basexx 0.2.0

basexx utilities. base32, base64, base58, base56...
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

use super::super::*;

#[inline(never)]
#[target_feature(enable = "avx2")]
pub(crate) unsafe fn _ascii_to_binary_128_avx2(
    lup: &[u8],
    buf: &mut [u8],
) -> Result<(), DecodeError> {
    assert_eq!(lup.len(), 128);
    unsafe { _ascii_to_binary_128_avx2_chunks32(lup, buf)? };
    Ok(())
}

#[inline(always)]
pub(crate) unsafe fn _ascii_to_binary_128_avx2_chunks32(
    lup: &[u8],
    buf: &mut [u8],
) -> Result<(), DecodeError> {
    let mut iter = buf.chunks_exact_mut(32);
    for chunk in iter.by_ref() {
        let mut buf2: [u64; 4] = bytemuck::pod_read_unaligned(chunk);
        unsafe {
            _ascii_to_binary_128_avx2_c32_chunks32(lup, &mut buf2)?;
        }
        chunk.copy_from_slice(bytemuck::bytes_of(&buf2));
    }
    let remaind = iter.into_remainder();
    _ascii_to_binary_scalar(lup, remaind)?;
    //
    Ok(())
}

#[cfg(target_feature = "sse2")]
#[inline(always)]
//#[target_feature(enable = "avx2")]
pub(crate) unsafe fn _ascii_to_binary_128_avx2_c32(
    lup: &[u8],
    buf: &mut [u64; 4],
) -> Result<(), DecodeError> {
    assert_eq!(lup.len(), 128);
    unsafe { _ascii_to_binary_128_avx2_c32_chunks32(lup, buf)? };
    Ok(())
}

macro_rules! dec_check_error {
    ($err_buf: expr, $buf_ptr: expr) => {
        if $err_buf[0] != 0 || $err_buf[1] != 0 || $err_buf[2] != 0 || $err_buf[3] != 0 {
            // on error
            for i in 0..4 {
                let cc = $err_buf[i];
                for j in 0..8 {
                    if (cc >> (8 * j)) & 0xFF != 0 {
                        return Err(DecodeError::InvalidByte(*($buf_ptr.add(j + 8 * i))));
                    }
                }
            }
        }
    };
}

