colconv 0.1.0

SIMD-dispatched color-conversion kernels covering the FFmpeg AVPixelFormat space, with a Sink-based API so consumers pick which derived outputs (RGB / Luma / HSV / custom) they want without paying for the ones they don't.
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

//! 8-bit YUV 4:1:0 → RGB / RGBA dispatchers (`yuv_410_to_rgb_row`,
//! `yuv_410_to_rgba_row`). Tier 1 P3 legacy (Cinepak / Sorenson).
//!
//! Backends: scalar (always) + NEON (aarch64) + SSE4.1 / AVX2 /
//! AVX-512 (x86_64) + simd128 (wasm32). Each SIMD backend follows
//! the same `block_size_y` / 4× horizontal chroma fan-out shape:
//! NEON / SSE4.1 / wasm 16 Y per iter, AVX2 32 Y, AVX-512 64 Y.
//! Math is byte-identical to scalar by construction.

#[cfg(any(
  target_arch = "aarch64",
  target_arch = "x86_64",
  target_arch = "wasm32"
))]
use crate::row::arch;
#[cfg(target_arch = "aarch64")]
use crate::row::neon_available;
#[cfg(target_arch = "wasm32")]
use crate::row::simd128_available;
#[cfg(target_arch = "x86_64")]
use crate::row::{avx2_available, avx512_available, sse41_available};
use crate::{
  ColorMatrix,
  row::{rgb_row_bytes, rgba_row_bytes, scalar},
};

/// Converts one row of 4:1:0 YUV to packed RGB.
///
/// Dispatches to the best available backend for the current target.
/// See `scalar::yuv_410_to_rgb_row` for the full semantic
/// specification (range handling, matrix definitions, output layout).
///
/// `use_simd = false` forces the scalar reference path.
#[cfg_attr(not(tarpaulin), inline(always))]
#[allow(clippy::too_many_arguments)]
pub fn yuv_410_to_rgb_row(
  y: &[u8],
  u_quarter: &[u8],
  v_quarter: &[u8],
  rgb_out: &mut [u8],
  width: usize,
  matrix: ColorMatrix,
  full_range: bool,
  use_simd: bool,
) {
  // Runtime asserts at the dispatcher boundary — same rationale as the
  // 4:2:0 sibling. Unsafe SIMD kernels rely on these in release builds.
  assert_eq!(width & 3, 0, "YUV 4:1:0 requires width % 4 == 0");
  let rgb_min = rgb_row_bytes(width);
  assert!(y.len() >= width, "y row too short");
  assert!(u_quarter.len() >= width / 4, "u_quarter row too short");
  assert!(v_quarter.len() >= width / 4, "v_quarter row too short");
  assert!(rgb_out.len() >= rgb_min, "rgb_out row too short");

  if use_simd {
    cfg_select! {
      target_arch = "aarch64" => {
        if neon_available() {
          // SAFETY: `neon_available()` verified NEON is present.
          // Bounds / parity invariants are the caller's obligation.
          unsafe {
            arch::neon::yuv_410_to_rgb_row(y, u_quarter, v_quarter, rgb_out, width, matrix, full_range);
          }
          return;
        }
      },
      target_arch = "x86_64" => {
        if avx512_available() {
          // SAFETY: `avx512_available()` verified AVX-512BW is present.
          unsafe {
            arch::x86_avx512::yuv_410_to_rgb_row(
              y, u_quarter, v_quarter, rgb_out, width, matrix, full_range,
            );
          }
          return;
        }
        if avx2_available() {
          // SAFETY: `avx2_available()` verified AVX2 is present.
          unsafe {
            arch::x86_avx2::yuv_410_to_rgb_row(
              y, u_quarter, v_quarter, rgb_out, width, matrix, full_range,
            );
          }
          return;
        }
        if sse41_available() {
          // SAFETY: `sse41_available()` verified SSE4.1 is present.
          unsafe {
            arch::x86_sse41::yuv_410_to_rgb_row(
              y, u_quarter, v_quarter, rgb_out, width, matrix, full_range,
            );
          }
          return;
        }
      },
      target_arch = "wasm32" => {
        if simd128_available() {
          // SAFETY: simd128 compile-time availability verified.
          unsafe {
            arch::wasm_simd128::yuv_410_to_rgb_row(
              y, u_quarter, v_quarter, rgb_out, width, matrix, full_range,
            );
          }
          return;
        }
      },
      _ => {
        // Targets without a SIMD backend (riscv64, powerpc, …) fall
        // through to scalar.
      }
    }
  }

