sonic-rs 0.1.3

Sonic-rs is a fast Rust JSON library based on SIMD
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
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168

#[cfg(target_arch = "x86_64")]
use std::arch::x86_64::*;

pub fn prefix_xor(bitmask: u64) -> u64 {
    unsafe {
        let all_ones = _mm_set1_epi8(-1i8);
        let result = _mm_clmulepi64_si128(_mm_set_epi64x(0, bitmask as i64), all_ones, 0);
        _mm_cvtsi128_si64(result) as u64
    }
}

#[inline(always)]
pub fn get_nonspace_bits(data: &[u8; 64]) -> u64 {
    unsafe {
        let lo: std::arch::x86_64::__m256i = _mm256_loadu_si256(data.as_ptr() as *const __m256i);
        let hi: std::arch::x86_64::__m256i =
            _mm256_loadu_si256(data.as_ptr().offset(32) as *const __m256i);
        let whitespace_data = _mm256_setr_epi8(
            b' ' as i8,
            100,
            100,
            100,
            17,
            100,
            113,
            2,
            100,
            b'\t' as i8,
            b'\n' as i8,
            112,
            100,
            b'\r' as i8,
            100,
            100,
            b' ' as i8,
            100,
            100,
            100,
            17,
            100,
            113,
            2,
            100,
            b'\t' as i8,
            b'\n' as i8,
            112,
            100,
            b'\r' as i8,
            100,
            100,
        );
        let shuf_lo = _mm256_shuffle_epi8(whitespace_data, lo);
        let shuf_hi = _mm256_shuffle_epi8(whitespace_data, hi);
        let lo = _mm256_cmpeq_epi8(lo, shuf_lo);
        let hi = _mm256_cmpeq_epi8(hi, shuf_hi);
        let space = _mm256_movemask_epi8(lo) as u32 as u64
            | ((_mm256_movemask_epi8(hi) as u32 as u64) << 32);
        !space
    }
}

macro_rules! packadd_1 {
    ($v:ident) => {
        let delta = _mm_set1_epi64x(0x010A010A010A010A);
        $v = _mm_maddubs_epi16($v, delta);
    };
}

macro_rules! packadd_2 {
    ($v:ident) => {
        let delta = _mm_set1_epi64x(0x0001006400010064);
        $v = _mm_madd_epi16($v, delta);
    };
}

macro_rules! packadd_4 {
    ($v:ident) => {
        $v = _mm_packus_epi32($v, $v);
        let delta = _mm_set_epi16(0, 0, 0, 0, 1, 10000, 1, 10000);
        $v = _mm_madd_epi16($v, delta);
    };
}

// simd add for 5 ~ 8 digits
macro_rules! simd_add_5_8 {
    ($v:ident, $nd:literal) => {{
        $v = _mm_slli_si128($v, 16 - $nd);
        packadd_1!($v);
        packadd_2!($v);
        (_mm_extract_epi32($v, 2) as u64) * 10000 + (_mm_extract_epi32($v, 3) as u64)
    }};
}

// simd add for 9 ~ 15 digits
macro_rules! simd_add_9_15 {
    ($v:ident, $nd:literal) => {{
        $v = _mm_slli_si128($v, 16 - $nd);
        packadd_1!($v);
        packadd_2!($v);
        packadd_4!($v);
        (_mm_extract_epi32($v, 0) as u64) * 100000000 + (_mm_extract_epi32($v, 1) as u64)
    }};
}

macro_rules! simd_add_16 {
    ($v:ident) => {{
        packadd_1!($v);
        packadd_2!($v);
        packadd_4!($v);
        (_mm_extract_epi32($v, 0) as u64) * 100000000 + (_mm_extract_epi32($v, 1) as u64)
    }};
}

#[inline(always)]
pub unsafe fn simd_str2int(c: &[u8], need: usize) -> (u64, usize) {
    debug_assert!(need <= 16);
    let data = _mm_loadu_si128(c.as_ptr() as *const __m128i);
    let zero = _mm_setzero_si128();
    let nine = _mm_set1_epi8(9);
    let zero_c = _mm_set1_epi8(b'0' as i8);

    let mut data = _mm_sub_epi8(data, zero_c);
    let lt_zero = _mm_cmpgt_epi8(zero, data);
    let gt_nine = _mm_cmpgt_epi8(data, nine);

    let is_num_end = _mm_or_si128(lt_zero, gt_nine);
    let is_num_end_int = _mm_movemask_epi8(is_num_end);

    // get the real parsed count
    let mut count = need;
    if is_num_end_int != 0 {
        let digits = is_num_end_int.trailing_zeros() as usize;
        if digits < need {
            count = digits;
        }
    }

    let sum = match count {
        1 => _mm_extract_epi8(data, 0) as u64,
        2 => (_mm_extract_epi8(data, 0) * 10 + _mm_extract_epi8(data, 1)) as u64,
        3 => {
            // shift to clear the non-digit ascii in vector
            data = _mm_slli_si128(data, 16 - 3);
            packadd_1!(data);
            // add the highest two lanes
            (_mm_extract_epi16(data, 6) * 100 + _mm_extract_epi16(data, 7)) as u64
        }
        4 => {
            data = _mm_slli_si128(data, 16 - 4);
            packadd_1!(data);
            (_mm_extract_epi16(data, 6) * 100 + _mm_extract_epi16(data, 7)) as u64
        }
        5 => simd_add_5_8!(data, 5),
        6 => simd_add_5_8!(data, 6),
        7 => simd_add_5_8!(data, 7),
        8 => simd_add_5_8!(data, 8),
        9 => simd_add_9_15!(data, 9),
        10 => simd_add_9_15!(data, 10),
        11 => simd_add_9_15!(data, 11),
        12 => simd_add_9_15!(data, 12),
        13 => simd_add_9_15!(data, 13),
        14 => simd_add_9_15!(data, 14),
        15 => simd_add_9_15!(data, 15),
        16 => simd_add_16!(data),
        _ => unreachable!(),
    };
    (sum, count)
}