threers-codec 0.1.0

Pure-Rust media codecs: H.264/HEVC/VP9 encoders, H.264/HEVC/VP9/AV1 decoders, AAC/MP3/Opus/Vorbis/FLAC/WAV decoders, AAC-LC and MP3 encoders, MP4/WebM/Matroska/Ogg, GIF/APNG, captions, and VideoToolbox acceleration on macOS
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
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
//! Loop-filter edge masks and levels (dav1d `src/lf_mask.c`).

use super::obu::{FrameHeader, Layout, LoopfilterModeRefDeltas};
use super::tables::{BLOCK_DIMENSIONS, TXFM_DIMENSIONS};

#[derive(Clone, Copy)]
pub struct Av1FilterLut {
    pub e: [u8; 64],
    pub i: [u8; 64],
    pub sharp: [u64; 2],
}

impl Default for Av1FilterLut {
    fn default() -> Self {
        Av1FilterLut {
            e: [0; 64],
            i: [0; 64],
            sharp: [0; 2],
        }
    }
}

#[derive(Clone, Copy, Default, Debug, PartialEq)]
pub struct RestorationUnit {
    /// SGR: `RESTORATION_SGRPROJ + sgr_idx`.
    pub typ: u8,
    pub filter_h: [i8; 3],
    pub filter_v: [i8; 3],
    pub sgr_weights: [i8; 2],
}

/// One 128x128 area (1 or 4 superblocks), pre-superres.
#[derive(Clone, Copy, Default)]
pub struct Av1Filter {
    /// `[col/row][pos][tx class][half]`, each bit one 4px line.
    pub filter_y: [[[[u16; 2]; 3]; 32]; 2],
    pub filter_uv: [[[[u16; 2]; 2]; 32]; 2],
    pub cdef_idx: [i8; 4],
    pub noskip_mask: [[u16; 2]; 16],
}

/// One 128x128 area, post-superres.
#[derive(Clone, Copy, Default)]
pub struct Av1Restoration {
    pub lr: [[RestorationUnit; 4]; 3],
}

/// `txa[edge][txsz|step][y][x]`.
type Txa = [[[[u8; 32]; 32]; 2]; 2];

#[allow(clippy::too_many_arguments)]
fn decomp_tx(
    txa: &mut Txa,
    oy: usize,
    ox: usize,
    from: usize,
    depth: usize,
    y_off: usize,
    x_off: usize,
    tx_masks: &[u16; 2],
) {
    let t = &TXFM_DIMENSIONS[from];
    let is_split = if from == 0 || depth > 1 {
        false
    } else {
        (tx_masks[depth] >> (y_off * 4 + x_off)) & 1 != 0
    };
    if is_split {
        let sub = t.sub as usize;
        let htw4 = (t.w >> 1) as usize;
        let hth4 = (t.h >> 1) as usize;
        decomp_tx(txa, oy, ox, sub, depth + 1, y_off * 2, x_off * 2, tx_masks);
        if t.w >= t.h {
            decomp_tx(
                txa,
                oy,
                ox + htw4,
                sub,
                depth + 1,
                y_off * 2,
                x_off * 2 + 1,
                tx_masks,
            );
        }
        if t.h >= t.w {
            decomp_tx(
                txa,
                oy + hth4,
                ox,
                sub,
                depth + 1,
                y_off * 2 + 1,
                x_off * 2,
                tx_masks,
            );
            if t.w >= t.h {
                decomp_tx(
                    txa,
                    oy + hth4,
                    ox + htw4,
                    sub,
                    depth + 1,
                    y_off * 2 + 1,
                    x_off * 2 + 1,
                    tx_masks,
                );
            }
        }
    } else {
        let lw = t.lw.min(2);
        let lh = t.lh.min(2);
        let n = 1usize << t.lw;
        for y in 0..t.h as usize {
            for x in 0..n {
                txa[0][0][oy + y][ox + x] = lw;
                txa[1][0][oy + y][ox + x] = lh;
            }
            txa[0][1][oy + y][ox] = t.w;
        }
        for x in 0..n {
            txa[1][1][oy][ox + x] = t.h;
        }
    }
}

#[inline]
fn set_bit(m: &mut [u16; 2], mask: u32) {
    let sidx = (mask >= 0x10000) as usize;
    m[sidx] |= (mask >> (sidx << 4)) as u16;
}

