ass-renderer 0.1.2

High-performance ASS subtitle renderer with modular backends
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

//! Low-level source-over blend helpers for coverage compositing.
//!
//! Fixed-point `a * b / 255` scaling plus per-row and per-pixel source-over
//! blends, with a `wide`-accelerated row blend behind the `simd` feature and a
//! scalar fallback. Shared by the compositing routines in
//! [`super::composite`].

/// Rounded fixed-point `a * b / 255` for `a, b` in `0..=255`.
///
/// The `((t + (t >> 8)) >> 8)` form is bit-identical to `(a*b + 127) / 255` over
/// that range but maps directly onto SIMD lanes (no per-lane division), so the
/// scalar and SIMD composite paths produce identical pixels.
#[inline]
pub(super) fn mul255(a: u16, b: u16) -> u16 {
    let t = a * b + 128;
    (t + (t >> 8)) >> 8
}

#[cfg(feature = "simd")]
#[inline]
fn mul255x16(a: wide::u16x16, b: wide::u16x16) -> wide::u16x16 {
    let t = a * b + 128_u16;
    (t + (t >> 8)) >> 8
}

/// Source-over blend of one tile row (premultiplied straight `color`, premult
/// channels `pr/pg/pb/pa`) onto `dst` starting at byte `dst_start`. Four pixels
/// at a time with `wide` when the `simd` feature is on; empty four-pixel groups
/// are skipped (text coverage is mostly empty).
#[cfg(feature = "simd")]
#[inline]
pub(super) fn blend_row(
    dst: &mut [u8],
    dst_start: usize,
    cov_row: &[u8],
    pr: u16,
    pg: u16,
    pb: u16,
    pa: u16,
) {
    use wide::u16x16;
    let pcolor = u16x16::new([
        pr, pg, pb, pa, pr, pg, pb, pa, pr, pg, pb, pa, pr, pg, pb, pa,
    ]);
    let run = cov_row.len();
    let mut t = 0;
    while t + 4 <= run {
        let (c0, c1, c2, c3) = (cov_row[t], cov_row[t + 1], cov_row[t + 2], cov_row[t + 3]);
        if (c0 | c1 | c2 | c3) != 0 {
            let cov = u16x16::new([
                u16::from(c0),
                u16::from(c0),
                u16::from(c0),
                u16::from(c0),
                u16::from(c1),
                u16::from(c1),
                u16::from(c1),
                u16::from(c1),
                u16::from(c2),
                u16::from(c2),
                u16::from(c2),
                u16::from(c2),
                u16::from(c3),
                u16::from(c3),
                u16::from(c3),
                u16::from(c3),
            ]);
            let src = mul255x16(pcolor, cov);
            let s = src.to_array();
            let inv = u16x16::new([
                255 - s[3],
                255 - s[3],
                255 - s[3],
                255 - s[3],
                255 - s[7],
                255 - s[7],
                255 - s[7],
                255 - s[7],
                255 - s[11],
                255 - s[11],
                255 - s[11],
                255 - s[11],
                255 - s[15],
                255 - s[15],
                255 - s[15],
                255 - s[15],
            ]);
            let di = dst_start + t * 4;
            let d = &dst[di..di + 16];
            let dpix = u16x16::new([
                u16::from(d[0]),
                u16::from(d[1]),
                u16::from(d[2]),
                u16::from(d[3]),
                u16::from(d[4]),
                u16::from(d[5]),
                u16::from(d[6]),
                u16::from(d[7]),
                u16::from(d[8]),
                u16::from(d[9]),
                u16::from(d[10]),
                u16::from(d[11]),
                u16::from(d[12]),
                u16::from(d[13]),
                u16::from(d[14]),
                u16::from(d[15]),
            ]);
            let out = (src + mul255x16(dpix, inv)).to_array();
            for (slot, &v) in dst[di..di + 16].iter_mut().zip(out.iter()) {
                *slot = v as u8;
            }
        }
        t += 4;
    }
    while t < run {
        blend_pixel(
            dst,
            dst_start + t * 4,
            u16::from(cov_row[t]),
            pr,
            pg,
            pb,
            pa,
        );
        t += 1;
    }
}

/// Scalar fallback when the `simd` feature is off.
#[cfg(not(feature = "simd"))]
#[inline]
pub(super) fn blend_row(
    dst: &mut [u8],
    dst_start: usize,
    cov_row: &[u8],
    pr: u16,
    pg: u16,
    pb: u16,
    pa: u16,
) {
    for (t, &c) in cov_row.iter().enumerate() {
        blend_pixel(dst, dst_start + t * 4, u16::from(c), pr, pg, pb, pa);
    }
}

/// Source-over one premultiplied RGBA pixel by coverage `cov`.
#[inline]
fn blend_pixel(dst: &mut [u8], di: usize, cov: u16, pr: u16, pg: u16, pb: u16, pa: u16) {
    if cov == 0 {
        return;
    }
    let inv = 255 - mul255(pa, cov);
    dst[di] = (mul255(pr, cov) + mul255(u16::from(dst[di]), inv)) as u8;
    dst[di + 1] = (mul255(pg, cov) + mul255(u16::from(dst[di + 1]), inv)) as u8;
    dst[di + 2] = (mul255(pb, cov) + mul255(u16::from(dst[di + 2]), inv)) as u8;
    dst[di + 3] = (mul255(pa, cov) + mul255(u16::from(dst[di + 3]), inv)) as u8;
}