zrip-core 0.9.1

Shared types and codecs for zrip (internal crate)
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

#[cfg(not(feature = "paranoid"))]
#[inline(always)]
pub(crate) fn read_u32_le(data: &[u8], offset: usize) -> u32 {
    let end = offset.checked_add(4).expect("read offset overflow");
    assert!(end <= data.len());
    // SAFETY: The assertion proves offset..offset+4 is inside data, and
    // read_unaligned accepts any alignment.
    u32::from_le(unsafe { (data.as_ptr().add(offset) as *const u32).read_unaligned() })
}

#[cfg(feature = "paranoid")]
#[inline(always)]
pub(crate) fn read_u32_le(data: &[u8], offset: usize) -> u32 {
    u32::from_le_bytes(*data[offset..].first_chunk::<4>().unwrap())
}

#[cfg(not(feature = "paranoid"))]
#[inline(always)]
pub(crate) fn read_u64_le(data: &[u8], offset: usize) -> u64 {
    let end = offset.checked_add(8).expect("read offset overflow");
    assert!(end <= data.len());
    // SAFETY: The assertion proves offset..offset+8 is inside data, and
    // read_unaligned accepts any alignment.
    u64::from_le(unsafe { (data.as_ptr().add(offset) as *const u64).read_unaligned() })
}

#[cfg(feature = "paranoid")]
#[inline(always)]
pub(crate) fn read_u64_le(data: &[u8], offset: usize) -> u64 {
    u64::from_le_bytes(*data[offset..].first_chunk::<8>().unwrap())
}

#[cfg(not(feature = "paranoid"))]
#[inline(always)]
pub(super) unsafe fn bulk_rounds(
    data: &[u8],
    v1: &mut u64,
    v2: &mut u64,
    v3: &mut u64,
    v4: &mut u64,
) {
    let len = data.len();
    debug_assert!(len >= 32);

    // SAFETY: bulk_end and unroll_end are rounded down from len, so every
    // pointer stays within data. Each loop body reads complete 8-byte lanes
    // before advancing beyond those rounded limits.
    unsafe {
        let mut p = data.as_ptr();
        let bulk_end = data.as_ptr().add(len & !31);
        let unroll_end = data.as_ptr().add(len & !127);

        while p < unroll_end {
            *v1 = super::xxh64_round(*v1, (p as *const u64).read_unaligned().to_le());
            *v2 = super::xxh64_round(*v2, (p.add(8) as *const u64).read_unaligned().to_le());
            *v3 = super::xxh64_round(*v3, (p.add(16) as *const u64).read_unaligned().to_le());
            *v4 = super::xxh64_round(*v4, (p.add(24) as *const u64).read_unaligned().to_le());

            *v1 = super::xxh64_round(*v1, (p.add(32) as *const u64).read_unaligned().to_le());
            *v2 = super::xxh64_round(*v2, (p.add(40) as *const u64).read_unaligned().to_le());
            *v3 = super::xxh64_round(*v3, (p.add(48) as *const u64).read_unaligned().to_le());
            *v4 = super::xxh64_round(*v4, (p.add(56) as *const u64).read_unaligned().to_le());

            *v1 = super::xxh64_round(*v1, (p.add(64) as *const u64).read_unaligned().to_le());
            *v2 = super::xxh64_round(*v2, (p.add(72) as *const u64).read_unaligned().to_le());
            *v3 = super::xxh64_round(*v3, (p.add(80) as *const u64).read_unaligned().to_le());
            *v4 = super::xxh64_round(*v4, (p.add(88) as *const u64).read_unaligned().to_le());

            *v1 = super::xxh64_round(*v1, (p.add(96) as *const u64).read_unaligned().to_le());
            *v2 = super::xxh64_round(*v2, (p.add(104) as *const u64).read_unaligned().to_le());
            *v3 = super::xxh64_round(*v3, (p.add(112) as *const u64).read_unaligned().to_le());
            *v4 = super::xxh64_round(*v4, (p.add(120) as *const u64).read_unaligned().to_le());

            p = p.add(128);
        }

        while p < bulk_end {
            *v1 = super::xxh64_round(*v1, (p as *const u64).read_unaligned().to_le());
            *v2 = super::xxh64_round(*v2, (p.add(8) as *const u64).read_unaligned().to_le());
            *v3 = super::xxh64_round(*v3, (p.add(16) as *const u64).read_unaligned().to_le());
            *v4 = super::xxh64_round(*v4, (p.add(24) as *const u64).read_unaligned().to_le());
            p = p.add(32);
        }
    }
}

#[cfg(feature = "paranoid")]
#[inline(always)]
pub(super) fn bulk_rounds(data: &[u8], v1: &mut u64, v2: &mut u64, v3: &mut u64, v4: &mut u64) {
    let len = data.len();
    debug_assert!(len >= 32);

    #[inline(always)]
    fn rd64(data: &[u8], offset: usize) -> u64 {
        u64::from_le_bytes(*data[offset..].first_chunk::<8>().unwrap())
    }

    let bulk_end = len & !31;
    let unroll_end = len & !127;
    let mut offset = 0;

    while offset < unroll_end {
        *v1 = super::xxh64_round(*v1, rd64(data, offset));
        *v2 = super::xxh64_round(*v2, rd64(data, offset + 8));
        *v3 = super::xxh64_round(*v3, rd64(data, offset + 16));
        *v4 = super::xxh64_round(*v4, rd64(data, offset + 24));

        *v1 = super::xxh64_round(*v1, rd64(data, offset + 32));
        *v2 = super::xxh64_round(*v2, rd64(data, offset + 40));
        *v3 = super::xxh64_round(*v3, rd64(data, offset + 48));
        *v4 = super::xxh64_round(*v4, rd64(data, offset + 56));

        *v1 = super::xxh64_round(*v1, rd64(data, offset + 64));
        *v2 = super::xxh64_round(*v2, rd64(data, offset + 72));
        *v3 = super::xxh64_round(*v3, rd64(data, offset + 80));
        *v4 = super::xxh64_round(*v4, rd64(data, offset + 88));

        *v1 = super::xxh64_round(*v1, rd64(data, offset + 96));
        *v2 = super::xxh64_round(*v2, rd64(data, offset + 104));
        *v3 = super::xxh64_round(*v3, rd64(data, offset + 112));
        *v4 = super::xxh64_round(*v4, rd64(data, offset + 120));

        offset += 128;
    }

    while offset < bulk_end {
        *v1 = super::xxh64_round(*v1, rd64(data, offset));
        *v2 = super::xxh64_round(*v2, rd64(data, offset + 8));
        *v3 = super::xxh64_round(*v3, rd64(data, offset + 16));
        *v4 = super::xxh64_round(*v4, rd64(data, offset + 24));
        offset += 32;
    }
}