moxcms/conversions/sse/
rgb_xyz.rs

1/*
2 * // Copyright (c) Radzivon Bartoshyk 3/2025. All rights reserved.
3 * //
4 * // Redistribution and use in source and binary forms, with or without modification,
5 * // are permitted provided that the following conditions are met:
6 * //
7 * // 1.  Redistributions of source code must retain the above copyright notice, this
8 * // list of conditions and the following disclaimer.
9 * //
10 * // 2.  Redistributions in binary form must reproduce the above copyright notice,
11 * // this list of conditions and the following disclaimer in the documentation
12 * // and/or other materials provided with the distribution.
13 * //
14 * // 3.  Neither the name of the copyright holder nor the names of its
15 * // contributors may be used to endorse or promote products derived from
16 * // this software without specific prior written permission.
17 * //
18 * // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
19 * // AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
21 * // DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
22 * // FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * // DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
24 * // SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
25 * // CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
26 * // OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
27 * // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28 */
29#![cfg(feature = "sse_shaper_paths")]
30use crate::conversions::TransformMatrixShaper;
31use crate::conversions::sse::SseAlignedU16;
32use crate::transform::PointeeSizeExpressible;
33use crate::{CmsError, Layout, TransformExecutor};
34use num_traits::AsPrimitive;
35#[cfg(target_arch = "x86")]
36use std::arch::x86::*;
37#[cfg(target_arch = "x86_64")]
38use std::arch::x86_64::*;
39
40pub(crate) struct TransformShaperRgbSse<
41    T: Clone + Copy + 'static + PointeeSizeExpressible + Default,
42    const SRC_LAYOUT: u8,
43    const DST_LAYOUT: u8,
44    // deleting linear cap is in effective here
45    const LINEAR_CAP: usize,
46> {
47    // removing linear cap here is not worth it, at least in previous attempts
48    pub(crate) profile: TransformMatrixShaper<T, LINEAR_CAP>,
49    pub(crate) bit_depth: usize,
50    pub(crate) gamma_lut: usize,
51}
52
53impl<
54    T: Clone + Copy + 'static + PointeeSizeExpressible + Default,
55    const SRC_LAYOUT: u8,
56    const DST_LAYOUT: u8,
57    const LINEAR_CAP: usize,
58> TransformShaperRgbSse<T, SRC_LAYOUT, DST_LAYOUT, LINEAR_CAP>
59where
60    u32: AsPrimitive<T>,
61{
62    #[target_feature(enable = "sse4.1")]
63    unsafe fn transform_impl(&self, src: &[T], dst: &mut [T]) -> Result<(), CmsError> {
64        let src_cn = Layout::from(SRC_LAYOUT);
65        let dst_cn = Layout::from(DST_LAYOUT);
66        let src_channels = src_cn.channels();
67        let dst_channels = dst_cn.channels();
68
69        let mut temporary = SseAlignedU16([0; 8]);
70
71        if src.len() / src_channels != dst.len() / dst_channels {
72            return Err(CmsError::LaneSizeMismatch);
73        }
74        if src.len() % src_channels != 0 {
75            return Err(CmsError::LaneMultipleOfChannels);
76        }
77        if dst.len() % dst_channels != 0 {
78            return Err(CmsError::LaneMultipleOfChannels);
79        }
80
81        let t = self.profile.adaptation_matrix.transpose();
82
83        let scale = (self.gamma_lut - 1) as f32;
84        let max_colors: T = ((1 << self.bit_depth) - 1).as_();
85
86        unsafe {
87            let m0 = _mm_setr_ps(t.v[0][0], t.v[0][1], t.v[0][2], 0f32);
88            let m1 = _mm_setr_ps(t.v[1][0], t.v[1][1], t.v[1][2], 0f32);
89            let m2 = _mm_setr_ps(t.v[2][0], t.v[2][1], t.v[2][2], 0f32);
90
91            let zeros = _mm_setzero_ps();
92
93            let v_scale = _mm_set1_ps(scale);
94
95            for (src, dst) in src
96                .chunks_exact(src_channels)
97                .zip(dst.chunks_exact_mut(dst_channels))
98            {
99                let rp = &self.profile.r_linear[src[src_cn.r_i()]._as_usize()];
100                let gp = &self.profile.g_linear[src[src_cn.g_i()]._as_usize()];
101                let bp = &self.profile.b_linear[src[src_cn.b_i()]._as_usize()];
102
103                let mut r = _mm_load_ss(rp);
104                let mut g = _mm_load_ss(gp);
105                let mut b = _mm_load_ss(bp);
106                let a = if src_channels == 4 {
107                    src[src_cn.a_i()]
108                } else {
109                    max_colors
110                };
111
112                r = _mm_shuffle_ps::<0>(r, r);
113                g = _mm_shuffle_ps::<0>(g, g);
114                b = _mm_shuffle_ps::<0>(b, b);
115
116                let v0 = _mm_mul_ps(r, m0);
117                let v1 = _mm_mul_ps(g, m1);
118                let v2 = _mm_mul_ps(b, m2);
119
120                let mut v = _mm_add_ps(_mm_add_ps(v0, v1), v2);
121                v = _mm_max_ps(v, zeros);
122                v = _mm_mul_ps(v, v_scale);
123                v = _mm_min_ps(v, v_scale);
124
125                let zx = _mm_cvtps_epi32(v);
126                _mm_store_si128(temporary.0.as_mut_ptr() as *mut _, zx);
127
128                dst[dst_cn.r_i()] = self.profile.r_gamma[temporary.0[0] as usize];
129                dst[dst_cn.g_i()] = self.profile.g_gamma[temporary.0[2] as usize];
130                dst[dst_cn.b_i()] = self.profile.b_gamma[temporary.0[4] as usize];
131                if dst_channels == 4 {
132                    dst[dst_cn.a_i()] = a;
133                }
134            }
135        }
136
137        Ok(())
138    }
139}
140
141impl<
142    T: Clone + Copy + 'static + PointeeSizeExpressible + Default,
143    const SRC_LAYOUT: u8,
144    const DST_LAYOUT: u8,
145    const LINEAR_CAP: usize,
146> TransformExecutor<T> for TransformShaperRgbSse<T, SRC_LAYOUT, DST_LAYOUT, LINEAR_CAP>
147where
148    u32: AsPrimitive<T>,
149{
150    fn transform(&self, src: &[T], dst: &mut [T]) -> Result<(), CmsError> {
151        unsafe { self.transform_impl(src, dst) }
152    }
153}
moxcms/conversions/sse/rgb_xyz.rs

moxcms/conversions/sse/
rgb_xyz.rs
