redstone-ml 0.0.0

High-performance Machine Learning, Auto-Differentiation and Tensor Algebra crate for Rust
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

use crate::IntegerDataType;
use rand_distr::num_traits::Zero;
use std::ops::{AddAssign, Mul};

pub(crate) trait DotProduct: AddAssign + Mul<Output=Self> + Zero + Copy {
    /// Performs a dot product between `count` elements beginning at `src0` and `src1`
    /// and writing the result to `dst`
    ///
    /// # Safety
    /// - `src0` and `src1` must represent a valid array of `count` elements.
    /// - `dst` must be a valid pointer
    unsafe fn dot_product(mut src0: *const Self, mut src1: *const Self, dst: *mut Self, count: usize) {
        let mut sum = Self::zero();

        for _ in 0..count {
            sum += (*src0) * (*src1);
            src0 = src0.add(1);
            src1 = src1.add(1);
        }

        *dst += sum;
    }

    /// Performs a dot product between `count` elements beginning at `src0` and `src1`
    /// with a space of `stride0` and `stride1` between elements respectively,
    /// and writing the result to `dst`
    ///
    /// # Safety
    /// - `src0` must point to a valid array of `count * stride0` elements.
    /// - `src1` must point to a valid array of `count * stride1` elements.
    /// - `dst` must be a valid pointer
    unsafe fn strided_dot_product(mut src0: *const Self, stride0: usize,
                                  mut src1: *const Self, stride1: usize,
                                  dst: *mut Self, count: usize) {
        let mut sum = Self::zero();

        for _ in 0..count {
            sum += (*src0) * (*src1);
            src0 = src0.add(stride0);
            src1 = src1.add(stride1);
        }

        *dst += sum;
    }
}

impl<T: IntegerDataType> DotProduct for T {}

impl DotProduct for f32 {
    #[cfg(apple_vdsp)]
    unsafe fn dot_product(src0: *const Self, src1: *const Self, dst: *mut Self, count: usize) {
        use crate::acceleration::vdsp::vDSP_dotpr;
        use std::ptr::addr_of_mut;

        let mut sum = Self::zero();
        vDSP_dotpr(src0, 1, src1, 1, addr_of_mut!(sum), count);
        *dst += sum;
    }

    #[cfg(all(neon_simd, not(apple_vdsp)))]
    unsafe fn dot_product(src0: *const Self, src1: *const Self, dst: *mut Self, count: usize) {
        use crate::ops::simd_sum_of_products::SIMDSumOfProducts;
        Self::simd_dot_product(src0, src1, dst, count);
    }

    #[cfg(all(not(apple_vdsp), not(neon_simd), blas))]
    unsafe fn dot_product(src0: *const Self, src1: *const Self, dst: *mut Self, count: usize) {
        use crate::acceleration::cblas::cblas_sdot;
        *dst += cblas_sdot(count as i32, src0, 1, src1, 1);
    }


    #[cfg(apple_vdsp)]
    unsafe fn strided_dot_product(src0: *const Self, stride0: usize,
                                  src1: *const Self, stride1: usize,
                                  dst: *mut Self, count: usize) {
        use crate::acceleration::vdsp::vDSP_dotpr;
        use std::ptr::addr_of_mut;

        let mut sum = Self::zero();
        vDSP_dotpr(src0, stride0 as isize, src1, stride1 as isize, addr_of_mut!(sum), count);
        *dst += sum;
    }

    #[cfg(all(not(apple_vdsp), not(neon_simd), blas))]
    unsafe fn strided_dot_product(src0: *const Self, stride0: usize,
                                  src1: *const Self, stride1: usize,
                                  dst: *mut Self, count: usize) {
        use crate::acceleration::cblas::cblas_sdot;
        *dst += cblas_sdot(count as i32, src0, stride0 as i32, src1, stride1 as i32);
    }
}

impl DotProduct for f64 {
    #[cfg(apple_vdsp)]
    unsafe fn dot_product(src0: *const Self, src1: *const Self, dst: *mut Self, count: usize) {
        use crate::acceleration::vdsp::vDSP_dotprD;
        use std::ptr::addr_of_mut;

        let mut sum = Self::zero();
        vDSP_dotprD(src0, 1, src1, 1, addr_of_mut!(sum), count);
        *dst += sum;
    }

    #[cfg(all(neon_simd, not(apple_vdsp)))]
    unsafe fn dot_product(src0: *const Self, src1: *const Self, dst: *mut Self, count: usize) {
        use crate::ops::simd_sum_of_products::SIMDSumOfProducts;
        Self::simd_dot_product(src0, src1, dst, count);
    }

    #[cfg(all(not(apple_vdsp), not(neon_simd), blas))]
    unsafe fn dot_product(src0: *const Self, src1: *const Self, dst: *mut Self, count: usize) {
        use crate::acceleration::cblas::cblas_ddot;
        *dst += cblas_ddot(count as i32, src0, 1, src1, 1);
    }

    #[cfg(apple_vdsp)]
    unsafe fn strided_dot_product(src0: *const Self, stride0: usize,
                                  src1: *const Self, stride1: usize,
                                  dst: *mut Self, count: usize) {
        use crate::acceleration::vdsp::vDSP_dotprD;
        use std::ptr::addr_of_mut;

        let mut sum = Self::zero();
        vDSP_dotprD(src0, stride0 as isize, src1, stride1 as isize, addr_of_mut!(sum), count);
        *dst += sum;
    }

    #[cfg(all(not(apple_vdsp), blas))]
    unsafe fn strided_dot_product(src0: *const Self, stride0: usize,
                                  src1: *const Self, stride1: usize,
                                  dst: *mut Self, count: usize) {
        use crate::acceleration::cblas::cblas_ddot;
        *dst += cblas_ddot(count as i32, src0, stride0 as i32, src1, stride1 as i32);
    }
}