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

use custos::{Buffer, Device, Shape};

pub trait ElementWise<T, D: Device, S: Shape = ()>: Device {
    fn add(&self, lhs: &Buffer<T, D, S>, rhs: &Buffer<T, D, S>) -> Buffer<T, Self, S>;
    fn mul(&self, lhs: &Buffer<T, D, S>, rhs: &Buffer<T, D, S>) -> Buffer<T, Self, S>;
}

pub trait ElementWiseBuf<T, D: Device> {
    fn add(&self, rhs: &Buffer<T, D>) -> Buffer<T, D>;
}

impl<'a, T, D: Device + ElementWise<T, D>> ElementWiseBuf<T, D> for Buffer<'a, T, D> {
    fn add(&self, rhs: &Buffer<T, D>) -> Buffer<T, D> {
        self.device().add(self, rhs)
    }
}

pub trait Add2<Rhs = Self> {
    type Output;

    fn add(self, rhs: Rhs) -> Self::Output;
}

impl<'a, T, D: Device + ElementWise<T, D>> Add2 for &Buffer<'a, T, D> {
    type Output = Buffer<'a, T, D>;

    fn add(self, rhs: Self) -> Self::Output {
        self.device().add(self, rhs)
    }
}

pub struct Matrix<'a, T, D: Device> {
    buf: Buffer<'a, T, D>,
}

impl<'a, T, D: Device> Matrix<'a, T, D> {
    #[inline]
    pub fn device(&self) -> &'a D {
        self.buf.device()
    }
}

impl<'a, T, D: Device + ElementWise<T, D>> Add2 for &Matrix<'a, T, D> {
    type Output = Matrix<'a, T, D>;

    fn add(self, rhs: Self) -> Self::Output {
        Matrix {
            buf: self.device().add(&self.buf, &rhs.buf),
        }
    }
}