zune-image 0.4.1

An image library, contiaining necessary capabilities to decode, manipulate and encode images
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
172
173
174
175
176
177
178
179
180
181
182
183
184

/*
 * Copyright (c) 2023.
 *
 * This software is free software;
 *
 * You can redistribute it or modify it under terms of the MIT, Apache License or Zlib license
 */

#![cfg(any(target_arch = "x86", target_arch = "x86_64"))]
#![cfg(feature = "simd")]

#[cfg(target_arch = "x86")]
use std::arch::x86::*;
#[cfg(target_arch = "x86_64")]
use std::arch::x86_64::*;

use crate::deinterleave::scalar::{
    de_interleave_four_channels_scalar, de_interleave_three_channels_scalar
};

#[target_feature(enable = "sse4.1")]
pub unsafe fn de_interleave_three_channels_sse3_u8(
    source: &[u8], c1: &mut [u8], c2: &mut [u8], c3: &mut [u8]
) {
    const CHUNK_SIZE: usize = 48;
    const OUT_CHUNK_SIZE: usize = CHUNK_SIZE / 3;

    assert_eq!(source.len() % 3, 0, "Source must be divisible by 3");
    assert_eq!(c1.len(), c2.len(), "Out sources must be of equal size");
    assert_eq!(c2.len(), c3.len(), "Out sources must be of equal size");

    let ssse3_red_indeces_0 =
        _mm_set_epi8(-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 15, 12, 9, 6, 3, 0);
    let ssse3_red_indeces_1 =
        _mm_set_epi8(-1, -1, -1, -1, -1, 14, 11, 8, 5, 2, -1, -1, -1, -1, -1, -1);
    let ssse3_red_indeces_2 =
        _mm_set_epi8(13, 10, 7, 4, 1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1);

    let ssse3_green_indeces_0 =
        _mm_set_epi8(-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 13, 10, 7, 4, 1);
    let ssse3_green_indeces_1 =
        _mm_set_epi8(-1, -1, -1, -1, -1, 15, 12, 9, 6, 3, 0, -1, -1, -1, -1, -1);
    let ssse3_green_indeces_2 =
        _mm_set_epi8(14, 11, 8, 5, 2, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1);

    let ssse3_blue_indeces_0 =
        _mm_set_epi8(-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 14, 11, 8, 5, 2);
    let ssse3_blue_indeces_1 =
        _mm_set_epi8(-1, -1, -1, -1, -1, -1, 13, 10, 7, 4, 1, -1, -1, -1, -1, -1);
    let ssse3_blue_indeces_2 =
        _mm_set_epi8(15, 12, 9, 6, 3, 0, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1);

    for (((source_chunk, a), b), c) in source
        .chunks_exact(CHUNK_SIZE)
        .zip(c1.chunks_exact_mut(OUT_CHUNK_SIZE))
        .zip(c2.chunks_exact_mut(OUT_CHUNK_SIZE))
        .zip(c3.chunks_exact_mut(OUT_CHUNK_SIZE))
    {
        // https://docs.google.com/presentation/d/1I0-SiHid1hTsv7tjLST2dYW5YF5AJVfs9l4Rg9rvz48/htmlpresent

        let chunk0 = _mm_loadu_si128(source_chunk.as_ptr().cast());
        let chunk1 = _mm_loadu_si128(source_chunk[16..].as_ptr().cast());
        let chunk2 = _mm_loadu_si128(source_chunk[32..].as_ptr().cast());

        let red = _mm_or_si128(
            _mm_or_si128(
                _mm_shuffle_epi8(chunk0, ssse3_red_indeces_0),
                _mm_shuffle_epi8(chunk1, ssse3_red_indeces_1)
            ),
            _mm_shuffle_epi8(chunk2, ssse3_red_indeces_2)
        );
        let green = _mm_or_si128(
            _mm_or_si128(
                _mm_shuffle_epi8(chunk0, ssse3_green_indeces_0),
                _mm_shuffle_epi8(chunk1, ssse3_green_indeces_1)
            ),
            _mm_shuffle_epi8(chunk2, ssse3_green_indeces_2)
        );

        let blue = _mm_or_si128(
            _mm_or_si128(
                _mm_shuffle_epi8(chunk0, ssse3_blue_indeces_0),
                _mm_shuffle_epi8(chunk1, ssse3_blue_indeces_1)
            ),
            _mm_shuffle_epi8(chunk2, ssse3_blue_indeces_2)
        );

