api.rs 28 KB
Newer Older
rzumer's avatar
rzumer committed
1 2 3 4 5 6 7 8 9
// Copyright (c) 2018, The rav1e contributors. All rights reserved
//
// This source code is subject to the terms of the BSD 2 Clause License and
// the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
// was not distributed with this source code in the LICENSE file, you can
// obtain it at www.aomedia.org/license/software. If the Alliance for Open
// Media Patent License 1.0 was not distributed with this source code in the
// PATENTS file, you can obtain it at www.aomedia.org/license/patent.

10
use arg_enum_proc_macro::ArgEnum;
Raphaël Zumer's avatar
Raphaël Zumer committed
11
use bitstream_io::*;
Raphaël Zumer's avatar
Raphaël Zumer committed
12 13 14
use crate::encoder::*;
use crate::metrics::calculate_frame_psnr;
use crate::partition::*;
15
use crate::rate::RCState;
16 17 18
use crate::rate::FRAME_NSUBTYPES;
use crate::rate::FRAME_SUBTYPE_I;
use crate::rate::FRAME_SUBTYPE_P;
Raphaël Zumer's avatar
Raphaël Zumer committed
19
use crate::scenechange::SceneChangeDetector;
20
use crate::util::Pixel;
Raphaël Zumer's avatar
Raphaël Zumer committed
21
use self::EncoderStatus::*;
Luca Barbato's avatar
Luca Barbato committed
22

Raphaël Zumer's avatar
Raphaël Zumer committed
23
use std::{cmp, fmt, io};
fbossen's avatar
fbossen committed
24
use std::collections::BTreeMap;
Luca Barbato's avatar
Luca Barbato committed
25
use std::sync::Arc;
Josh Holmer's avatar
Josh Holmer committed
26
use std::collections::BTreeSet;
27
use std::path::PathBuf;
Josh Holmer's avatar
Josh Holmer committed
28 29

const LOOKAHEAD_FRAMES: u64 = 10;
Luca Barbato's avatar
Luca Barbato committed
30 31 32

// TODO: use the num crate?
#[derive(Clone, Copy, Debug)]
33
#[repr(C)]
Luca Barbato's avatar
Luca Barbato committed
34 35 36
pub struct Rational {
  pub num: u64,
  pub den: u64
Luca Barbato's avatar
Luca Barbato committed
37 38
}

Luca Barbato's avatar
Luca Barbato committed
39 40 41
impl Rational {
  pub fn new(num: u64, den: u64) -> Self {
    Rational { num, den }
Luca Barbato's avatar
Luca Barbato committed
42 43 44
  }
}

45 46 47 48 49 50 51
#[derive(Clone, Copy, Debug)]
#[repr(C)]
pub struct Point {
  pub x: u16,
  pub y: u16
}

52
#[derive(Clone, Debug)]
53
pub struct EncoderConfig {
54
  // output size
55 56
  pub width: usize,
  pub height: usize,
57 58

  // data format and ancillary color information
59 60 61
  pub bit_depth: usize,
  pub chroma_sampling: ChromaSampling,
  pub chroma_sample_position: ChromaSamplePosition,
62 63 64 65 66 67
  pub pixel_range: PixelRange,
  pub color_description: Option<ColorDescription>,
  pub mastering_display: Option<MasteringDisplay>,
  pub content_light: Option<ContentLight>,

  // encoder configuration
68
  pub time_base: Rational,
Josh Holmer's avatar
Josh Holmer committed
69 70 71 72
  /// The *minimum* interval between two keyframes
  pub min_key_frame_interval: u64,
  /// The *maximum* interval between two keyframes
  pub max_key_frame_interval: u64,
73
  pub low_latency: bool,
74
  pub quantizer: usize,
75
  pub bitrate: i32,
76 77
  pub tune: Tune,
  pub speed_settings: SpeedSettings,
78 79
  /// `None` for one-pass encode. `Some(1)` or `Some(2)` for two-pass encoding.
  pub pass: Option<u8>,
80
  pub show_psnr: bool,
81
  pub stats_file: Option<PathBuf>,
82
  pub train_rdo: bool,
83 84 85 86
}

impl Default for EncoderConfig {
  fn default() -> Self {
87
    const DEFAULT_SPEED: usize = 5;
88
    Self::with_speed_preset(DEFAULT_SPEED)
89 90 91
  }
}

92 93 94
impl EncoderConfig {
  pub fn with_speed_preset(speed: usize) -> Self {
    EncoderConfig {
95 96
      width: 640,
      height: 480,
97

98 99 100
      bit_depth: 8,
      chroma_sampling: ChromaSampling::Cs420,
      chroma_sample_position: ChromaSamplePosition::Unknown,
101 102 103 104 105
      pixel_range: PixelRange::Unspecified,
      color_description: None,
      mastering_display: None,
      content_light: None,

106
      time_base: Rational { num: 30, den: 1 },
Josh Holmer's avatar
Josh Holmer committed
107 108
      min_key_frame_interval: 12,
      max_key_frame_interval: 240,
109
      low_latency: false,
110
      quantizer: 100,
111
      bitrate: 0,
112
      tune: Tune::default(),
113
      speed_settings: SpeedSettings::from_preset(speed),
114
      pass: None,
115
      show_psnr: false,
116
      stats_file: None,
117
      train_rdo: false
118 119 120 121 122 123 124 125 126 127 128
    }
  }
}

