custos 0.7.0

A minimal OpenCL, WGPU, CUDA and host CPU array manipulation engine.
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

use custos::{range, Buffer, GraphOpt, CPU};

#[cfg(feature = "opencl")]
use custos::OpenCL;

use crate::graph::AddOp;

#[test]
fn test_graph() -> custos::Result<()> {
    let device = CPU::new();

    // idx: 0
    let a = Buffer::from((&device, [1, 2, 3, 4, 5, 6]));
    // idx: 1
    let b = Buffer::from((&device, [2, 3, 1, 4, 0, 5]));

    for ep in range(1) {
        // idx: 2, deps: [0, 1]
        let c = a.add(&b);
        assert_eq!(vec![3, 5, 4, 8, 5, 11], c.read());

        // idx: 3, deps: [2, 2]
        let d = c.relu();

        assert_eq!(vec![3, 5, 4, 8, 5, 11], d.read());

        // idx: 4, deps: [3, 1]
        let e = d.add(&b);

        if ep == 1 {
            assert_eq!(c.ptr, d.ptr);
            assert_eq!(c.ptr, e.ptr);
        }
        device.optimize()?;
    }
    Ok(())
}

#[cfg(feature = "opencl")]
#[test]
fn test_graph_cl() -> custos::Result<()> {
    let device = OpenCL::new(0)?;

    // idx: 0
    let a = Buffer::from((&device, [1, 2, 3, 4, 5, 6]));
    // idx: 1
    let b = Buffer::from((&device, [2, 3, 1, 4, 0, 5]));

    for ep in range(1) {
        // idx: 2, deps: [0, 1]
        let c = a.add(&b);
        assert_eq!(vec![3, 5, 4, 8, 5, 11], c.read());

        // idx: 3, deps: [2, 2]
        let d = c.relu();

        assert_eq!(vec![3, 5, 4, 8, 5, 11], d.read());

        // idx: 4, deps: [3, 1]
        let e = d.add(&b);

        if ep == 1 {
            assert_eq!(c.ptr, d.ptr);
            assert_eq!(c.ptr, e.ptr);
        }
        device.optimize()?;
    }
    Ok(())
}

#[cfg(feature = "cuda")]
#[test]
fn test_graph_cu() -> custos::Result<()> {
    use custos::CUDA;

    let device = CUDA::new(0)?;

    // idx: 0
    let a = Buffer::from((&device, [1, 2, 3, 4, 5, 6]));
    // idx: 1
    let b = Buffer::from((&device, [2, 3, 1, 4, 0, 5]));

    for ep in range(1) {
        // idx: 2, deps: [0, 1]
        let c = a.add(&b);
        assert_eq!(vec![3, 5, 4, 8, 5, 11], c.read());

        // idx: 3, deps: [2, 2]
        let d = c.relu();

        assert_eq!(vec![3, 5, 4, 8, 5, 11], d.read());

        // idx: 4, deps: [3, 1]
        let e = d.add(&b);

        if ep == 1 {
            assert_eq!(c.ptr, d.ptr);
            assert_eq!(c.ptr, e.ptr);
        }
        device.optimize()?;
    }
    Ok(())
}