#[allow(clippy::too_many_arguments)]
fn mask_edges_inter(
    masks: &mut [[[[u16; 2]; 3]; 32]; 2],
    by4: usize,
    bx4: usize,
    w4: usize,
    h4: usize,
    skip: bool,
    max_tx: usize,
    tx_masks: &[u16; 2],
    a: &mut [u8],
    l: &mut [u8],
) {
    let t = &TXFM_DIMENSIONS[max_tx];
    let mut txa: Txa = [[[[0; 32]; 32]; 2]; 2];
    let mut y_off = 0;
    let mut y = 0;
    while y < h4 {
        let mut x_off = 0;
        let mut x = 0;
        while x < w4 {
            decomp_tx(&mut txa, y, x, max_tx, 0, y_off, x_off, tx_masks);
            x += t.w as usize;
            x_off += 1;
        }
        y += t.h as usize;
        y_off += 1;
    }
    let mut mask = 1u32 << by4;
    for y in 0..h4 {
        set_bit(&mut masks[0][bx4][txa[0][0][y][0].min(l[y]) as usize], mask);
        mask <<= 1;
    }
    mask = 1u32 << bx4;
    for x in 0..w4 {
        set_bit(&mut masks[1][by4][txa[1][0][0][x].min(a[x]) as usize], mask);
        mask <<= 1;
    }
    if !skip {
        mask = 1u32 << by4;
        for y in 0..h4 {
            let mut ltx = txa[0][0][y][0];
            let mut step = txa[0][1][y][0] as usize;
            let mut x = step;
            while x < w4 {
                let rtx = txa[0][0][y][x];
                set_bit(&mut masks[0][bx4 + x][rtx.min(ltx) as usize], mask);
                ltx = rtx;
                step = txa[0][1][y][x] as usize;
                x += step;
            }
            mask <<= 1;
        }
        mask = 1u32 << bx4;
        for x in 0..w4 {
            let mut ttx = txa[1][0][0][x];
            let mut step = txa[1][1][0][x] as usize;
            let mut y = step;
            while y < h4 {
                let btx = txa[1][0][y][x];
                set_bit(&mut masks[1][by4 + y][ttx.min(btx) as usize], mask);
                ttx = btx;
                step = txa[1][1][y][x] as usize;
                y += step;
            }
            mask <<= 1;
        }
    }
    for y in 0..h4 {
        l[y] = txa[0][0][y][w4 - 1];
    }
    a[..w4].copy_from_slice(&txa[1][0][h4 - 1][..w4]);
}

#[allow(clippy::too_many_arguments)]
fn mask_edges_intra(
    masks: &mut [[[[u16; 2]; 3]; 32]; 2],
    by4: usize,
    bx4: usize,
    w4: usize,
    h4: usize,
    tx: usize,
    a: &mut [u8],
    l: &mut [u8],
) {
    let t = &TXFM_DIMENSIONS[tx];
    let twl4c = t.lw.min(2);
    let thl4c = t.lh.min(2);
    let mut mask = 1u32 << by4;
    for y in 0..h4 {
        set_bit(&mut masks[0][bx4][twl4c.min(l[y]) as usize], mask);
        mask <<= 1;
    }
    mask = 1u32 << bx4;
    for x in 0..w4 {
        set_bit(&mut masks[1][by4][thl4c.min(a[x]) as usize], mask);
        mask <<= 1;
    }
    let hstep = t.w as usize;
    let tt = 1u64 << by4;
    let inner = ((tt << h4) - tt) as u32;
    let (i1, i2) = ((inner & 0xffff) as u16, (inner >> 16) as u16);
    let mut x = hstep;
    while x < w4 {
        masks[0][bx4 + x][twl4c as usize][0] |= i1;
        masks[0][bx4 + x][twl4c as usize][1] |= i2;
        x += hstep;
    }
    let vstep = t.h as usize;
    let tt = 1u64 << bx4;
    let inner = ((tt << w4) - tt) as u32;
    let (i1, i2) = ((inner & 0xffff) as u16, (inner >> 16) as u16);
    let mut y = vstep;
    while y < h4 {
        masks[1][by4 + y][thl4c as usize][0] |= i1;
        masks[1][by4 + y][thl4c as usize][1] |= i2;
        y += vstep;
    }
    a[..w4].fill(thl4c);
    l[..h4].fill(twl4c);
}