#[derive(Clone, Copy, Debug)]
pub struct SpeedSettings {
  pub min_block_size: BlockSize,
  pub multiref: bool,
  pub fast_deblock: bool,
  pub reduced_tx_set: bool,
  pub tx_domain_distortion: bool,
129
  pub tx_domain_rate: bool,
130 131 132 133
  pub encode_bottomup: bool,
  pub rdo_tx_decision: bool,
  pub prediction_modes: PredictionModesSetting,
  pub include_near_mvs: bool,
134
  pub no_scene_detection: bool,
135 136
  pub diamond_me: bool,
  pub cdef: bool
137 138 139 140 141
}

impl Default for SpeedSettings {
  fn default() -> Self {
    SpeedSettings {
142
      min_block_size: BlockSize::BLOCK_16X16,
143 144 145 146
      multiref: false,
      fast_deblock: false,
      reduced_tx_set: false,
      tx_domain_distortion: false,
147
      tx_domain_rate: false,
148 149 150 151 152
      encode_bottomup: false,
      rdo_tx_decision: false,
      prediction_modes: PredictionModesSetting::Simple,
      include_near_mvs: false,
      no_scene_detection: false,
153
      diamond_me: false,
154
      cdef: false,
155 156
    }
  }
157 158 159 160 161 162 163 164 165 166
}

impl SpeedSettings {
  pub fn from_preset(speed: usize) -> Self {
    SpeedSettings {
      min_block_size: Self::min_block_size_preset(speed),
      multiref: Self::multiref_preset(speed),
      fast_deblock: Self::fast_deblock_preset(speed),
      reduced_tx_set: Self::reduced_tx_set_preset(speed),
      tx_domain_distortion: Self::tx_domain_distortion_preset(speed),
167
      tx_domain_rate: Self::tx_domain_rate_preset(speed),
168 169 170 171
      encode_bottomup: Self::encode_bottomup_preset(speed),
      rdo_tx_decision: Self::rdo_tx_decision_preset(speed),
      prediction_modes: Self::prediction_modes_preset(speed),
      include_near_mvs: Self::include_near_mvs_preset(speed),
172
      no_scene_detection: Self::no_scene_detection_preset(speed),
173
      diamond_me: Self::diamond_me_preset(speed),
174
      cdef: Self::cdef_preset(speed),
175 176 177
    }
  }

178 179
  /// This preset is set this way because 8x8 with reduced TX set is faster but with equivalent
  /// or better quality compared to 16x16 or 32x32 (to which reduced TX set does not apply).
180
  fn min_block_size_preset(speed: usize) -> BlockSize {
181
    if speed == 0 {
182
      BlockSize::BLOCK_4X4
183
    } else if speed <= 8 {
184 185 186 187 188 189
      BlockSize::BLOCK_8X8
    } else {
      BlockSize::BLOCK_64X64
    }
  }

190 191 192 193
  /// Multiref is enabled automatically if low_latency is false,
  /// but if someone is setting low_latency to true manually,
  /// multiref has a large speed penalty with low quality gain.
  /// Because low_latency can be set manually, this setting is conservative.
194
  fn multiref_preset(speed: usize) -> bool {
195
    speed <= 1
196 197 198
  }

  fn fast_deblock_preset(speed: usize) -> bool {
199
    speed >= 8
200 201 202
  }

  fn reduced_tx_set_preset(speed: usize) -> bool {
203
    speed >= 5
204 205
  }

206 207 208
  /// TX domain distortion is always faster, with no significant quality change
  fn tx_domain_distortion_preset(_speed: usize) -> bool {
    true
209 210
  }

211 212 213 214
  fn tx_domain_rate_preset(_speed: usize) -> bool {
    false
  }

215 216 217 218 219 220 221 222 223
  fn encode_bottomup_preset(speed: usize) -> bool {
    speed == 0
  }

  fn rdo_tx_decision_preset(speed: usize) -> bool {
    speed <= 3
  }

  fn prediction_modes_preset(speed: usize) -> PredictionModesSetting {
224
    if speed <= 1 {
225
      PredictionModesSetting::ComplexAll
226
    } else if speed <= 5 {
227 228 229 230 231 232 233 234 235
      PredictionModesSetting::ComplexKeyframes
    } else {
      PredictionModesSetting::Simple
    }
  }

  fn include_near_mvs_preset(speed: usize) -> bool {
    speed <= 2
  }
236 237 238 239

  fn no_scene_detection_preset(speed: usize) -> bool {
    speed == 10
  }
240

241 242 243 244 245 246 247
  /// Currently Diamond ME gives better quality than full search on most videos,
  /// in addition to being faster.
  /// There are a few outliers, such as the Wikipedia test clip.
  ///
  /// TODO: Revisit this setting if full search quality improves in the future.
  fn diamond_me_preset(_speed: usize) -> bool {
    true
248
  }
249 250 251 252

  fn cdef_preset(_speed: usize) -> bool {
    true
  }
253 254
}

255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276
#[allow(dead_code, non_camel_case_types)]
#[derive(Debug, PartialEq, Clone, Copy, Serialize, Deserialize)]
#[repr(C)]
pub enum FrameType {
  KEY,
  INTER,
  INTRA_ONLY,
  SWITCH
}

impl fmt::Display for FrameType {
  fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
    use self::FrameType::*;
    match self {
      KEY => write!(f, "Key frame"),
      INTER => write!(f, "Inter frame"),
      INTRA_ONLY => write!(f, "Intra only frame"),
      SWITCH => write!(f, "Switching frame"),
    }
  }
}

