helicase 0.1.1

SIMD-accelerated library for FASTA/FASTQ parsing and bitpacking
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
169
170
171

#![allow(clippy::missing_transmute_annotations)]

use crate::config::{advanced::*, *};
use crate::lexer::*;
use core::arch::x86_64::*;
use core::mem::transmute;

const GREATER_THAN: __m256i = unsafe { transmute([b'>'; 32]) };
const LINE_FEED: __m256i = unsafe { transmute([b'\n'; 32]) };
const ASCII_N: __m256i = unsafe { transmute([b'N'; 32]) };
const LUT_ACTG: __m256i = unsafe { transmute(*b"A_C_T_G_________A_C_T_G_________") };

#[inline(always)]
pub fn extract_fasta_bitmask<const CONFIG: Config>(buf: &[u8]) -> FastaBitmask {
    unsafe {
        let ptr = buf.as_ptr() as *const __m256i;
        let v1 = _mm256_loadu_si256(ptr);
        let v2 = _mm256_loadu_si256(ptr.add(1));

        let open_bracket = u8_mask(v1, v2, GREATER_THAN);
        let line_feeds = u8_mask(v1, v2, LINE_FEED);

        let (is_dna, two_bits, high_bit, low_bit, mask_non_actg, mask_n) =
            bitpack_dna::<CONFIG>(v1, v2);

        FastaBitmask {
            open_bracket,
            line_feeds,
            is_dna,
            two_bits,
            high_bit,
            low_bit,
            mask_non_actg,
            mask_n,
        }
    }
}

#[inline(always)]
pub fn extract_fastq_bitmask<const CONFIG: Config>(buf: &[u8]) -> FastqBitmask {
    unsafe {
        let ptr = buf.as_ptr() as *const __m256i;
        let v1 = _mm256_loadu_si256(ptr);
        let v2 = _mm256_loadu_si256(ptr.add(1));

        let line_feeds = u8_mask(v1, v2, LINE_FEED);

        let (is_dna, two_bits, high_bit, low_bit, mask_non_actg, mask_n) =
            bitpack_dna::<CONFIG>(v1, v2);

        FastqBitmask {
            line_feeds,
            is_dna,
            two_bits,
            high_bit,
            low_bit,
            mask_non_actg,
            mask_n,
        }
    }
}

#[inline(always)]
fn bitpack_dna<const CONFIG: Config>(v1: __m256i, v2: __m256i) -> (u64, u128, u64, u64, u64, u64) {
    unsafe {
        let mut is_dna = !0;
        let mut two_bits = 0;
        let mut high_bit = 0;
        let mut low_bit = 0;
        let mut mask_non_actg = 0;
        let mut mask_n = 0;

        if flag_is_set(CONFIG, COMPUTE_DNA_COLUMNAR) {
            let (mm_hi_1, mm_lo_1, mm_hi_2, mm_lo_2) = (
                _mm256_movemask_epi8(_mm256_slli_epi16(v1, 5)) as u32 as u64,
                _mm256_movemask_epi8(_mm256_slli_epi16(v1, 6)) as u32 as u64,
                _mm256_movemask_epi8(_mm256_slli_epi16(v2, 5)) as u32 as u64,
                _mm256_movemask_epi8(_mm256_slli_epi16(v2, 6)) as u32 as u64,
            );
            high_bit = mm_hi_1 | (mm_hi_2 << 32);
            low_bit = mm_lo_1 | (mm_lo_2 << 32);
        }

