sark-core 0.11.0

the L7 layer for dope
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
137
138
139

use core::arch::x86_64::{
    __m128i, _mm_andnot_si128, _mm_cmpeq_epi8, _mm_loadu_si128, _mm_max_epu8, _mm_min_epu8,
    _mm_movemask_epi8, _mm_or_si128, _mm_set1_epi8,
};

use super::{HeaderNameOutcome, HeaderValueOutcome, scalar};

const LANE_MASK: u32 = 0xFFFF;

pub(super) fn scan_header_name(bytes: &[u8], start: usize) -> HeaderNameOutcome {
    // SAFETY: SSE2 is part of the x86_64 baseline, and the backend loads only
    // after proving that a complete 16-byte lane remains in `bytes`.
    unsafe { scan_header_name_sse2(bytes, start) }
}

pub(super) fn scan_header_value(bytes: &[u8], start: usize) -> HeaderValueOutcome {
    // SAFETY: same baseline and lane-bounds proof as `scan_header_name`.
    unsafe { scan_header_value_sse2(bytes, start) }
}

pub(super) fn request_target_is_valid(bytes: &[u8]) -> bool {
    // SAFETY: same baseline and lane-bounds proof as `scan_header_name`.
    unsafe { request_target_is_valid_sse2(bytes) }
}

#[target_feature(enable = "sse2")]
fn lane_eq(chunk: __m128i, byte: u8) -> __m128i {
    _mm_cmpeq_epi8(chunk, _mm_set1_epi8(byte as i8))
}

#[target_feature(enable = "sse2")]
fn in_range(chunk: __m128i, lo: u8, hi: u8) -> __m128i {
    let clamped = _mm_min_epu8(
        _mm_max_epu8(chunk, _mm_set1_epi8(lo as i8)),
        _mm_set1_epi8(hi as i8),
    );
    _mm_cmpeq_epi8(chunk, clamped)
}

#[target_feature(enable = "sse2")]
fn valid_mask(chunk: __m128i) -> __m128i {
    let alpha_num = _mm_or_si128(
        _mm_or_si128(in_range(chunk, b'A', b'Z'), in_range(chunk, b'a', b'z')),
        in_range(chunk, b'0', b'9'),
    );
    let punct = _mm_or_si128(
        _mm_or_si128(
            _mm_or_si128(
                _mm_or_si128(lane_eq(chunk, b'!'), lane_eq(chunk, b'#')),
                _mm_or_si128(lane_eq(chunk, b'$'), lane_eq(chunk, b'%')),
            ),
            _mm_or_si128(
                _mm_or_si128(lane_eq(chunk, b'&'), lane_eq(chunk, b'\'')),
                _mm_or_si128(lane_eq(chunk, b'*'), lane_eq(chunk, b'+')),
            ),
        ),
        _mm_or_si128(
            _mm_or_si128(
                _mm_or_si128(lane_eq(chunk, b'-'), lane_eq(chunk, b'.')),
                _mm_or_si128(lane_eq(chunk, b'^'), lane_eq(chunk, b'_')),
            ),
            _mm_or_si128(
                lane_eq(chunk, b'`'),
                _mm_or_si128(lane_eq(chunk, b'|'), lane_eq(chunk, b'~')),
            ),
        ),
    );
    _mm_or_si128(alpha_num, punct)
}

#[target_feature(enable = "sse2")]
fn scan_header_name_sse2(bytes: &[u8], start: usize) -> HeaderNameOutcome {
    let mut idx = start;
    let len = bytes.len();
    while idx + 16 <= len {
        // SAFETY: the loop condition proves the full unaligned lane is in `bytes`.
        let chunk = unsafe { _mm_loadu_si128(bytes.as_ptr().add(idx).cast::<__m128i>()) };
        let hit_vector = _mm_or_si128(lane_eq(chunk, b':'), lane_eq(chunk, b'\r'));
        let valid_or_hit = _mm_or_si128(valid_mask(chunk), hit_vector);
        let hit = _mm_movemask_epi8(hit_vector) as u32;
        let invalid = _mm_movemask_epi8(valid_or_hit) as u32 ^ LANE_MASK;
        if invalid != 0 {
            let invalid_offset = invalid.trailing_zeros();
            if hit != 0 && hit.trailing_zeros() < invalid_offset {
                let pos = idx + hit.trailing_zeros() as usize;
                return HeaderNameOutcome::Found {
                    pos,
                    byte: bytes[pos],
                };
            }
            return HeaderNameOutcome::Invalid;
        }
        if hit != 0 {
            let pos = idx + hit.trailing_zeros() as usize;
            return HeaderNameOutcome::Found {
                pos,
                byte: bytes[pos],
            };
        }
        idx += 16;
    }
    scalar::scan_header_name(bytes, idx)
}

#[target_feature(enable = "sse2")]
fn scan_header_value_sse2(bytes: &[u8], start: usize) -> HeaderValueOutcome {
    let mut idx = start;
    let len = bytes.len();
    while idx + 16 <= len {
        // SAFETY: the loop condition proves the full unaligned lane is in `bytes`.
        let chunk = unsafe { _mm_loadu_si128(bytes.as_ptr().add(idx).cast::<__m128i>()) };
        let low_control = _mm_andnot_si128(
            lane_eq(chunk, b'\t'),
            _mm_cmpeq_epi8(_mm_min_epu8(chunk, _mm_set1_epi8(0x1f)), chunk),
        );
        let special = _mm_or_si128(low_control, lane_eq(chunk, 0x7f));
        let mask = _mm_movemask_epi8(special) as u32;
        if mask != 0 {
            return scalar::scan_header_value(bytes, idx + mask.trailing_zeros() as usize);
        }
        idx += 16;
    }
    scalar::scan_header_value(bytes, idx)
}

#[target_feature(enable = "sse2")]
fn request_target_is_valid_sse2(bytes: &[u8]) -> bool {
    let mut chunks = bytes.chunks_exact(16);
    for chunk in &mut chunks {
        // SAFETY: `chunks_exact(16)` yields a complete 16-byte lane.
        let vector = unsafe { _mm_loadu_si128(chunk.as_ptr().cast::<__m128i>()) };
        let low = _mm_cmpeq_epi8(_mm_min_epu8(vector, _mm_set1_epi8(0x20)), vector);
        let invalid = _mm_or_si128(low, lane_eq(vector, 0x7f));
        if _mm_movemask_epi8(invalid) != 0 {
            return false;
        }
    }
    scalar::request_target_is_valid(chunks.remainder())
}