277 278 279 280 281 282 283
#[derive(Clone, Copy, Debug, PartialOrd, PartialEq)]
pub enum PredictionModesSetting {
  Simple,
  ComplexKeyframes,
  ComplexAll,
}

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
#[derive(Copy, Clone, Debug, PartialEq)]
#[repr(C)]
pub enum ChromaSampling {
  Cs420,
  Cs422,
  Cs444,
  Cs400,
}

impl Default for ChromaSampling {
  fn default() -> Self {
    ChromaSampling::Cs420
  }
}

impl ChromaSampling {
  // Provides the sampling period in the horizontal and vertical axes.
  pub fn sampling_period(self) -> (usize, usize) {
    use self::ChromaSampling::*;
    match self {
      Cs420 => (2, 2),
      Cs422 => (2, 1),
      Cs444 => (1, 1),
      Cs400 => (2, 2),
    }
  }
}

#[derive(Copy, Clone, Debug, PartialEq)]
#[repr(C)]
pub enum ChromaSamplePosition {
  Unknown,
  Vertical,
  Colocated
}

impl Default for ChromaSamplePosition {
  fn default() -> Self {
    ChromaSamplePosition::Unknown
  }
}

326 327 328 329 330 331
#[derive(ArgEnum, Debug, Clone, Copy, PartialEq)]
#[repr(C)]
pub enum PixelRange {
    Unspecified = 0,
    Limited,
    Full,
332 333 334 335 336 337 338 339
}

impl Default for PixelRange {
    fn default() -> Self {
        PixelRange::Unspecified
    }
}

340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356
#[derive(ArgEnum, Debug, Clone, Copy, PartialEq)]
#[repr(C)]
pub enum MatrixCoefficients {
    Identity = 0,
    BT709,
    Unspecified,
    BT470M = 4,
    BT470BG,
    ST170M,
    ST240M,
    YCgCo,
    BT2020NonConstantLuminance,
    BT2020ConstantLuminance,
    ST2085,
    ChromaticityDerivedNonConstantLuminance,
    ChromaticityDerivedConstantLuminance,
    ICtCp,
357 358 359 360 361 362 363 364
}

impl Default for MatrixCoefficients {
    fn default() -> Self {
        MatrixCoefficients::Unspecified
    }
}

365 366 367 368 369 370 371 372 373 374 375 376 377 378 379
#[derive(ArgEnum, Debug, Clone, Copy, PartialEq)]
#[repr(C)]
pub enum ColorPrimaries {
    BT709 = 1,
    Unspecified,
    BT470M = 4,
    BT470BG,
    ST170M,
    ST240M,
    Film,
    BT2020,
    ST428,
    P3DCI,
    P3Display,
    Tech3213 = 22,
380 381 382 383 384 385 386 387
}

impl Default for ColorPrimaries {
    fn default() -> Self {
        ColorPrimaries::Unspecified
    }
}

388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407
#[derive(ArgEnum, Debug, Clone, Copy, PartialEq)]
#[repr(C)]
pub enum TransferCharacteristics {
    BT1886 = 1,
    Unspecified,
    BT470M = 4,
    BT470BG,
    ST170M,
    ST240M,
    Linear,
    Logarithmic100,
    Logarithmic316,
    XVYCC,
    BT1361E,
    SRGB,
    BT2020Ten,
    BT2020Twelve,
    PerceptualQuantizer,
    ST428,
    HybridLogGamma,
408 409 410 411 412 413 414 415
}

impl Default for TransferCharacteristics {
    fn default() -> Self {
        TransferCharacteristics::Unspecified
    }
}

416 417 418 419 420 421 422
#[derive(Copy, Clone, Debug)]
pub struct ColorDescription {
    pub color_primaries: ColorPrimaries,
    pub transfer_characteristics: TransferCharacteristics,
    pub matrix_coefficients: MatrixCoefficients
}

423 424 425 426 427 428 429 430 431 432 433 434 435 436
#[derive(Copy, Clone, Debug)]
pub struct MasteringDisplay {
    pub primaries: [Point; 3],
    pub white_point: Point,
    pub max_luminance: u32,
    pub min_luminance: u32,
}

#[derive(Copy, Clone, Debug)]
pub struct ContentLight {
    pub max_content_light_level: u16,
    pub max_frame_average_light_level: u16,
}