#[inline(always)]
pub(crate) unsafe fn _ascii_to_binary_128_avx2_c32_chunks32(
    lup: &[u8],
    buf: &mut [u64; 4],
) -> Result<(), DecodeError> {
    //panic!("PASS: SSE2");
    #[cfg(target_arch = "x86")]
    use core::arch::x86::*;
    #[cfg(target_arch = "x86_64")]
    use core::arch::x86_64::*;
    //
    unsafe {
        let buf_ptr = buf.as_mut_ptr() as *mut u8;
        //
        // ASCII graphic: U+0021 '!' ..= U+007E '~': 33..=126
        // skip range: 0..=31
        let mm_map03 = _mm_loadu_si128(lup.as_ptr().add(16 * 2) as *const __m128i);
        let mm_map04 = _mm_loadu_si128(lup.as_ptr().add(16 * 3) as *const __m128i);
        let mm_map05 = _mm_loadu_si128(lup.as_ptr().add(16 * 4) as *const __m128i);
        let mm_map06 = _mm_loadu_si128(lup.as_ptr().add(16 * 5) as *const __m128i);
        let mm_map07 = _mm_loadu_si128(lup.as_ptr().add(16 * 6) as *const __m128i);
        let mm_map08 = _mm_loadu_si128(lup.as_ptr().add(16 * 7) as *const __m128i);
        let mm_map3 = _mm256_broadcastsi128_si256(mm_map03);
        let mm_map4 = _mm256_broadcastsi128_si256(mm_map04);
        let mm_map5 = _mm256_broadcastsi128_si256(mm_map05);
        let mm_map6 = _mm256_broadcastsi128_si256(mm_map06);
        let mm_map7 = _mm256_broadcastsi128_si256(mm_map07);
        let mm_map8 = _mm256_broadcastsi128_si256(mm_map08);
        let err_buf = [0u64; 4];
        //
        {
            // ..|00ffffff|00eeeeee|00dddddd|00cccccc|00bbbbbb|00aaaaaa
            let mm_in = _mm256_loadu_si256(buf_ptr as *const __m256i);
            //
            // check error
            let mm_err = _mm256_or_si256(
                _mm256_cmpgt_epi8(_mm256_set1_epi8(33), mm_in),
                _mm256_cmpeq_epi8(mm_in, _mm256_set1_epi8(127)),
            );
            _mm256_storeu_si256(err_buf.as_ptr() as *mut __m256i, mm_err);
            dec_check_error!(err_buf, buf_ptr);
            //_mm_prefetch(buf_ptr.cast::<i8>(), _MM_HINT_T0);
            let mm_in_p1 = _mm256_add_epi8(mm_in, _mm256_set1_epi8(1));
            //
            let mm_mask33 = _mm256_cmpgt_epi8(_mm256_set1_epi8(16 * 3), mm_in);
            let mm_mask44 = _mm256_cmpgt_epi8(_mm256_set1_epi8(16 * 4), mm_in);
            let mm_mask55 = _mm256_cmpgt_epi8(_mm256_set1_epi8(16 * 5), mm_in);
            let mm_mask66 = _mm256_cmpgt_epi8(_mm256_set1_epi8(16 * 6), mm_in);
            let mm_mask77 = _mm256_cmpgt_epi8(_mm256_set1_epi8(16 * 7), mm_in);
            let mm_mask88 = _mm256_set1_epi8(-1);
            //
            let mm_mask3 = mm_mask33;
            let mm_and3 = _mm256_and_si256(mm_in_p1, mm_mask3);
            let mm_idx3 = _mm256_sub_epi8(mm_and3, _mm256_set1_epi8(1));
            let mm_out3 = _mm256_shuffle_epi8(mm_map3, mm_idx3);
            //
            let mm_mask4 = _mm256_andnot_si256(mm_mask33, mm_mask44);
            let mm_and4 = _mm256_and_si256(mm_in_p1, mm_mask4);
            let mm_idx4 = _mm256_sub_epi8(mm_and4, _mm256_set1_epi8(1));
            let mm_out4 = _mm256_shuffle_epi8(mm_map4, mm_idx4);
            //
            let mm_mask5 = _mm256_andnot_si256(mm_mask44, mm_mask55);
            let mm_and5 = _mm256_and_si256(mm_in_p1, mm_mask5);
            let mm_idx5 = _mm256_sub_epi8(mm_and5, _mm256_set1_epi8(1));
            let mm_out5 = _mm256_shuffle_epi8(mm_map5, mm_idx5);
            //
            let mm_mask6 = _mm256_andnot_si256(mm_mask55, mm_mask66);
            let mm_and6 = _mm256_and_si256(mm_in_p1, mm_mask6);
            let mm_idx6 = _mm256_sub_epi8(mm_and6, _mm256_set1_epi8(1));
            let mm_out6 = _mm256_shuffle_epi8(mm_map6, mm_idx6);
            //
            let mm_mask7 = _mm256_andnot_si256(mm_mask66, mm_mask77);
            let mm_and7 = _mm256_and_si256(mm_in_p1, mm_mask7);
            let mm_idx7 = _mm256_sub_epi8(mm_and7, _mm256_set1_epi8(1));
            let mm_out7 = _mm256_shuffle_epi8(mm_map7, mm_idx7);
            //
            let mm_mask8 = _mm256_andnot_si256(mm_mask77, mm_mask88);
            let mm_and8 = _mm256_and_si256(mm_in_p1, mm_mask8);
            let mm_idx8 = _mm256_sub_epi8(mm_and8, _mm256_set1_epi8(1));
            let mm_out8 = _mm256_shuffle_epi8(mm_map8, mm_idx8);
            //
            // make out
            let mm_out = _mm256_or_si256(
                _mm256_or_si256(
                    _mm256_or_si256(mm_out3, mm_out4),
                    _mm256_or_si256(mm_out5, mm_out6),
                ),
                _mm256_or_si256(mm_out7, mm_out8),
            );
            let mm_err = _mm256_cmpgt_epi8(_mm256_set1_epi8(0), mm_out);
            _mm256_storeu_si256(err_buf.as_ptr() as *mut __m256i, mm_err);
            dec_check_error!(err_buf, buf_ptr);
            _mm256_storeu_si256(buf_ptr as *mut __m256i, mm_out);
        }
    };
    //
    Ok(())
}

#[cfg(all(test, not(feature = "bench")))]
mod tests;

#[cfg(all(test, feature = "ubench"))]
mod benches;
#[cfg(all(test, feature = "ubench"))]
#[allow(unused_imports)]
pub(crate) use benches::*;