fast_image_resize 6.1.0

Library for fast image resizing with using of SIMD instructions
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

use core::arch::x86_64::*;

use super::sse4;
use crate::convolution::optimisations::{CoefficientsI16Chunk, Normalizer16};
use crate::image_view::ImageViewMut;
use crate::pixels::InnerPixel;
use crate::{simd_utils, ImageView};

#[inline]
pub(crate) fn vert_convolution<T>(
    src_view: &impl ImageView<Pixel = T>,
    dst_view: &mut impl ImageViewMut<Pixel = T>,
    offset: u32,
    normalizer: &Normalizer16,
) where
    T: InnerPixel<Component = u8>,
{
    let precision = normalizer.precision();

    macro_rules! call {
        ($imm8:expr) => {{
            vert_convolution_p::<T, $imm8>(src_view, dst_view, offset, normalizer);
        }};
    }
    constify_imm8!(precision, call);
}

fn vert_convolution_p<T, const PRECISION: i32>(
    src_view: &impl ImageView<Pixel = T>,
    dst_view: &mut impl ImageViewMut<Pixel = T>,
    offset: u32,
    normalizer: &Normalizer16,
) where
    T: InnerPixel<Component = u8>,
{
    let coefficients_chunks = normalizer.chunks();
    let src_x = offset as usize * T::count_of_components();
    let dst_rows = dst_view.iter_rows_mut(0);
    let dst_row_and_coefs = dst_rows.zip(coefficients_chunks);

    for (dst_row, coeffs_chunk) in dst_row_and_coefs {
        unsafe {
            vert_convolution_into_one_row::<T, PRECISION>(
                src_view,
                dst_row,
                src_x,
                coeffs_chunk,
                normalizer,
            );
        }
    }
}

#[inline]
#[target_feature(enable = "avx2")]
unsafe fn vert_convolution_into_one_row<T, const PRECISION: i32>(
    src_view: &impl ImageView<Pixel = T>,
    dst_row: &mut [T],
    mut src_x: usize,
    coeffs_chunk: &CoefficientsI16Chunk,
    normalizer: &Normalizer16,
) where
    T: InnerPixel<Component = u8>,
{
    let y_start = coeffs_chunk.start;
    let coeffs = coeffs_chunk.values();
    let max_rows = coeffs.len() as u32;
    let y_last = (y_start + max_rows).max(1) - 1;
    let mut dst_u8 = T::components_mut(dst_row);

    let initial_256 = _mm256_set1_epi32(1 << (PRECISION as u8 - 1));

    let mut dst_chunks_64 = dst_u8.chunks_exact_mut(64);
    for dst_chunk in &mut dst_chunks_64 {
        let mut sss0 = [initial_256; 4];
        let mut sss1 = [initial_256; 4];

        let coeffs_chunks = coeffs.chunks_exact(2);
        let coeffs_reminder = coeffs_chunks.remainder();

        for (src_rows, two_coeffs) in src_view.iter_2_rows(y_start, max_rows).zip(coeffs_chunks) {
            let components1 = T::components(src_rows[0]);
            let components2 = T::components(src_rows[1]);

            // Load two coefficients at once
            let mmk = simd_utils::mm256_load_and_clone_i16x2(two_coeffs);

            conv_32_components_two_rows(components1, components2, src_x, &mut sss0, mmk);
            conv_32_components_two_rows(components1, components2, src_x + 32, &mut sss1, mmk);
        }

        if let Some(&k) = coeffs_reminder.first() {
            if let Some(s_row) = src_view.iter_rows(y_last).next() {
                let components = T::components(s_row);
                let mmk = _mm256_set1_epi32(k as i32);

                conv_32_components_one_row(components, src_x, &mut sss0, mmk);
                conv_32_components_one_row(components, src_x + 32, &mut sss1, mmk);
            }
        }

        let (dst0, dst1) = dst_chunk.split_at_mut(32);
        for (mut sss, dst) in [(sss0, dst0), (sss1, dst1)] {
            sss = sss.map(|v| _mm256_srai_epi32::<PRECISION>(v));
            let half0 = _mm256_packs_epi32(sss[0], sss[1]);
            let half1 = _mm256_packs_epi32(sss[2], sss[3]);
            let components = _mm256_packus_epi16(half0, half1);
            let dst_ptr = dst.as_mut_ptr() as *mut __m256i;
            _mm256_storeu_si256(dst_ptr, components);
        }