437
/// Contain all the encoder configuration
438
#[derive(Clone, Debug)]
439
pub struct Config {
Luca Barbato's avatar
Luca Barbato committed
440 441 442
  pub enc: EncoderConfig,
  /// The number of threads in the threadpool.
  pub threads: usize
Luca Barbato's avatar
Luca Barbato committed
443 444
}

445 446
const MAX_USABLE_THREADS: usize = 4;

Luca Barbato's avatar
Luca Barbato committed
447
impl Config {
Luca Barbato's avatar
Luca Barbato committed
448 449
  pub fn parse(&mut self, key: &str, value: &str) -> Result<(), EncoderStatus> {
    match key {
450
      "low_latency" => self.enc.low_latency = value.parse().map_err(|_e| ParseError)?,
Josh Holmer's avatar
Josh Holmer committed
451 452
      "min_key_frame_interval" => self.enc.min_key_frame_interval = value.parse().map_err(|_e| ParseError)?,
      "key_frame_interval" => self.enc.max_key_frame_interval = value.parse().map_err(|_e| ParseError)?,
453
      "quantizer" => self.enc.quantizer = value.parse().map_err(|_e| ParseError)?,
454
      "speed" => self.enc.speed_settings = SpeedSettings::from_preset(value.parse().map_err(|_e| ParseError)?),
455
      "tune" => self.enc.tune = value.parse().map_err(|_e| ParseError)?,
Luca Barbato's avatar
Luca Barbato committed
456
      "threads" => self.threads = value.parse().map_err(|_e| ParseError)?,
457 458
      "width" => self.enc.width = value.parse().map_err(|_e| ParseError)?,
      "height" => self.enc.height = value.parse().map_err(|_e| ParseError)?,
459
      _ => return Err(InvalidKey)
Luca Barbato's avatar
Luca Barbato committed
460 461 462 463 464
    }

    Ok(())
  }

465
  pub fn new_context<T: Pixel>(&self) -> Context<T> {
466 467 468
    // initialize with temporal delimiter
    let packet_data = TEMPORAL_DELIMITER.to_vec();

469 470 471 472 473 474
    let maybe_ac_qi_max = if self.enc.quantizer < 255 {
      Some(self.enc.quantizer as u8)
    } else {
      None
    };

475 476 477 478 479 480 481
    let threads = if self.threads == 0 {
      rayon::current_num_threads().min(MAX_USABLE_THREADS)
    } else {
      self.threads
    };

    let pool = rayon::ThreadPoolBuilder::new().num_threads(threads).build().unwrap();
Luca Barbato's avatar
Luca Barbato committed
482

Josh Holmer's avatar
Josh Holmer committed
483 484
    Context {
      frame_count: 0,
485
      limit: 0,
Josh Holmer's avatar
Josh Holmer committed
486
      idx: 0,
Josh Holmer's avatar
Josh Holmer committed
487
      frames_processed: 0,
Josh Holmer's avatar
Josh Holmer committed
488
      frame_q: BTreeMap::new(),
Josh Holmer's avatar
Josh Holmer committed
489 490
      frame_data: BTreeMap::new(),
      keyframes: BTreeSet::new(),
491
      packet_data,
Luca Barbato's avatar
Luca Barbato committed
492
      pool,
Josh Holmer's avatar
Josh Holmer committed
493 494
      segment_start_idx: 0,
      segment_start_frame: 0,
495
      keyframe_detector: SceneChangeDetector::new(self.enc.bit_depth),
496
      config: self.clone(),
497 498 499 500 501 502 503 504 505 506
      rc_state: RCState::new(
        self.enc.width as i32,
        self.enc.height as i32,
        self.enc.time_base.num as i64,
        self.enc.time_base.den as i64,
        self.enc.bitrate,
        maybe_ac_qi_max,
        self.enc.max_key_frame_interval as i32
      ),
      maybe_prev_log_base_q: None,
507 508 509
      first_pass_data: FirstPassData {
        frames: Vec::new(),
      },
Josh Holmer's avatar
Josh Holmer committed
510
    }
Luca Barbato's avatar
Luca Barbato committed
511 512 513
  }
}

514
pub struct Context<T: Pixel> {
Luca Barbato's avatar
Luca Barbato committed
515
  //    timebase: Rational,
fbossen's avatar
fbossen committed
516
  frame_count: u64,
517
  limit: u64,
518
  pub(crate) idx: u64,
Josh Holmer's avatar
Josh Holmer committed
519 520
  frames_processed: u64,
  /// Maps frame *number* to frames
521
  frame_q: BTreeMap<u64, Option<Arc<Frame<T>>>>, //    packet_q: VecDeque<Packet>
Josh Holmer's avatar
Josh Holmer committed
522
  /// Maps frame *idx* to frame data
523
  frame_data: BTreeMap<u64, FrameInvariants<T>>,
Josh Holmer's avatar
Josh Holmer committed
524 525 526
  /// A list of keyframe *numbers* in this encode. Needed so that we don't
  /// need to keep all of the frame_data in memory for the whole life of the encode.
  keyframes: BTreeSet<u64>,
527
  /// A storage space for reordered frames.
Josh Holmer's avatar
Josh Holmer committed
528 529 530
  packet_data: Vec<u8>,
  segment_start_idx: u64,
  segment_start_frame: u64,
531
  keyframe_detector: SceneChangeDetector<T>,
532
  pub config: Config,
533
  rc_state: RCState,
534
  maybe_prev_log_base_q: Option<i64>,
535
  pub first_pass_data: FirstPassData,
Luca Barbato's avatar
Luca Barbato committed
536
  pool: rayon::ThreadPool,
Luca Barbato's avatar
Luca Barbato committed
537 538 539 540
}