#[allow(clippy::too_many_arguments)]
fn mask_edges_chroma(
    masks: &mut [[[[u16; 2]; 2]; 32]; 2],
    cby4: usize,
    cbx4: usize,
    cw4: usize,
    ch4: usize,
    skip_inter: bool,
    tx: usize,
    a: &mut [u8],
    l: &mut [u8],
    ss_hor: usize,
    ss_ver: usize,
) {
    let t = &TXFM_DIMENSIONS[tx];
    let twl4c = (t.lw != 0) as u8;
    let thl4c = (t.lh != 0) as u8;
    let vbits = 4 - ss_ver;
    let hbits = 4 - ss_hor;
    let vmask = 16 >> ss_ver;
    let hmask = 16 >> ss_hor;
    let vmax = 1u32 << vmask;
    let hmax = 1u32 << hmask;
    let mut mask = 1u32 << cby4;
    for y in 0..ch4 {
        let sidx = (mask >= vmax) as usize;
        masks[0][cbx4][twl4c.min(l[y]) as usize][sidx] |= (mask >> (sidx << vbits)) as u16;
        mask <<= 1;
    }
    mask = 1u32 << cbx4;
    for x in 0..cw4 {
        let sidx = (mask >= hmax) as usize;
        masks[1][cby4][thl4c.min(a[x]) as usize][sidx] |= (mask >> (sidx << hbits)) as u16;
        mask <<= 1;
    }
    if !skip_inter {
        let hstep = t.w as usize;
        let tt = 1u64 << cby4;
        let inner = ((tt << ch4) - tt) as u32;
        let (i1, i2) = ((inner & ((1 << vmask) - 1)) as u16, (inner >> vmask) as u16);
        let mut x = hstep;
        while x < cw4 {
            if i1 != 0 {
                masks[0][cbx4 + x][twl4c as usize][0] |= i1;
            }
            if i2 != 0 {
                masks[0][cbx4 + x][twl4c as usize][1] |= i2;
            }
            x += hstep;
        }
        let vstep = t.h as usize;
        let tt = 1u64 << cbx4;
        let inner = ((tt << cw4) - tt) as u32;
        let (i1, i2) = ((inner & ((1 << hmask) - 1)) as u16, (inner >> hmask) as u16);
        let mut y = vstep;
        while y < ch4 {
            if i1 != 0 {
                masks[1][cby4 + y][thl4c as usize][0] |= i1;
            }
            if i2 != 0 {
                masks[1][cby4 + y][thl4c as usize][1] |= i2;
            }
            y += vstep;
        }
    }
    a[..cw4].fill(thl4c);
    l[..ch4].fill(twl4c);
}

/// Edge contexts handed to the mask builders: (above, left) luma and chroma.
pub struct LfEdges<'a> {
    pub ay: &'a mut [u8],
    pub ly: &'a mut [u8],
    pub uv: Option<(&'a mut [u8], &'a mut [u8])>,
}

/// `dav1d_create_lf_mask_intra` / `_inter` (`inter` = Some((skip, tx_masks))).
#[allow(clippy::too_many_arguments)]
pub fn create_lf_mask(
    lflvl: &mut Av1Filter,
    level_cache: &mut [[u8; 4]],
    b4_stride: usize,
    filter_level: [u8; 4],
    bx: usize,
    by: usize,
    iw: usize,
    ih: usize,
    bs: usize,
    ytx: usize,
    inter: Option<(bool, [u16; 2])>,
    uvtx: usize,
    layout: Layout,
    e: LfEdges,
) {
    let b_dim = BLOCK_DIMENSIONS[bs];
    let bw4 = (iw - bx).min(b_dim[0] as usize);
    let bh4 = (ih - by).min(b_dim[1] as usize);
    let bx4 = bx & 31;
    let by4 = by & 31;
    if bw4 != 0 && bh4 != 0 {
        for y in 0..bh4 {
            let o = (by + y) * b4_stride + bx;
            for v in &mut level_cache[o..o + bw4] {
                v[0] = filter_level[0];
                v[1] = filter_level[1];
            }
        }
        match inter {
            None => mask_edges_intra(&mut lflvl.filter_y, by4, bx4, bw4, bh4, ytx, e.ay, e.ly),
            Some((skip, tx_masks)) => mask_edges_inter(
                &mut lflvl.filter_y,
                by4,
                bx4,
                bw4,
                bh4,
                skip,
                ytx,
                &tx_masks,
                e.ay,
                e.ly,
            ),
        }
    }
    let Some((auv, luv)) = e.uv else { return };
    let ss_ver = (layout == Layout::I420) as usize;
    let ss_hor = (layout != Layout::I444) as usize;
    let cbw4 =
        (((iw + ss_hor) >> ss_hor) - (bx >> ss_hor)).min((b_dim[0] as usize + ss_hor) >> ss_hor);
    let cbh4 =
        (((ih + ss_ver) >> ss_ver) - (by >> ss_ver)).min((b_dim[1] as usize + ss_ver) >> ss_ver);
    if cbw4 == 0 || cbh4 == 0 {
        return;
    }
    let cbx4 = bx4 >> ss_hor;
    let cby4 = by4 >> ss_ver;
    for y in 0..cbh4 {
        let o = ((by >> ss_ver) + y) * b4_stride + (bx >> ss_hor);
        for v in &mut level_cache[o..o + cbw4] {
            v[2] = filter_level[2];
            v[3] = filter_level[3];
        }
    }
    let skip_inter = inter.map(|(s, _)| s).unwrap_or(false);
    mask_edges_chroma(
        &mut lflvl.filter_uv,
        cby4,
        cbx4,
        cbw4,
        cbh4,
        skip_inter,
        uvtx,
        auv,
        luv,
        ss_hor,
        ss_ver,
    );
}