        src_x += 64;
    }

    // 32 components in one register
    dst_u8 = dst_chunks_64.into_remainder();
    let mut dst_chunks_32 = dst_u8.chunks_exact_mut(32);
    for dst_chunk in &mut dst_chunks_32 {
        let mut sss = [initial_256; 4];

        let coeffs_chunks = coeffs.chunks_exact(2);
        let coeffs_reminder = coeffs_chunks.remainder();

        for (src_rows, two_coeffs) in src_view.iter_2_rows(y_start, max_rows).zip(coeffs_chunks) {
            let components1 = T::components(src_rows[0]); // top line
            let components2 = T::components(src_rows[1]); // bottom line

            // Load two coefficients at once
            let mmk = simd_utils::mm256_load_and_clone_i16x2(two_coeffs);

            conv_32_components_two_rows(components1, components2, src_x, &mut sss, mmk);
        }

        if let Some(&k) = coeffs_reminder.first() {
            if let Some(s_row) = src_view.iter_rows(y_last).next() {
                let components = T::components(s_row);
                let mmk = _mm256_set1_epi32(k as i32);

                conv_32_components_one_row(components, src_x, &mut sss, mmk);
            }
        }

        let dst_ptr = dst_chunk.as_mut_ptr() as *mut __m256i;
        sss = sss.map(|v| _mm256_srai_epi32::<PRECISION>(v));
        let half0 = _mm256_packs_epi32(sss[0], sss[1]);
        let half1 = _mm256_packs_epi32(sss[2], sss[3]);
        let dst = _mm256_packus_epi16(half0, half1);
        _mm256_storeu_si256(dst_ptr, dst);

        src_x += 32;
    }

    dst_u8 = dst_chunks_32.into_remainder();
    sse4::conv_less_that_32_components::<T, PRECISION>(
        src_view,
        dst_u8,
        src_x,
        coeffs_chunk,
        normalizer,
    );
}

#[inline]
#[target_feature(enable = "avx2")]
unsafe fn conv_32_components_two_rows(
    components1: &[u8],
    components2: &[u8],
    src_x: usize,
    sss: &mut [__m256i; 4],
    mmk: __m256i,
) {
    let zero_256 = _mm256_setzero_si256();
    let source1 = simd_utils::loadu_si256(components1, src_x); // top line
    let source2 = simd_utils::loadu_si256(components2, src_x); // bottom line
    conv_loaded_32_components(source1, source2, sss, mmk, zero_256);
}

#[inline]
#[target_feature(enable = "avx2")]
unsafe fn conv_32_components_one_row(
    components1: &[u8],
    src_x: usize,
    sss: &mut [__m256i; 4],
    mmk: __m256i,
) {
    let zero_256 = _mm256_setzero_si256();
    let source1 = simd_utils::loadu_si256(components1, src_x); // top line
    conv_loaded_32_components(source1, zero_256, sss, mmk, zero_256);
}

#[inline]
#[target_feature(enable = "avx2")]
unsafe fn conv_loaded_32_components(
    source1: __m256i,
    source2: __m256i,
    sss: &mut [__m256i; 4],
    mmk: __m256i,
    zero_256: __m256i,
) {
    let source = _mm256_unpacklo_epi8(source1, source2);
    let pix = _mm256_unpacklo_epi8(source, zero_256);
    sss[0] = _mm256_add_epi32(sss[0], _mm256_madd_epi16(pix, mmk));
    let pix = _mm256_unpackhi_epi8(source, zero_256);
    sss[1] = _mm256_add_epi32(sss[1], _mm256_madd_epi16(pix, mmk));

    let source = _mm256_unpackhi_epi8(source1, source2);
    let pix = _mm256_unpacklo_epi8(source, zero_256);
    sss[2] = _mm256_add_epi32(sss[2], _mm256_madd_epi16(pix, mmk));
    let pix = _mm256_unpackhi_epi8(source, zero_256);
    sss[3] = _mm256_add_epi32(sss[3], _mm256_madd_epi16(pix, mmk));
}