#[derive(Clone, Copy, Debug)]
pub enum EncoderStatus {
541
  /// The encoder needs more data to produce an output Packet--used with frame reordering
Luca Barbato's avatar
Luca Barbato committed
542
  NeedMoreData,
543 544
  /// The encoder needs more Frames to analyze lookahead
  NeedMoreFrames,
Luca Barbato's avatar
Luca Barbato committed
545 546 547
  /// There are enough Frames queue
  EnoughData,
  ///
Luca Barbato's avatar
Luca Barbato committed
548 549 550
  Failure,
  InvalidKey,
  ParseError
Luca Barbato's avatar
Luca Barbato committed
551 552
}

553
pub struct Packet<T: Pixel> {
Luca Barbato's avatar
Luca Barbato committed
554
  pub data: Vec<u8>,
555
  pub rec: Option<Frame<T>>,
fbossen's avatar
fbossen committed
556
  pub number: u64,
557 558 559
  pub frame_type: FrameType,
  /// PSNR for Y, U, and V planes
  pub psnr: Option<(f64, f64, f64)>,
560 561
}

562
impl<T: Pixel> fmt::Display for Packet<T> {
563
  fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
Luca Barbato's avatar
Luca Barbato committed
564 565 566 567 568 569 570
    write!(
      f,
      "Frame {} - {} - {} bytes",
      self.number,
      self.frame_type,
      self.data.len()
    )
571
  }
Luca Barbato's avatar
Luca Barbato committed
572 573
}

574 575
impl<T: Pixel> Context<T> {
  pub fn new_frame(&self) -> Arc<Frame<T>> {
Josh Holmer's avatar
Josh Holmer committed
576
    Arc::new(Frame::new(
577 578
      self.config.enc.width,
      self.config.enc.height,
579
      self.config.enc.chroma_sampling
Josh Holmer's avatar
Josh Holmer committed
580
    ))
Luca Barbato's avatar
Luca Barbato committed
581 582 583 584
  }

  pub fn send_frame<F>(&mut self, frame: F) -> Result<(), EncoderStatus>
  where
585 586
    F: Into<Option<Arc<Frame<T>>>>,
    T: Pixel,
Luca Barbato's avatar
Luca Barbato committed
587
  {
Josh Holmer's avatar
Josh Holmer committed
588 589
    let idx = self.frame_count;
    self.frame_q.insert(idx, frame.into());
590
    self.frame_count += 1;
Luca Barbato's avatar
Luca Barbato committed
591 592 593
    Ok(())
  }

594
  pub fn get_frame(&self, frame_number: u64) -> Arc<Frame<T>> {
595 596 597 598
    // Clones only the arc, so low cost overhead
    self.frame_q.get(&frame_number).as_ref().unwrap().as_ref().unwrap().clone()
  }

599 600 601 602
  pub fn get_frame_count(&self) -> u64 {
    self.frame_count
  }

603 604
  pub fn set_limit(&mut self, limit: u64) {
    self.limit = limit;
605 606
  }

Josh Holmer's avatar
Josh Holmer committed
607 608 609 610
  pub fn needs_more_lookahead(&self) -> bool {
    self.needs_more_frames(self.frame_count) && self.frames_processed + LOOKAHEAD_FRAMES > self.frame_q.keys().last().cloned().unwrap_or(0)
  }

611
  pub fn needs_more_frames(&self, frame_count: u64) -> bool {
612
    self.limit == 0 || frame_count < self.limit
613 614
  }

615 616 617
  pub fn container_sequence_header(&mut self) -> Vec<u8> {
    fn sequence_header_inner(seq: &Sequence) -> io::Result<Vec<u8>> {
      let mut buf = Vec::new();
618

619
      {
620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636
        let mut bw = BitWriter::endian(&mut buf, BigEndian);
        bw.write_bit(true)?; // marker
        bw.write(7, 1)?; // version
        bw.write(3, seq.profile)?;
        bw.write(5, 32)?; // level
        bw.write_bit(false)?; // tier
        bw.write_bit(seq.bit_depth > 8)?; // high_bitdepth
        bw.write_bit(seq.bit_depth == 12)?; // twelve_bit
        bw.write_bit(seq.bit_depth == 1)?; // monochrome
        bw.write_bit(seq.bit_depth == 12)?; // twelve_bit
        bw.write_bit(seq.chroma_sampling != ChromaSampling::Cs444)?; // chroma_subsampling_x
        bw.write_bit(seq.chroma_sampling == ChromaSampling::Cs420)?; // chroma_subsampling_y
        bw.write(2, 0)?; // sample_position
        bw.write(3, 0)?; // reserved
        bw.write_bit(false)?; // initial_presentation_delay_present

        bw.write(4, 0)?; // reserved
637
      }
638

639 640 641
      Ok(buf)
    }

Josh Holmer's avatar
Josh Holmer committed
642
    sequence_header_inner(&self.frame_data[&0].sequence).unwrap()
643 644
  }