        _mm_storeu_si128(a.as_mut_ptr().cast(), red);
        _mm_storeu_si128(b.as_mut_ptr().cast(), green);
        _mm_storeu_si128(c.as_mut_ptr().cast(), blue);
    }
    if source.len() % CHUNK_SIZE != 0 {
        // do the remainder
        let rem = source.len() % CHUNK_SIZE;
        let start = source.len() - rem;
        let c_start = c1.len() - (rem / 3);

        let c1 = &mut c1[c_start..];
        let c2 = &mut c2[c_start..];
        let c3 = &mut c3[c_start..];

        de_interleave_three_channels_scalar(&source[start..], c1, c2, c3);
    }
}

#[allow(clippy::zero_prefixed_literal)]
#[target_feature(enable = "sse4.1")]
#[rustfmt::skip]
pub unsafe fn de_interleave_four_channels_sse41(
    source: &[u8], c1: &mut [u8], c2: &mut [u8], c3: &mut [u8], c4: &mut [u8],
)
{
    // https://godbolt.org/z/hhTarvMds

    const CHUNK_SIZE: usize = 64;
    const OUT_CHUNK_SIZE: usize = CHUNK_SIZE / 4;

    let shuffle_mask = _mm_set_epi8(15, 11, 7, 3,
                                    14, 10, 6, 2,
                                    13, 09, 5, 1,
                                    12, 08, 4, 0);

    for ((((source_chunk, rr), gg), bb), aa) in source
        .chunks_exact(CHUNK_SIZE)
        .zip(c1.chunks_exact_mut(OUT_CHUNK_SIZE))
        .zip(c2.chunks_exact_mut(OUT_CHUNK_SIZE))
        .zip(c3.chunks_exact_mut(OUT_CHUNK_SIZE))
        .zip(c4.chunks_exact_mut(OUT_CHUNK_SIZE))
    {
        // Data is in [r,g,b,a,r,g,b,r,g,b,a,r,g,b,a]
        //

        // We load 64 bytes to ensure that when we write, we do a write of 16 which
        // fits nicely into a sse register.
        let t1 = _mm_loadu_si128(source_chunk[00..].as_ptr().cast());
        let t2 = _mm_loadu_si128(source_chunk[08..].as_ptr().cast());
        let t3 = _mm_loadu_si128(source_chunk[16..].as_ptr().cast());
        let t4 = _mm_loadu_si128(source_chunk[24..].as_ptr().cast());

        // convert data into
        // rrrr,gggg,bbbb,aaaa
        let x1 = _mm_shuffle_epi8(t1, shuffle_mask);
        let x2 = _mm_shuffle_epi8(t2, shuffle_mask);
        let x3 = _mm_shuffle_epi8(t3, shuffle_mask);
        let x4 = _mm_shuffle_epi8(t4, shuffle_mask);

        // Each register contains rrrr,gggg bbbb and aaaa
        // so now bring them all together. so that a register has
        // rrrrrrrrr,ggggggg,bbbbbb and aaaaaa

        // rrrrrr,gggggg
        let p1 = _mm_unpacklo_epi32(x1, x2);
        // rrrrr,gggggg
        let p2 = _mm_unpacklo_epi32(x3, x4);
        // bbbbb,aaaaaa
        let p3 = _mm_unpackhi_epi32(x1, x2);
        // bbbbb,aaaaaa
        let p4 = _mm_unpackhi_epi32(x3, x4);

        let rrrr = _mm_unpacklo_epi64(p1, p2);
        let gggg = _mm_unpackhi_epi64(p1, p2);
        let bbbb = _mm_unpacklo_epi64(p3, p4);
        let aaaa = _mm_unpackhi_epi64(p3, p4);

        _mm_storeu_si128(rr.as_mut_ptr().cast(), rrrr);
        _mm_storeu_si128(gg.as_mut_ptr().cast(), gggg);
        _mm_storeu_si128(bb.as_mut_ptr().cast(), bbbb);
        _mm_storeu_si128(aa.as_mut_ptr().cast(), aaaa);
    }
    if source.len() % CHUNK_SIZE != 0
    {
        // do the remainder
        let rem = source.len() % CHUNK_SIZE;
        let start = source.len() - rem;
        let c_start = c1.len() - (rem / 4);

        let c1 = &mut c1[c_start..];
        let c2 = &mut c2[c_start..];
        let c3 = &mut c3[c_start..];
        let c4 = &mut c4[c_start..];

        de_interleave_four_channels_scalar(&source[start..], c1, c2, c3, c4);
    }
}