ndarray-parallel 0.8.0

Parallelization for ndarray (using rayon).
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


#![feature(test)]

extern crate num_cpus;
extern crate test;
use test::Bencher;

extern crate rayon;
#[macro_use] extern crate ndarray;
#[macro_use] extern crate ndarray_parallel;
use ndarray::prelude::*;
use ndarray_parallel::prelude::*;

use ndarray::Zip;

const EXP_N: usize = 256;
const ADDN: usize = 512;

use std::cmp::max;

fn set_threads() {
    let n = max(1, num_cpus::get() / 2);
    //println!("Using {} threads", n);
    let _ = rayon::ThreadPoolBuilder::new().num_threads(n).build_global();
}

#[bench]
fn map_exp_regular(bench: &mut Bencher)
{
    let mut a = Array2::<f64>::zeros((EXP_N, EXP_N));
    a.swap_axes(0, 1);
    bench.iter(|| {
        a.mapv_inplace(|x| x.exp());
    });
}

#[bench]
fn rayon_exp_regular(bench: &mut Bencher)
{
    set_threads();
    let mut a = Array2::<f64>::zeros((EXP_N, EXP_N));
    a.swap_axes(0, 1);
    bench.iter(|| {
        a.view_mut().into_par_iter().for_each(|x| *x = x.exp());
    });
}

const FASTEXP: usize = EXP_N;

#[inline]
fn fastexp(x: f64) -> f64 {
    let x = 1. + x/1024.;
    x.powi(1024)
}

#[bench]
fn map_fastexp_regular(bench: &mut Bencher)
{
    let mut a = Array2::<f64>::zeros((FASTEXP, FASTEXP));
    bench.iter(|| {
        a.mapv_inplace(|x| fastexp(x))
    });
}

#[bench]
fn rayon_fastexp_regular(bench: &mut Bencher)
{
    set_threads();
    let mut a = Array2::<f64>::zeros((FASTEXP, FASTEXP));
    bench.iter(|| {
        a.view_mut().into_par_iter().for_each(|x| *x = fastexp(*x));
    });
}

#[bench]
fn map_fastexp_cut(bench: &mut Bencher)
{
    let mut a = Array2::<f64>::zeros((FASTEXP, FASTEXP));
    let mut a = a.slice_mut(s![.., ..-1]);
    bench.iter(|| {
        a.mapv_inplace(|x| fastexp(x))
    });
}

#[bench]
fn rayon_fastexp_cut(bench: &mut Bencher)
{
    set_threads();
    let mut a = Array2::<f64>::zeros((FASTEXP, FASTEXP));
    let mut a = a.slice_mut(s![.., ..-1]);
    bench.iter(|| {
        a.view_mut().into_par_iter().for_each(|x| *x = fastexp(*x));
    });
}

#[bench]
fn map_fastexp_by_axis(bench: &mut Bencher)
{
    let mut a = Array2::<f64>::zeros((FASTEXP, FASTEXP));
    bench.iter(|| {
        for mut sheet in a.axis_iter_mut(Axis(0)) {
            sheet.mapv_inplace(fastexp)
        }
    });
}

#[bench]
fn rayon_fastexp_by_axis(bench: &mut Bencher)
{
    set_threads();
    let mut a = Array2::<f64>::zeros((FASTEXP, FASTEXP));
    bench.iter(|| {
        a.axis_iter_mut(Axis(0)).into_par_iter()
            .for_each(|mut sheet| sheet.mapv_inplace(fastexp));
    });
}

#[bench]
fn rayon_fastexp_zip(bench: &mut Bencher)
{
    set_threads();
    let mut a = Array2::<f64>::zeros((FASTEXP, FASTEXP));
    bench.iter(|| {
        Zip::from(&mut a).into_par_iter().for_each(|(elt, )| *elt = fastexp(*elt));
    });
}

#[bench]
fn add(bench: &mut Bencher)
{
    let mut a = Array2::<f64>::zeros((ADDN, ADDN));
    let b = Array2::<f64>::zeros((ADDN, ADDN));
    let c = Array2::<f64>::zeros((ADDN, ADDN));
    let d = Array2::<f64>::zeros((ADDN, ADDN));
    bench.iter(|| {
        azip!(mut a, b, c, d in {
            *a += b.exp() + c.exp() + d.exp();
        });
    });
}

#[bench]
fn rayon_add(bench: &mut Bencher)
{
    set_threads();
    let mut a = Array2::<f64>::zeros((ADDN, ADDN));
    let b = Array2::<f64>::zeros((ADDN, ADDN));
    let c = Array2::<f64>::zeros((ADDN, ADDN));
    let d = Array2::<f64>::zeros((ADDN, ADDN));
    bench.iter(|| {
        par_azip!(mut a, b, c, d in {
            *a += b.exp() + c.exp() + d.exp();
        });
    });
}