Josh Holmer's avatar
Josh Holmer committed
645
  fn next_keyframe(&self) -> u64 {
Josh Holmer's avatar
Josh Holmer committed
646 647 648 649
    let next_detected = self.frame_data.values()
      .find(|fi| fi.frame_type == FrameType::KEY && fi.number > self.segment_start_frame)
      .map(|fi| fi.number);
    let next_limit = self.segment_start_frame + self.config.enc.max_key_frame_interval;
Josh Holmer's avatar
Josh Holmer committed
650 651 652 653 654
    if next_detected.is_none() {
      return next_limit;
    }
    cmp::min(next_detected.unwrap(), next_limit)
  }
fbossen's avatar
fbossen committed
655

656
  fn set_frame_properties(&mut self, idx: u64) -> bool {
657
    let (fi, end_of_subgop) = self.build_frame_properties(idx);
658 659 660
    self.frame_data.insert(idx, fi);

    end_of_subgop
Josh Holmer's avatar
Josh Holmer committed
661 662
  }

663
  fn build_frame_properties(&mut self, idx: u64) -> (FrameInvariants<T>, bool) {
Josh Holmer's avatar
Josh Holmer committed
664
    if idx == 0 {
665
      let seq = Sequence::new(&self.config.enc);
666

Josh Holmer's avatar
Josh Holmer committed
667
      // The first frame will always be a key frame
Josh Holmer's avatar
Josh Holmer committed
668 669
      let fi = FrameInvariants::new_key_frame(
        &FrameInvariants::new(
670
          self.config.enc.clone(),
671
          seq
Josh Holmer's avatar
Josh Holmer committed
672 673 674
        ),
        0
      );
675
      return (fi, true);
Josh Holmer's avatar
Josh Holmer committed
676 677
    }

Josh Holmer's avatar
Josh Holmer committed
678 679
    let mut fi = self.frame_data[&(idx - 1)].clone();

680 681 682 683 684
    // FIXME: inter unsupported with 4:2:2 and 4:4:4 chroma sampling
    let chroma_sampling = self.config.enc.chroma_sampling;
    let keyframe_only = chroma_sampling == ChromaSampling::Cs444 ||
      chroma_sampling == ChromaSampling::Cs422;

Josh Holmer's avatar
Josh Holmer committed
685 686 687 688 689 690
    // Initially set up the frame as an inter frame.
    // We need to determine what the frame number is before we can
    // look up the frame type. If reordering is enabled, the idx
    // may not match the frame number.
    let idx_in_segment = idx - self.segment_start_idx;
    if idx_in_segment > 0 {
691
      let next_keyframe = if keyframe_only { self.segment_start_frame + 1 } else { self.next_keyframe() };
692
      let (fi_temp, end_of_subgop) = FrameInvariants::new_inter_frame(
Josh Holmer's avatar
Josh Holmer committed
693
        &fi,
694 695 696 697
        self.segment_start_frame,
        idx_in_segment,
        next_keyframe
      );
Josh Holmer's avatar
Josh Holmer committed
698
      fi = fi_temp;
699
      if !end_of_subgop {
Josh Holmer's avatar
Josh Holmer committed
700 701 702
        if !fi.inter_cfg.unwrap().reorder
          || ((idx_in_segment - 1) % fi.inter_cfg.unwrap().group_len == 0
          && fi.number == (next_keyframe - 1))
703
        {
Josh Holmer's avatar
Josh Holmer committed
704
          self.segment_start_idx = idx;
Josh Holmer's avatar
Josh Holmer committed
705
          self.segment_start_frame = next_keyframe;
Josh Holmer's avatar
Josh Holmer committed
706
          fi.number = next_keyframe;
Josh Holmer's avatar
Josh Holmer committed
707
        } else {
708
          return (fi, false);
Josh Holmer's avatar
Josh Holmer committed
709 710
        }
      }
fbossen's avatar
fbossen committed
711
    }
Luca Barbato's avatar
Luca Barbato committed
712

713 714
    match self.frame_q.get(&fi.number) {
      Some(Some(_)) => {},
Josh Holmer's avatar
Josh Holmer committed
715
      _ => { return (fi, false); }
716 717
    }

Josh Holmer's avatar
Josh Holmer committed
718
    // Now that we know the frame number, look up the correct frame type
Josh Holmer's avatar
Josh Holmer committed
719 720 721 722 723 724 725
    let frame_type = self.determine_frame_type(fi.number);
    if frame_type == FrameType::KEY {
      self.segment_start_idx = idx;
      self.segment_start_frame = fi.number;
      self.keyframes.insert(fi.number);
    }
    fi.frame_type = frame_type;
Josh Holmer's avatar
Josh Holmer committed
726

