yuv 0.8.13

High performance utilities for YUV format handling and conversion.
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
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227

/*
 * Copyright (c) Radzivon Bartoshyk, 12/2024. All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without modification,
 * are permitted provided that the following conditions are met:
 *
 * 1.  Redistributions of source code must retain the above copyright notice, this
 * list of conditions and the following disclaimer.
 *
 * 2.  Redistributions in binary form must reproduce the above copyright notice,
 * this list of conditions and the following disclaimer in the documentation
 * and/or other materials provided with the distribution.
 *
 * 3.  Neither the name of the copyright holder nor the names of its
 * contributors may be used to endorse or promote products derived from
 * this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */
use crate::avx2::avx2_utils::{_mm256_load_deinterleave_rgb, _mm256_store_interleave_rgb_for_yuv};
use crate::shuffle::ShuffleConverter;
use crate::yuv_support::YuvSourceChannels;
#[cfg(target_arch = "x86")]
use std::arch::x86::*;
#[cfg(target_arch = "x86_64")]
use std::arch::x86_64::*;

/// This is default shuffling with interleaving and de-interleaving.
///
/// For the same channels count there is more fast approach on x86 with reshuffling table
pub(crate) struct ShuffleConverterAvx2<const SRC: u8, const DST: u8> {}

impl<const SRC: u8, const DST: u8> Default for ShuffleConverterAvx2<SRC, DST> {
    fn default() -> Self {
        ShuffleConverterAvx2 {}
    }
}

impl<const SRC: u8, const DST: u8> ShuffleConverter<u8, SRC, DST>
    for ShuffleConverterAvx2<SRC, DST>
{
    fn convert(&self, src: &[u8], dst: &mut [u8], width: usize) {
        unsafe { shuffle_channels8_avx_impl::<SRC, DST>(src, dst, width) }
    }
}

#[target_feature(enable = "avx2")]
unsafe fn shuffle_channels8_avx_impl<const SRC: u8, const DST: u8>(
    src: &[u8],
    dst: &mut [u8],
    _: usize,
) {
    let src_channels: YuvSourceChannels = SRC.into();
    let dst_channels: YuvSourceChannels = DST.into();

    for (src, dst) in src
        .chunks_exact(32 * src_channels.get_channels_count())
        .zip(dst.chunks_exact_mut(32 * dst_channels.get_channels_count()))
    {
        let (a0, b0, c0, d0) = _mm256_load_deinterleave_rgb::<SRC>(src.as_ptr());
        _mm256_store_interleave_rgb_for_yuv::<DST>(dst.as_mut_ptr(), a0, b0, c0, d0);
    }

    let src = src
        .chunks_exact(32 * src_channels.get_channels_count())
        .remainder();
    let dst = dst
        .chunks_exact_mut(32 * dst_channels.get_channels_count())
        .into_remainder();

    if !src.is_empty() && !dst.is_empty() {
        assert!(src.len() < 32 * 4);
        assert!(dst.len() < 32 * 4);
        let mut transient_src: [u8; 32 * 4] = [0; 32 * 4];
        let mut transient_dst: [u8; 32 * 4] = [0; 32 * 4];
        std::ptr::copy_nonoverlapping(src.as_ptr(), transient_src.as_mut_ptr(), src.len());
        let (a0, b0, c0, d0) = _mm256_load_deinterleave_rgb::<SRC>(transient_src.as_ptr());
        _mm256_store_interleave_rgb_for_yuv::<DST>(transient_dst.as_mut_ptr(), a0, b0, c0, d0);
        std::ptr::copy_nonoverlapping(transient_dst.as_ptr(), dst.as_mut_ptr(), dst.len());
    }
}

/// This is shuffling only for 4 channels image
///
/// This is more fast method that just swaps channel positions
pub(crate) struct ShuffleQTableConverterAvx2<const SRC: u8, const DST: u8> {
    q_table_avx: [u8; 32],
}

const RGBA_TO_BGRA_TABLE_AVX2: [u8; 32] = [
    2,
    1,
    0,
    3,
    2 + 4,
    1 + 4,
    4,
    3 + 4,
    2 + 8,
    1 + 8,
    8,
    3 + 8,
    2 + 12,
    1 + 12,
    12,
    3 + 12,
    2,
    1,
    0,
    3,
    2 + 4,
    1 + 4,
    4,
    3 + 4,
    2 + 8,
    1 + 8,
    8,
    3 + 8,
    2 + 12,
    1 + 12,
    12,
    3 + 12,
];

