fastlanes 0.5.0

Rust implementation of the FastLanes compression layout
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

#![allow(unused_assignments)]

use crate::{iterate, supported_bit_width, unpack, BitPacking, FastLanes};
use paste::paste;

pub trait Delta: BitPacking {
    fn delta<const LANES: usize>(
        input: &[Self; 1024],
        base: &[Self; LANES],
        output: &mut [Self; 1024],
    );

    fn undelta<const LANES: usize>(
        input: &[Self; 1024],
        base: &[Self; LANES],
        output: &mut [Self; 1024],
    );

    fn undelta_pack<const LANES: usize, const W: usize, const B: usize>(
        input: &[Self; B],
        base: &[Self; LANES],
        output: &mut [Self; 1024],
    );
}

macro_rules! impl_delta {
    ($T:ty) => {
        impl Delta for $T {
            #[inline(never)]
            fn delta<const LANES: usize>(
                input: &[Self; 1024],
                base: &[Self; LANES],
                output: &mut [Self; 1024],
            ) {
                const {
                    assert!(LANES == Self::LANES);
                }

                for lane in 0..Self::LANES {
                    let mut prev = base[lane];
                    iterate!($T, lane, |$idx| {
                        let next = input[$idx];
                        output[$idx] = next.wrapping_sub(prev);
                        prev = next;
                    });
                }
            }

            #[inline(never)]
            fn undelta<const LANES: usize>(
                input: &[Self; 1024],
                base: &[Self; LANES],
                output: &mut [Self; 1024],
            ) {
                const {
                    assert!(LANES == Self::LANES);
                }
                for lane in 0..LANES {
                    let mut prev = base[lane];
                    iterate!($T, lane, |$idx| {
                        let next = input[$idx].wrapping_add(prev);
                        output[$idx] = next;
                        prev = next;
                    });
                }
            }

            #[inline(never)]
            fn undelta_pack<const LANES: usize, const W: usize, const B: usize>(
                input: &[Self; B],
                base: &[Self; LANES],
                output: &mut [Self; 1024],
            ) {
                const {
                    assert!(LANES == Self::LANES);
                    assert!(supported_bit_width(W, 8 * core::mem::size_of::<$T>()));
                    assert!(B == 1024 * W / Self::T);
                }

                for lane in 0..Self::LANES {
                    let mut prev = base[lane];
                    unpack!($T, W, input, lane, |$idx, $elem| {
                        let next = $elem.wrapping_add(prev);
                        output[$idx] = next;
                        prev = next;
                    });
                }
            }
        }
    };
}

impl_delta!(u8);
impl_delta!(u16);
impl_delta!(u32);
impl_delta!(u64);

#[cfg(test)]
mod test {
    use super::*;
    use crate::Transpose;
    use core::mem::size_of;

    #[test]
    fn test_delta() {
        const LANES: usize = u16::LANES;
        const W: usize = 15;
        const B: usize = 1024 * W / u16::T;

        let mut values: [u16; 1024] = [0; 1024];
        for i in 0..1024 {
            values[i] = (i / 8) as u16;
        }

        let mut transposed = [0; 1024];
        Transpose::transpose(&values, &mut transposed);

        let mut deltas = [0; 1024];
        Delta::delta(&transposed, &[0; 64], &mut deltas);

        let mut packed = [0; 128 * W / size_of::<u16>()];
        BitPacking::pack::<W, B>(&deltas, &mut packed);

        // Fused kernel
        let mut unpacked = [0; 1024];
        Delta::undelta_pack::<LANES, W, B>(&packed, &[0; 64], &mut unpacked);
        assert_eq!(transposed, unpacked);

        // Unfused kernel
        BitPacking::unpack::<W, B>(&packed, &mut unpacked);
        let mut undelta = [0; 1024];
        Delta::undelta(&unpacked, &[0; 64], &mut undelta);
        assert_eq!(transposed, undelta);
    }
}