pub fn calc_eih(lut: &mut Av1FilterLut, sharp: i32) {
    for level in 0..64 {
        let mut limit = level;
        if sharp > 0 {
            limit >>= (sharp + 3) >> 2;
            limit = limit.min(9 - sharp);
        }
        limit = limit.max(1);
        lut.i[level as usize] = limit as u8;
        lut.e[level as usize] = (2 * (level + 2) + limit) as u8;
    }
    lut.sharp[0] = ((sharp + 3) >> 2) as u64;
    lut.sharp[1] = if sharp != 0 { (9 - sharp) as u64 } else { 0xff };
}

/// `[seg][dir][ref][is_gmv]`.
pub type LfLevels = [[[[u8; 2]; 8]; 4]; 8];

fn calc_lf_value(
    v: &mut [[u8; 2]; 8],
    base_lvl: i32,
    lf_delta: i32,
    seg_delta: i32,
    mr: Option<&LoopfilterModeRefDeltas>,
) {
    let base = ((base_lvl + lf_delta).clamp(0, 63) + seg_delta).clamp(0, 63);
    match mr {
        None => *v = [[base as u8; 2]; 8],
        Some(mr) => {
            let sh = (base >= 32) as i32;
            let l0 = (base + mr.ref_delta[0] as i32 * (1 << sh)).clamp(0, 63) as u8;
            v[0] = [l0, l0];
            for r in 1..8 {
                for m in 0..2 {
                    let delta = mr.mode_delta[m] as i32 + mr.ref_delta[r] as i32;
                    v[r][m] = (base + delta * (1 << sh)).clamp(0, 63) as u8;
                }
            }
        }
    }
}

pub fn calc_lf_values(values: &mut LfLevels, hdr: &FrameHeader, lf_delta: [i8; 4]) {
    let n_seg = if hdr.segmentation.enabled { 8 } else { 1 };
    let lf = &hdr.loopfilter;
    if lf.level_y[0] == 0 && lf.level_y[1] == 0 {
        for v in values.iter_mut().take(n_seg) {
            *v = [[[0; 2]; 8]; 4];
        }
        return;
    }
    let mr = if lf.mode_ref_delta_enabled {
        Some(&lf.mode_ref_deltas)
    } else {
        None
    };
    let multi = hdr.delta.lf_multi;
    for s in 0..n_seg {
        let segd = if hdr.segmentation.enabled {
            Some(&hdr.segmentation.seg_data.d[s])
        } else {
            None
        };
        let sd =
            |f: fn(&super::obu::SegmentationData) -> i8| segd.map(|d| f(d) as i32).unwrap_or(0);
        calc_lf_value(
            &mut values[s][0],
            lf.level_y[0] as i32,
            lf_delta[0] as i32,
            sd(|d| d.delta_lf_y_v),
            mr,
        );
        calc_lf_value(
            &mut values[s][1],
            lf.level_y[1] as i32,
            lf_delta[if multi { 1 } else { 0 }] as i32,
            sd(|d| d.delta_lf_y_h),
            mr,
        );
        for (k, (lvl, dl)) in [(lf.level_u, 2usize), (lf.level_v, 3)]
            .into_iter()
            .enumerate()
        {
            let seg = if k == 0 {
                sd(|d| d.delta_lf_u)
            } else {
                sd(|d| d.delta_lf_v)
            };
            if lvl == 0 {
                values[s][2 + k] = [[0; 2]; 8];
            } else {
                calc_lf_value(
                    &mut values[s][2 + k],
                    lvl as i32,
                    lf_delta[if multi { dl } else { 0 }] as i32,
                    seg,
                    mr,
                );
            }
        }
    }
}