pic-scale 0.7.8

High performance image scaling
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

/*
 * Copyright (c) Radzivon Bartoshyk. 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::alpha_handle_f16::{premultiply_pixel_f16_row, unpremultiply_pixel_f16_row};
use crate::sse::f16_utils::{_mm_cvtph_psx, _mm_cvtps_phx};
use crate::sse::{sse_deinterleave_rgba_epi16, sse_interleave_rgba_epi16};
use core::f16;
#[cfg(target_arch = "x86")]
use std::arch::x86::*;
#[cfg(target_arch = "x86_64")]
use std::arch::x86_64::*;

pub(crate) fn sse_premultiply_alpha_rgba_f16(dst: &mut [f16], src: &[f16]) {
    unsafe {
        sse_premultiply_alpha_rgba_f16_regular(dst, src);
    }
}

#[target_feature(enable = "sse4.1")]
fn sse_premultiply_alpha_rgba_f16_regular(dst: &mut [f16], src: &[f16]) {
    sse_premultiply_alpha_rgba_row_f16_impl::<false>(dst, src);
}

#[inline(always)]
fn sse_premultiply_alpha_rgba_row_f16_impl<const F16C: bool>(dst: &mut [f16], src: &[f16]) {
    unsafe {
        let mut rem = dst;
        let mut src_rem = src;

        for (dst, src) in rem.chunks_exact_mut(8 * 4).zip(src_rem.chunks_exact(8 * 4)) {
            let src_ptr = src.as_ptr();
            let lane0 = _mm_loadu_si128(src_ptr as *const __m128i);
            let lane1 = _mm_loadu_si128(src_ptr.add(8) as *const __m128i);
            let lane2 = _mm_loadu_si128(src_ptr.add(16) as *const __m128i);
            let lane3 = _mm_loadu_si128(src_ptr.add(24) as *const __m128i);
            let pixel = sse_deinterleave_rgba_epi16(lane0, lane1, lane2, lane3);

            let low_alpha = _mm_cvtph_psx::<F16C>(pixel.3);
            let low_r = _mm_mul_ps(_mm_cvtph_psx::<F16C>(pixel.0), low_alpha);
            let low_g = _mm_mul_ps(_mm_cvtph_psx::<F16C>(pixel.1), low_alpha);
            let low_b = _mm_mul_ps(_mm_cvtph_psx::<F16C>(pixel.2), low_alpha);

            let high_alpha = _mm_cvtph_psx::<F16C>(_mm_srli_si128::<8>(pixel.3));
            let high_r = _mm_mul_ps(
                _mm_cvtph_psx::<F16C>(_mm_srli_si128::<8>(pixel.0)),
                high_alpha,
            );
            let high_g = _mm_mul_ps(
                _mm_cvtph_psx::<F16C>(_mm_srli_si128::<8>(pixel.1)),
                high_alpha,
            );
            let high_b = _mm_mul_ps(
                _mm_cvtph_psx::<F16C>(_mm_srli_si128::<8>(pixel.2)),
                high_alpha,
            );
            let r_values =
                _mm_unpacklo_epi64(_mm_cvtps_phx::<F16C>(low_r), _mm_cvtps_phx::<F16C>(high_r));
            let g_values =
                _mm_unpacklo_epi64(_mm_cvtps_phx::<F16C>(low_g), _mm_cvtps_phx::<F16C>(high_g));
            let b_values =
                _mm_unpacklo_epi64(_mm_cvtps_phx::<F16C>(low_b), _mm_cvtps_phx::<F16C>(high_b));
            let dst_ptr = dst.as_mut_ptr();
            let (d_lane0, d_lane1, d_lane2, d_lane3) =
                sse_interleave_rgba_epi16(r_values, g_values, b_values, pixel.3);
            _mm_storeu_si128(dst_ptr as *mut __m128i, d_lane0);
            _mm_storeu_si128(dst_ptr.add(8) as *mut __m128i, d_lane1);
            _mm_storeu_si128(dst_ptr.add(16) as *mut __m128i, d_lane2);
            _mm_storeu_si128(dst_ptr.add(24) as *mut __m128i, d_lane3);
        }

        rem = rem.chunks_exact_mut(8 * 4).into_remainder();
        src_rem = src_rem.chunks_exact(8 * 4).remainder();

        premultiply_pixel_f16_row(rem, src_rem);
    }
}

pub(crate) fn sse_unpremultiply_alpha_rgba_f16(in_place: &mut [f16]) {
    unsafe {
        if is_x86_feature_detected!("f16c") {
            sse_unpremultiply_alpha_rgba_f16_regular(in_place);
        } else {
            sse_unpremultiply_alpha_rgba_f16c(in_place);
        }
    }
}

