dyn-stack 0.11.0

Dynamic stack wrapper for unsized allocations
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

use criterion::{black_box, criterion_group, criterion_main, Criterion};
use dyn_stack::{DynStack, GlobalMemBuffer, GlobalPodBuffer, PodStack, StackReq};

pub fn criterion_benchmark(c: &mut Criterion) {
    c.bench_function("memalloc", |b| {
        b.iter(|| {
            black_box(GlobalMemBuffer::new(StackReq::new_aligned::<i32>(
                1 << 20,
                16,
            )))
        })
    });
    c.bench_function("memalloc-zeroed", |b| {
        b.iter(|| {
            black_box(GlobalPodBuffer::new(StackReq::new_aligned::<i32>(
                1 << 20,
                16,
            )))
        })
    });

    for n in [32, 64, 1024, 16384] {
        let align = 64;
        let single_scratch = StackReq::new_aligned::<f32>(n, align);
        let scratch = single_scratch.and(single_scratch);

        let mut mem = GlobalMemBuffer::new(scratch);
        let stack = DynStack::new(&mut *mem);

        {
            let (src, stack) = stack.make_aligned_with(n, align, |_| 0.0_f32);
            let (mut dst, _) = stack.make_aligned_with(n, align, |_| 0.0_f32);
            c.bench_function(&format!("preallocated-{}", n), |b| {
                b.iter(|| {
                    for (d, s) in dst.iter_mut().zip(src.iter()) {
                        *d = s + s;
                    }
                    black_box(&mut dst);
                })
            });
        }

        {
            let (src, stack) = stack.make_aligned_with(n, align, |_| 0.0_f32);
            c.bench_function(&format!("allocate-on-demand-init-{}", n), |b| {
                b.iter(|| {
                    let (mut dst, _) = stack.make_aligned_with(n, align, |_| 0.0_f32);
                    for (d, s) in dst.iter_mut().zip(src.iter()) {
                        *d = s + s;
                    }
                    black_box(&mut dst);
                })
            });
        }

        {
            let (src, stack) = stack.make_aligned_with(n, align, |_| 0.0_f32);
            c.bench_function(&format!("allocate-on-demand-uninit-{}", n), |b| {
                b.iter(|| {
                    let (mut dst, _) = stack.make_aligned_uninit::<f32>(n, align);
                    for (d, s) in dst.iter_mut().zip(src.iter()) {
                        d.write(s + s);
                    }
                    black_box(&mut dst);
                })
            });
        }

        {
            let (src, stack) = stack.make_aligned_with(n, align, |_| 0.0_f32);
            c.bench_function(&format!("allocate-on-demand-collect-{}", n), |b| {
                b.iter(|| {
                    let (mut dst, _) = stack
                        .collect_aligned(align, src.iter().zip(src.iter()).map(|(a, b)| a + b));
                    black_box(&mut dst);
                })
            });
        }
    }

    for n in [32, 64, 1024, 16384] {
        let align = 64;
        let single_scratch = StackReq::new_aligned::<f32>(n, align);
        let scratch = single_scratch.and(single_scratch);

        let mut mem = GlobalPodBuffer::new(scratch);
        let stack = PodStack::new(&mut *mem);

        {
            let (src, stack) = stack.make_aligned_with(n, align, |_| 0.0_f32);
            let (mut dst, _) = stack.make_aligned_with(n, align, |_| 0.0_f32);
            c.bench_function(&format!("pod-preallocated-{}", n), |b| {
                b.iter(|| {
                    for (d, s) in dst.iter_mut().zip(src.iter()) {
                        *d = s + s;
                    }
                    black_box(&mut dst);
                })
            });
        }

        {
            let (src, stack) = stack.make_aligned_with(n, align, |_| 0.0_f32);
            c.bench_function(&format!("pod-allocate-on-demand-init-{}", n), |b| {
                b.iter(|| {
                    let (mut dst, _) = stack.make_aligned_with(n, align, |_| 0.0_f32);
                    for (d, s) in dst.iter_mut().zip(src.iter()) {
                        *d = s + s;
                    }
                    black_box(&mut dst);
                })
            });
        }

        {
            let (src, stack) = stack.make_aligned_with(n, align, |_| 0.0_f32);
            c.bench_function(&format!("pod-allocate-on-demand-uninit-{}", n), |b| {
                b.iter(|| {
                    let (mut dst, _) = stack.make_aligned_raw::<f32>(n, align);
                    for (d, s) in dst.iter_mut().zip(src.iter()) {
                        *d = s + s;
                    }
                    black_box(&mut dst);
                })
            });
        }

        {
            let (src, stack) = stack.make_aligned_with(n, align, |_| 0.0_f32);
            c.bench_function(&format!("pod-allocate-on-demand-collect-{}", n), |b| {
                b.iter(|| {
                    let (mut dst, _) = stack
                        .collect_aligned(align, src.iter().zip(src.iter()).map(|(a, b)| a + b));
                    black_box(&mut dst);
                })
            });
        }
    }
}

criterion_group!(benches, criterion_benchmark);
criterion_main!(benches);