moxcms/conversions/sse/
rgb_xyz_opt.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 */
29use crate::conversions::rgbxyz::TransformMatrixShaperOptimizedV;
30use crate::conversions::sse::rgb_xyz::SseAlignedU16;
31use crate::transform::PointeeSizeExpressible;
32use crate::{CmsError, Layout, TransformExecutor};
33use num_traits::AsPrimitive;
34#[cfg(target_arch = "x86")]
35use std::arch::x86::*;
36#[cfg(target_arch = "x86_64")]
37use std::arch::x86_64::*;
38
39pub(crate) struct TransformShaperRgbOptSse<
40    T: Clone + Copy + 'static + PointeeSizeExpressible + Default,
41    const SRC_LAYOUT: u8,
42    const DST_LAYOUT: u8,
43> {
44    pub(crate) profile: TransformMatrixShaperOptimizedV<T>,
45    pub(crate) bit_depth: usize,
46    pub(crate) gamma_lut: usize,
47}
48
49impl<
50    T: Clone + Copy + 'static + PointeeSizeExpressible + Default,
51    const SRC_LAYOUT: u8,
52    const DST_LAYOUT: u8,
53> TransformShaperRgbOptSse<T, SRC_LAYOUT, DST_LAYOUT>
54where
55    u32: AsPrimitive<T>,
56{
57    #[target_feature(enable = "sse4.1")]
58    unsafe fn transform_impl(&self, src: &[T], dst: &mut [T]) -> Result<(), CmsError> {
59        let src_cn = Layout::from(SRC_LAYOUT);
60        let dst_cn = Layout::from(DST_LAYOUT);
61        let src_channels = src_cn.channels();
62        let dst_channels = dst_cn.channels();
63
64        let mut temporary = SseAlignedU16([0; 8]);
65
66        if src.len() / src_channels != dst.len() / dst_channels {
67            return Err(CmsError::LaneSizeMismatch);
68        }
69        if src.len() % src_channels != 0 {
70            return Err(CmsError::LaneMultipleOfChannels);
71        }
72        if dst.len() % dst_channels != 0 {
73            return Err(CmsError::LaneMultipleOfChannels);
74        }
75
76        let t = self.profile.adaptation_matrix.transpose();
77
78        let scale = (self.gamma_lut - 1) as f32;
79        let max_colors: T = ((1 << self.bit_depth) - 1).as_();
80
81        // safety precondition for linearization table
82        if T::FINITE {
83            let cap = (1 << self.bit_depth) - 1;
84            assert!(self.profile.linear.len() >= cap);
85        } else {
86            assert!(self.profile.linear.len() >= T::NOT_FINITE_LINEAR_TABLE_SIZE);
87        }
88
89        let lut_lin = &self.profile.linear;
90
91        unsafe {
92            let m0 = _mm_setr_ps(t.v[0][0], t.v[0][1], t.v[0][2], 0f32);
93            let m1 = _mm_setr_ps(t.v[1][0], t.v[1][1], t.v[1][2], 0f32);
94            let m2 = _mm_setr_ps(t.v[2][0], t.v[2][1], t.v[2][2], 0f32);
95
96            let zeros = _mm_setzero_ps();
97
98            let v_scale = _mm_set1_ps(scale);
99
100            for (src, dst) in src
101                .chunks_exact(src_channels)
102                .zip(dst.chunks_exact_mut(dst_channels))
103            {
104                let rp = lut_lin.get_unchecked(src[src_cn.r_i()]._as_usize());
105                let gp = lut_lin.get_unchecked(src[src_cn.g_i()]._as_usize());
106                let bp = lut_lin.get_unchecked(src[src_cn.b_i()]._as_usize());
107
108                let mut r = _mm_load_ss(rp);
109                let mut g = _mm_load_ss(gp);
110                let mut b = _mm_load_ss(bp);
111                let a = if src_channels == 4 {
112                    src[src_cn.a_i()]
113                } else {
114                    max_colors
115                };
116
117                r = _mm_shuffle_ps::<0>(r, r);
118                g = _mm_shuffle_ps::<0>(g, g);
119                b = _mm_shuffle_ps::<0>(b, b);
120
121                let v0 = _mm_mul_ps(r, m0);
122                let v1 = _mm_mul_ps(g, m1);
123                let v2 = _mm_mul_ps(b, m2);
124
125                let mut v = _mm_add_ps(_mm_add_ps(v0, v1), v2);
126                v = _mm_max_ps(v, zeros);
127                v = _mm_mul_ps(v, v_scale);
128                v = _mm_min_ps(v, v_scale);
129
130                let zx = _mm_cvtps_epi32(v);
131                _mm_store_si128(temporary.0.as_mut_ptr() as *mut _, zx);
132
133                dst[dst_cn.r_i()] = self.profile.gamma[temporary.0[0] as usize];
134                dst[dst_cn.g_i()] = self.profile.gamma[temporary.0[2] as usize];
135                dst[dst_cn.b_i()] = self.profile.gamma[temporary.0[4] as usize];
136                if dst_channels == 4 {
137                    dst[dst_cn.a_i()] = a;
138                }
139            }
140        }
141
142        Ok(())
143    }
144}
145
146impl<
147    T: Clone + Copy + 'static + PointeeSizeExpressible + Default,
148    const SRC_LAYOUT: u8,
149    const DST_LAYOUT: u8,
150> TransformExecutor<T> for TransformShaperRgbOptSse<T, SRC_LAYOUT, DST_LAYOUT>
151where
152    u32: AsPrimitive<T>,
153{
154    fn transform(&self, src: &[T], dst: &mut [T]) -> Result<(), CmsError> {
155        unsafe { self.transform_impl(src, dst) }
156    }
157}
moxcms/conversions/sse/rgb_xyz_opt.rs

moxcms/conversions/sse/
rgb_xyz_opt.rs
