trueno 0.17.4

High-performance SIMD compute library with GPU support for matrix operations
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

//! WebAssembly SIMD128 reduction operations (dot, sum, max, min, argmax, argmin).

use crate::backends::VectorBackend;
#[cfg(target_arch = "wasm32")]
use std::arch::wasm32::*;

/// WASM SIMD128 dot product.
#[target_feature(enable = "simd128")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub unsafe fn dot(a: &[f32], b: &[f32]) -> f32 {
    let len = a.len();
    let mut i = 0;
    let mut acc = f32x4_splat(0.0);

    while i + 4 <= len {
        let va = v128_load(a.as_ptr().add(i) as *const v128);
        let vb = v128_load(b.as_ptr().add(i) as *const v128);
        acc = f32x4_add(acc, f32x4_mul(va, vb));
        i += 4;
    }

    let mut result = f32x4_extract_lane::<0>(acc)
        + f32x4_extract_lane::<1>(acc)
        + f32x4_extract_lane::<2>(acc)
        + f32x4_extract_lane::<3>(acc);
    result += a[i..].iter().zip(&b[i..]).map(|(x, y)| x * y).sum::<f32>();
    result
}

/// WASM SIMD128 vector sum.
#[target_feature(enable = "simd128")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub unsafe fn sum(a: &[f32]) -> f32 {
    let len = a.len();
    let mut i = 0;
    let mut acc = f32x4_splat(0.0);

    while i + 4 <= len {
        acc = f32x4_add(acc, v128_load(a.as_ptr().add(i) as *const v128));
        i += 4;
    }

    let mut result = f32x4_extract_lane::<0>(acc)
        + f32x4_extract_lane::<1>(acc)
        + f32x4_extract_lane::<2>(acc)
        + f32x4_extract_lane::<3>(acc);
    result += a[i..].iter().sum::<f32>();
    result
}

/// WASM SIMD128 vector max.
#[target_feature(enable = "simd128")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub unsafe fn max(a: &[f32]) -> f32 {
    let len = a.len();
    let mut i = 0;
    let mut vmax = f32x4_splat(a[0]);

    while i + 4 <= len {
        vmax = f32x4_pmax(vmax, v128_load(a.as_ptr().add(i) as *const v128));
        i += 4;
    }

    let mut result = f32x4_extract_lane::<0>(vmax)
        .max(f32x4_extract_lane::<1>(vmax))
        .max(f32x4_extract_lane::<2>(vmax))
        .max(f32x4_extract_lane::<3>(vmax));
    for &val in &a[i..] {
        if val > result {
            result = val;
        }
    }
    result
}

/// WASM SIMD128 vector min.
#[target_feature(enable = "simd128")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub unsafe fn min(a: &[f32]) -> f32 {
    let len = a.len();
    let mut i = 0;
    let mut vmin = f32x4_splat(a[0]);

    while i + 4 <= len {
        vmin = f32x4_pmin(vmin, v128_load(a.as_ptr().add(i) as *const v128));
        i += 4;
    }

    let mut result = f32x4_extract_lane::<0>(vmin)
        .min(f32x4_extract_lane::<1>(vmin))
        .min(f32x4_extract_lane::<2>(vmin))
        .min(f32x4_extract_lane::<3>(vmin));
    for &val in &a[i..] {
        if val < result {
            result = val;
        }
    }
    result
}

/// WASM SIMD128 argmax.
#[target_feature(enable = "simd128")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub unsafe fn argmax(a: &[f32]) -> usize {
    let mut max_idx: usize = 0;
    let mut max_val = a[0];
    for (i, &val) in a.iter().enumerate() {
        if val > max_val {
            max_val = val;
            max_idx = i;
        }
    }
    max_idx
}

/// WASM SIMD128 argmin.
#[target_feature(enable = "simd128")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub unsafe fn argmin(a: &[f32]) -> usize {
    let mut min_idx: usize = 0;
    let mut min_val = a[0];
    for (i, &val) in a.iter().enumerate() {
        if val < min_val {
            min_val = val;
            min_idx = i;
        }
    }
    min_idx
}

/// Kahan sum (delegates to scalar).
// SAFETY: caller ensures preconditions are met for this unsafe function
pub unsafe fn sum_kahan(a: &[f32]) -> f32 {
    crate::backends::scalar::ScalarBackend::sum_kahan(a)
}