#[target_feature(enable = "sse4.1")]
fn sse_unpremultiply_alpha_rgba_f16_regular(in_place: &mut [f16]) {
    sse_unpremultiply_alpha_rgba_f16_row_impl::<false>(in_place);
}

#[target_feature(enable = "sse4.1", enable = "f16c")]
fn sse_unpremultiply_alpha_rgba_f16c(in_place: &mut [f16]) {
    sse_unpremultiply_alpha_rgba_f16_row_impl::<true>(in_place);
}

#[inline(always)]
fn sse_unpremultiply_alpha_rgba_f16_row_impl<const F16C: bool>(in_place: &mut [f16]) {
    unsafe {
        let mut rem = in_place;

        for dst in rem.chunks_exact_mut(8 * 4) {
            let src_ptr = dst.as_ptr();
            let lane0 = _mm_loadu_si128(src_ptr as *const __m128i);
            let lane1 = _mm_loadu_si128(src_ptr.add(8) as *const __m128i);
            let lane2 = _mm_loadu_si128(src_ptr.add(16) as *const __m128i);
            let lane3 = _mm_loadu_si128(src_ptr.add(24) as *const __m128i);
            let pixel = sse_deinterleave_rgba_epi16(lane0, lane1, lane2, lane3);

            let low_alpha = _mm_cvtph_psx::<F16C>(pixel.3);
            let zeros = _mm_setzero_ps();
            let low_alpha_zero_mask = _mm_cmpeq_ps(low_alpha, zeros);
            let low_r = _mm_blendv_ps(
                _mm_mul_ps(_mm_cvtph_psx::<F16C>(pixel.0), low_alpha),
                zeros,
                low_alpha_zero_mask,
            );
            let low_g = _mm_blendv_ps(
                _mm_mul_ps(_mm_cvtph_psx::<F16C>(pixel.1), low_alpha),
                zeros,
                low_alpha_zero_mask,
            );
            let low_b = _mm_blendv_ps(
                _mm_mul_ps(_mm_cvtph_psx::<F16C>(pixel.2), low_alpha),
                zeros,
                low_alpha_zero_mask,
            );

            let high_alpha = _mm_cvtph_psx::<F16C>(_mm_srli_si128::<8>(pixel.3));
            let high_alpha_zero_mask = _mm_cmpeq_ps(high_alpha, zeros);
            let high_r = _mm_blendv_ps(
                _mm_mul_ps(
                    _mm_cvtph_psx::<F16C>(_mm_srli_si128::<8>(pixel.0)),
                    high_alpha,
                ),
                zeros,
                high_alpha_zero_mask,
            );
            let high_g = _mm_blendv_ps(
                _mm_mul_ps(
                    _mm_cvtph_psx::<F16C>(_mm_srli_si128::<8>(pixel.1)),
                    high_alpha,
                ),
                zeros,
                high_alpha_zero_mask,
            );
            let high_b = _mm_blendv_ps(
                _mm_mul_ps(
                    _mm_cvtph_psx::<F16C>(_mm_srli_si128::<8>(pixel.2)),
                    high_alpha,
                ),
                zeros,
                high_alpha_zero_mask,
            );
            let r_values =
                _mm_unpacklo_epi64(_mm_cvtps_phx::<F16C>(low_r), _mm_cvtps_phx::<F16C>(high_r));
            let g_values =
                _mm_unpacklo_epi64(_mm_cvtps_phx::<F16C>(low_g), _mm_cvtps_phx::<F16C>(high_g));
            let b_values =
                _mm_unpacklo_epi64(_mm_cvtps_phx::<F16C>(low_b), _mm_cvtps_phx::<F16C>(high_b));
            let dst_ptr = dst.as_mut_ptr();
            let (d_lane0, d_lane1, d_lane2, d_lane3) =
                sse_interleave_rgba_epi16(r_values, g_values, b_values, pixel.3);
            _mm_storeu_si128(dst_ptr as *mut __m128i, d_lane0);
            _mm_storeu_si128(dst_ptr.add(8) as *mut __m128i, d_lane1);
            _mm_storeu_si128(dst_ptr.add(16) as *mut __m128i, d_lane2);
            _mm_storeu_si128(dst_ptr.add(24) as *mut __m128i, d_lane3);
        }

        rem = rem.chunks_exact_mut(8 * 4).into_remainder();

        unpremultiply_pixel_f16_row(rem);
    }
}