  scalar::yuv_410_to_rgb_row(y, u_quarter, v_quarter, rgb_out, width, matrix, full_range);
}

/// Converts one row of 4:1:0 YUV to packed **RGBA** (8-bit).
///
/// Same numerical contract as [`yuv_410_to_rgb_row`]; the only
/// differences are the per-pixel stride (4 vs 3) and the alpha byte
/// (`0xFF`, opaque, for every pixel — sources without an alpha plane
/// produce opaque output). The first three bytes per pixel are
/// byte-identical to what [`yuv_410_to_rgb_row`] would write.
///
/// `rgba_out.len() >= 4 * width`. `use_simd = false` forces the
/// scalar reference path.
#[cfg_attr(not(tarpaulin), inline(always))]
#[allow(clippy::too_many_arguments)]
pub fn yuv_410_to_rgba_row(
  y: &[u8],
  u_quarter: &[u8],
  v_quarter: &[u8],
  rgba_out: &mut [u8],
  width: usize,
  matrix: ColorMatrix,
  full_range: bool,
  use_simd: bool,
) {
  assert_eq!(width & 3, 0, "YUV 4:1:0 requires width % 4 == 0");
  let rgba_min = rgba_row_bytes(width);
  assert!(y.len() >= width, "y row too short");
  assert!(u_quarter.len() >= width / 4, "u_quarter row too short");
  assert!(v_quarter.len() >= width / 4, "v_quarter row too short");
  assert!(rgba_out.len() >= rgba_min, "rgba_out row too short");

  if use_simd {
    cfg_select! {
      target_arch = "aarch64" => {
        if neon_available() {
          // SAFETY: `neon_available()` verified NEON is present.
          unsafe {
            arch::neon::yuv_410_to_rgba_row(
              y, u_quarter, v_quarter, rgba_out, width, matrix, full_range,
            );
          }
          return;
        }
      },
      target_arch = "x86_64" => {
        if avx512_available() {
          // SAFETY: `avx512_available()` verified AVX-512BW is present.
          unsafe {
            arch::x86_avx512::yuv_410_to_rgba_row(
              y, u_quarter, v_quarter, rgba_out, width, matrix, full_range,
            );
          }
          return;
        }
        if avx2_available() {
          // SAFETY: `avx2_available()` verified AVX2 is present.
          unsafe {
            arch::x86_avx2::yuv_410_to_rgba_row(
              y, u_quarter, v_quarter, rgba_out, width, matrix, full_range,
            );
          }
          return;
        }
        if sse41_available() {
          // SAFETY: `sse41_available()` verified SSE4.1 is present.
          unsafe {
            arch::x86_sse41::yuv_410_to_rgba_row(
              y, u_quarter, v_quarter, rgba_out, width, matrix, full_range,
            );
          }
          return;
        }
      },
      target_arch = "wasm32" => {
        if simd128_available() {
          // SAFETY: simd128 compile-time availability verified.
          unsafe {
            arch::wasm_simd128::yuv_410_to_rgba_row(
              y, u_quarter, v_quarter, rgba_out, width, matrix, full_range,
            );
          }
          return;
        }
      },
      _ => {
        // Targets without a SIMD backend fall through to scalar.
      }
    }
  }

  scalar::yuv_410_to_rgba_row(y, u_quarter, v_quarter, rgba_out, width, matrix, full_range);
}