Josh Holmer's avatar
Josh Holmer committed
727 728 729 730 731
    let idx_in_segment = idx - self.segment_start_idx;
    if idx_in_segment == 0 {
      fi = FrameInvariants::new_key_frame(&fi, self.segment_start_frame);
    } else {
      let next_keyframe = self.next_keyframe();
732
      let (fi_temp, end_of_subgop) = FrameInvariants::new_inter_frame(
Josh Holmer's avatar
Josh Holmer committed
733 734 735 736 737 738
        &fi,
        self.segment_start_frame,
        idx_in_segment,
        next_keyframe
      );
      fi = fi_temp;
739 740
      if !end_of_subgop {
        return (fi, false);
Josh Holmer's avatar
Josh Holmer committed
741 742
      }
    }
743
    (fi, true)
fbossen's avatar
fbossen committed
744
  }
Luca Barbato's avatar
Luca Barbato committed
745

746
  pub fn receive_packet(&mut self) -> Result<Packet<T>, EncoderStatus> {
747
    if self.needs_more_lookahead() {
748
      return Err(EncoderStatus::NeedMoreFrames);
749 750
    }

751 752 753 754 755 756
    let idx = {
      let mut idx = self.idx;
      while !self.set_frame_properties(idx) {
        self.idx += 1;
        idx = self.idx;
      }
Luca Barbato's avatar
Luca Barbato committed
757

758
      if !self.needs_more_frames(self.frame_data[&idx].number) {
759 760 761 762 763
        self.idx += 1;
        return Err(EncoderStatus::EnoughData);
      }
      idx
    };
764

765
    let ret = {
Luca Barbato's avatar
Luca Barbato committed
766 767 768
      let fi = self.frame_data.get_mut(&idx).unwrap();
      if fi.show_existing_frame {
        self.idx += 1;
Luca Barbato's avatar
Luca Barbato committed
769

Luca Barbato's avatar
Luca Barbato committed
770
        let mut fs = FrameState::new(fi);
fbossen's avatar
fbossen committed
771

772 773
        // TODO: Record the bits spent here against the original frame for rate
        //  control purposes, or add a new frame subtype?
Luca Barbato's avatar
Luca Barbato committed
774
        let sef_data = self.pool.install(|| encode_frame(fi, &mut fs));
775
        self.packet_data.extend(sef_data);
Luca Barbato's avatar
Luca Barbato committed
776

Luca Barbato's avatar
Luca Barbato committed
777
        let rec = if fi.show_frame { Some(fs.rec) } else { None };
778
        let fi = fi.clone();
779
        self.finalize_packet(rec, &fi)
Luca Barbato's avatar
Luca Barbato committed
780 781 782
      } else {
        if let Some(f) = self.frame_q.get(&fi.number) {
          self.idx += 1;
fbossen's avatar
fbossen committed
783

784
          if let Some(frame) = f.clone() {
785
            let fti = fi.get_frame_subtype();
786 787 788
            let qps =
              self.rc_state.select_qi(self, fti, self.maybe_prev_log_base_q);
            let fi = self.frame_data.get_mut(&idx).unwrap();
789
            fi.set_quantizers(&qps);
Luca Barbato's avatar
Luca Barbato committed
790
            let mut fs = FrameState::new_with_frame(fi, frame.clone());
fbossen's avatar
fbossen committed
791

792
            // TODO: Trial encoding for first frame of each type.
Luca Barbato's avatar
Luca Barbato committed
793
            let data = self.pool.install(||encode_frame(fi, &mut fs));
794 795 796 797 798 799 800 801
            self.maybe_prev_log_base_q = Some(qps.log_base_q);
            // TODO: Add support for dropping frames.
            self.rc_state.update_state(
              (data.len() * 8) as i64,
              fti,
              qps.log_target_q,
              false
            );
Luca Barbato's avatar
Luca Barbato committed
802
            self.packet_data.extend(data);
fbossen's avatar
fbossen committed
803

Luca Barbato's avatar
Luca Barbato committed
804
            fs.rec.pad(fi.width, fi.height);
fbossen's avatar
fbossen committed
805

Luca Barbato's avatar
Luca Barbato committed
806 807
            // TODO avoid the clone by having rec Arc.
            let rec = if fi.show_frame { Some(fs.rec.clone()) } else { None };
fbossen's avatar
fbossen committed
808

Luca Barbato's avatar
Luca Barbato committed
809
            update_rec_buffer(fi, fs);
fbossen's avatar
fbossen committed
810

Luca Barbato's avatar
Luca Barbato committed
811
            if fi.show_frame {
812
              let fi = fi.clone();
813
              self.finalize_packet(rec, &fi)
Luca Barbato's avatar
Luca Barbato committed
814 815 816
            } else {
              Err(EncoderStatus::NeedMoreData)
            }
817 818 819
          } else {
            Err(EncoderStatus::NeedMoreData)
          }
fbossen's avatar
fbossen committed
820 821 822 823
        } else {
          Err(EncoderStatus::NeedMoreData)
        }
      }
824 825 826
    };

    if let Ok(ref pkt) = ret {
Luca Barbato's avatar
Luca Barbato committed
827
      self.garbage_collect(pkt.number);
828 829 830
    }