        if flag_is_set(CONFIG, COMPUTE_DNA_PACKED) {
            #[cfg(all(target_feature = "bmi2", not(feature = "no-pdep")))]
            {
                let (mm_hi_1, mm_lo_1, mm_hi_2, mm_lo_2) = (
                    _mm256_movemask_epi8(_mm256_slli_epi16(v1, 5)) as u32 as u64,
                    _mm256_movemask_epi8(_mm256_slli_epi16(v1, 6)) as u32 as u64,
                    _mm256_movemask_epi8(_mm256_slli_epi16(v2, 5)) as u32 as u64,
                    _mm256_movemask_epi8(_mm256_slli_epi16(v2, 6)) as u32 as u64,
                );
                let mm_1 =
                    _pdep_u64(mm_hi_1, 0xAAAAAAAAAAAAAAAA) | _pdep_u64(mm_lo_1, 0x5555555555555555);
                let mm_2 =
                    _pdep_u64(mm_hi_2, 0xAAAAAAAAAAAAAAAA) | _pdep_u64(mm_lo_2, 0x5555555555555555);
                two_bits = (mm_1 as u128) | ((mm_2 as u128) << 64);
            }
            #[cfg(any(not(target_feature = "bmi2"), feature = "no-pdep"))]
            {
                // Adapted from https://github.com/Daniel-Liu-c0deb0t/cute-nucleotides/commit/007164bce68f671188fa5c607982fbd306112cb3
                let (iv1, iv2) = (
                    _mm256_permute4x64_epi64(v1, 0xD8),
                    _mm256_permute4x64_epi64(v2, 0xD8),
                );
                let (hi_1, lo_1, hi_2, lo_2) = (
                    _mm256_slli_epi16(iv1, 5),
                    _mm256_slli_epi16(iv1, 6),
                    _mm256_slli_epi16(iv2, 5),
                    _mm256_slli_epi16(iv2, 6),
                );
                let (mm_hi_1, mm_lo_1, mm_hi_2, mm_lo_2) = (
                    _mm256_movemask_epi8(_mm256_unpackhi_epi8(lo_1, hi_1)) as u32 as u64,
                    _mm256_movemask_epi8(_mm256_unpacklo_epi8(lo_1, hi_1)) as u32 as u64,
                    _mm256_movemask_epi8(_mm256_unpackhi_epi8(lo_2, hi_2)) as u32 as u64,
                    _mm256_movemask_epi8(_mm256_unpacklo_epi8(lo_2, hi_2)) as u32 as u64,
                );
                let mm_1 = (mm_hi_1 << 32) | mm_lo_1;
                let mm_2 = (mm_hi_2 << 32) | mm_lo_2;
                two_bits = (mm_1 as u128) | ((mm_2 as u128) << 64);
            }
        }

        if flag_is_set(CONFIG, SPLIT_NON_ACTG | COMPUTE_MASK_NON_ACTG) {
            let mask_two_bits = _mm256_set1_epi8(0b110i8);
            let mask_upper = _mm256_set1_epi8(0b11011111u8 as i8);
            let uv1 = _mm256_and_si256(v1, mask_upper);
            let uv2 = _mm256_and_si256(v2, mask_upper);

            let mask_actg = movemask_64(
                _mm256_cmpeq_epi8(
                    _mm256_shuffle_epi8(LUT_ACTG, _mm256_and_si256(v1, mask_two_bits)),
                    uv1,
                ),
                _mm256_cmpeq_epi8(
                    _mm256_shuffle_epi8(LUT_ACTG, _mm256_and_si256(v2, mask_two_bits)),
                    uv2,
                ),
            );
            if flag_is_set(CONFIG, SPLIT_NON_ACTG) {
                is_dna = mask_actg;
            }
            if flag_is_set(CONFIG, COMPUTE_MASK_NON_ACTG) {
                mask_non_actg = !mask_actg;
            }
            if flag_is_set(CONFIG, COMPUTE_MASK_N) {
                mask_n = u8_mask(v1, v2, ASCII_N);
            }
        }

        (is_dna, two_bits, high_bit, low_bit, mask_non_actg, mask_n)
    }
}

#[inline(always)]
fn movemask_64(v1: __m256i, v2: __m256i) -> u64 {
    unsafe {
        (_mm256_movemask_epi8(v1) as u32 as u64) | ((_mm256_movemask_epi8(v2) as u32 as u64) << 32)
    }
}

#[inline(always)]
pub fn u8_mask(v1: __m256i, v2: __m256i, vc: __m256i) -> u64 {
    unsafe {
        let cmp_c1 = _mm256_cmpeq_epi8(v1, vc);
        let cmp_c2 = _mm256_cmpeq_epi8(v2, vc);
        let a = _mm256_movemask_epi8(cmp_c1) as u32 as u64;
        let b = _mm256_movemask_epi8(cmp_c2) as u32 as u64;
        a | (b << 32)
    }
}