rules! def_benchs {
($N:expr,
$FUN:ident,
$BENCH_NAME:ident,
) => {
mod $BENCH_NAME {
use super::*;
pub fn sum(c: &mut Criterion) {
let v: Vec<u32> = (0..$N).collect();
c.bench_function(
&(stringify!($BENCH_NAME).replace('_', " ") + " sum"),
move |b| b.iter(|| cloned(&v).$FUN(|x, y| x + y)),
);
}
pub fn complex_iter(c: &mut Criterion) {
let u = (3..).take($N / 2);
let v = (5..).take($N / 2);
let it = u.chain(v);
c.bench_function(
&(stringify!($BENCH_NAME).replace('_', " ") + " complex iter"),
move |b| b.iter(|| it.clone().map(|x| x as f32).$FUN(f32::atan2)),
);
}
pub fn string_format(c: &mut Criterion) {
// This goes quadratic with linear `fold1`, so use a smaller
// size to not waste too much time in travis. The allocations
// in here are so expensive anyway that it'll still take
// way longer per iteration than the other two benchmarks.
let v: Vec<u32> = (0..($N / 4)).collect();
c.bench_function(
&(stringify!($BENCH_NAME).replace('_', " ") + " string format"),
move |b| {
b.iter(|| {
cloned(&v)
.map(|x| x.to_string())
.$FUN(|x, y| format!("{} + {}", x, y))
})
},
);
}
}
criterion_group!(
$BENCH_NAME,
$BENCH_NAME::sum,
$BENCH_NAME::complex_iter,
$BENCH_NAME::string_format,
);
};
}
def_benchs! {
10_000,
fold1,
fold1_10k,
}
def_benchs! {
10_000,
tree_fold1,
tree_fold1_stack_10k,
}
def_benchs! {
10_000,
tree_fold1_vec,
tree_fold1_vec_10k,
}
def_benchs! {
100,
fold1,
fold1_100,
}
def_benchs! {
100,
tree_fold1,
tree_fold1_stack_100,
}
def_benchs! {
100,
tree_fold1_vec,
tree_fold1_vec_100,
}
def_benchs! {
8,
fold1,
fold1_08,
}
def_benchs! {
8,
tree_fold1,
tree_fold1_stack_08,
}
def_benchs! {
8,
tree_fold1_vec,
tree_fold1_vec_08,
}
criterion_main!(
fold1_10k,
tree_fold1_stack_10k,
tree_fold1_vec_10k,
fold1_100,
tree_fold1_stack_100,
tree_fold1_vec_100,
fold1_08,
tree_fold1_stack_08,
tree_fold1_vec_08,
);
|