impl<const SRC: u8, const DST: u8> ShuffleQTableConverterAvx2<SRC, DST> {
    pub(crate) fn create() -> Self {
        let src_channels: YuvSourceChannels = SRC.into();
        let dst_channels: YuvSourceChannels = DST.into();
        if src_channels.get_channels_count() != 4 || dst_channels.get_channels_count() != 4 {
            unimplemented!("Shuffle table implemented only for 4 channels");
        }
        let new_table_avx: [u8; 32] = match src_channels {
            YuvSourceChannels::Rgb => unreachable!(),
            YuvSourceChannels::Rgba => match dst_channels {
                YuvSourceChannels::Rgb => unreachable!(),
                YuvSourceChannels::Rgba => RGBA_TO_BGRA_TABLE_AVX2,
                YuvSourceChannels::Bgra => RGBA_TO_BGRA_TABLE_AVX2,
                YuvSourceChannels::Bgr => unreachable!(),
            },
            YuvSourceChannels::Bgra => match dst_channels {
                YuvSourceChannels::Rgb => unreachable!(),
                YuvSourceChannels::Rgba => RGBA_TO_BGRA_TABLE_AVX2,
                YuvSourceChannels::Bgra => RGBA_TO_BGRA_TABLE_AVX2,
                YuvSourceChannels::Bgr => unreachable!(),
            },
            YuvSourceChannels::Bgr => unreachable!(),
        };
        ShuffleQTableConverterAvx2 {
            q_table_avx: new_table_avx,
        }
    }
}

impl<const SRC: u8, const DST: u8> ShuffleConverter<u8, SRC, DST>
    for ShuffleQTableConverterAvx2<SRC, DST>
{
    fn convert(&self, src: &[u8], dst: &mut [u8], width: usize) {
        unsafe { shuffle_qtable_channels8_avx_impl::<SRC, DST>(src, dst, width, self.q_table_avx) }
    }
}

#[target_feature(enable = "avx2")]
unsafe fn shuffle_qtable_channels8_avx_impl<const SRC: u8, const DST: u8>(
    src: &[u8],
    dst: &mut [u8],
    _: usize,
    vq_table_avx: [u8; 32],
) {
    let src_channels: YuvSourceChannels = SRC.into();
    let dst_channels: YuvSourceChannels = DST.into();
    assert_eq!(src_channels.get_channels_count(), 4);
    assert_eq!(dst_channels.get_channels_count(), 4);

    let q_table_avx = _mm256_loadu_si256(vq_table_avx.as_ptr() as *const _);

    for (src, dst) in src.chunks_exact(32 * 4).zip(dst.chunks_exact_mut(32 * 4)) {
        let mut row_1 = _mm256_loadu_si256(src.as_ptr() as *const __m256i);
        let mut row_2 = _mm256_loadu_si256(src.as_ptr().add(32) as *const __m256i);
        let mut row_3 = _mm256_loadu_si256(src.as_ptr().add(64) as *const __m256i);
        let mut row_4 = _mm256_loadu_si256(src.as_ptr().add(96) as *const __m256i);

        row_1 = _mm256_shuffle_epi8(row_1, q_table_avx);
        row_2 = _mm256_shuffle_epi8(row_2, q_table_avx);
        row_3 = _mm256_shuffle_epi8(row_3, q_table_avx);
        row_4 = _mm256_shuffle_epi8(row_4, q_table_avx);

        _mm256_storeu_si256(dst.as_mut_ptr() as *mut __m256i, row_1);
        _mm256_storeu_si256(dst.as_mut_ptr().add(32) as *mut __m256i, row_2);
        _mm256_storeu_si256(dst.as_mut_ptr().add(64) as *mut __m256i, row_3);
        _mm256_storeu_si256(dst.as_mut_ptr().add(96) as *mut __m256i, row_4);
    }

    let src = src.chunks_exact(32 * 4).remainder();
    let dst = dst.chunks_exact_mut(32 * 4).into_remainder();

    for (src, dst) in src.chunks_exact(32).zip(dst.chunks_exact_mut(32)) {
        let mut row_1 = _mm256_loadu_si256(src.as_ptr() as *const __m256i);

        row_1 = _mm256_shuffle_epi8(row_1, q_table_avx);

        _mm256_storeu_si256(dst.as_mut_ptr() as *mut __m256i, row_1);
    }

    let src = src.chunks_exact(32).remainder();
    let dst = dst.chunks_exact_mut(32).into_remainder();

    if !src.is_empty() && !dst.is_empty() {
        assert!(src.len() < 32);
        assert!(dst.len() < 32);
        let mut transient_src: [u8; 32] = [0; 32];
        let mut transient_dst: [u8; 32] = [0; 32];
        std::ptr::copy_nonoverlapping(src.as_ptr(), transient_src.as_mut_ptr(), src.len());
        let mut row_1 = _mm256_loadu_si256(transient_src.as_ptr() as *const __m256i);
        row_1 = _mm256_shuffle_epi8(row_1, q_table_avx);
        _mm256_storeu_si256(transient_dst.as_mut_ptr() as *mut _, row_1);
        std::ptr::copy_nonoverlapping(transient_dst.as_ptr(), dst.as_mut_ptr(), dst.len());
    }
}