    ret
Luca Barbato's avatar
Luca Barbato committed
831 832
  }

833
  fn finalize_packet(&mut self, rec: Option<Frame<T>>, fi: &FrameInvariants<T>) -> Result<Packet<T>, EncoderStatus> {
834 835 836 837 838 839 840 841 842
    let data = self.packet_data.clone();
    self.packet_data.clear();
    if write_temporal_delimiter(&mut self.packet_data).is_err() {
      return Err(EncoderStatus::Failure);
    }

    let mut psnr = None;
    if self.config.enc.show_psnr {
      if let Some(ref rec) = rec {
843
        let original_frame = self.get_frame(fi.number);
844 845 846 847 848 849 850 851
        psnr = Some(calculate_frame_psnr(
          &*original_frame,
          rec,
          fi.sequence.bit_depth
        ));
      }
    }

852 853 854 855
    if self.config.enc.pass == Some(1) {
      self.first_pass_data.frames.push(FirstPassFrame::from(fi));
    }

856 857 858 859 860 861 862 863 864 865
    self.frames_processed += 1;
    Ok(Packet {
      data,
      rec,
      number: fi.number,
      frame_type: fi.frame_type,
      psnr
    })
  }

866
  fn garbage_collect(&mut self, cur_frame: u64) {
Josh Holmer's avatar
Josh Holmer committed
867 868 869 870 871 872 873 874 875 876 877 878 879 880
    if cur_frame == 0 {
      return;
    }
    for i in 0..cur_frame {
      self.frame_q.remove(&i);
    }
    if self.idx < 2 {
      return;
    }
    for i in 0..(self.idx - 1) {
      self.frame_data.remove(&i);
    }
  }

Luca Barbato's avatar
Luca Barbato committed
881
  pub fn flush(&mut self) {
fbossen's avatar
fbossen committed
882
    self.frame_q.insert(self.frame_count, None);
883
    self.frame_count += 1;
Luca Barbato's avatar
Luca Barbato committed
884
  }
Josh Holmer's avatar
Josh Holmer committed
885

Josh Holmer's avatar
Josh Holmer committed
886 887
  fn determine_frame_type(&mut self, frame_number: u64) -> FrameType {
    if frame_number == 0 {
Josh Holmer's avatar
Josh Holmer committed
888 889
      return FrameType::KEY;
    }
890 891 892 893 894 895 896
    if self.config.enc.speed_settings.no_scene_detection {
      if frame_number % self.config.enc.max_key_frame_interval == 0 {
        return FrameType::KEY;
      } else {
        return FrameType::INTER;
      }
    }
Josh Holmer's avatar
Josh Holmer committed
897

Josh Holmer's avatar
Josh Holmer committed
898 899 900 901 902 903 904 905
    let prev_keyframe = self.keyframes.iter()
      .rfind(|&&keyframe| keyframe < frame_number)
      .cloned()
      .unwrap_or(0);
    let frame = match self.frame_q.get(&frame_number).cloned() {
      Some(frame) => frame,
      None => { return FrameType::KEY; }
    };
Josh Holmer's avatar
Josh Holmer committed
906
    if let Some(frame) = frame {
Josh Holmer's avatar
Josh Holmer committed
907 908 909
      let distance = frame_number - prev_keyframe;
      if distance < self.config.enc.min_key_frame_interval {
        if distance + 1 == self.config.enc.min_key_frame_interval {
Josh Holmer's avatar
Josh Holmer committed
910 911
          // Run the detector for the current frame, so that it will contain this frame's information
          // to compare against the next frame. We can ignore the results for this frame.
Josh Holmer's avatar
Josh Holmer committed
912
          self.keyframe_detector.detect_scene_change(frame, frame_number as usize);
Josh Holmer's avatar
Josh Holmer committed
913 914 915
        }
        return FrameType::INTER;
      }
Josh Holmer's avatar
Josh Holmer committed
916
      if distance >= self.config.enc.max_key_frame_interval {
Josh Holmer's avatar
Josh Holmer committed
917 918
        return FrameType::KEY;
      }
Josh Holmer's avatar
Josh Holmer committed
919
      if self.keyframe_detector.detect_scene_change(frame, frame_number as usize) {
Josh Holmer's avatar
Josh Holmer committed
920 921 922 923 924
        return FrameType::KEY;
      }
    }
    FrameType::INTER
  }
925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997

  // Count the number of frames of each subtype in the next
  //  reservoir_frame_delay frames.
  // Returns the number of frames until the last keyframe in the next
  //  reservoir_frame_delay frames, or the end of the interval, whichever
  //  comes first.
  pub(crate) fn guess_frame_subtypes(
    &self, nframes: &mut [i32; FRAME_NSUBTYPES], reservoir_frame_delay: i32
  ) -> i32 {
    // TODO: Ideally this logic should be centralized, but the actual code used
    //  to determine a frame's subtype is spread over many places and
    //  intertwined with mutable state changes that occur when the frame is
    //  actually encoded.
    // So for now we just duplicate it here in stateless fashion.
    for fti in 0..FRAME_NSUBTYPES {
      nframes[fti] = 0;
    }
    let mut prev_keyframe = self.segment_start_idx;
    let mut acc: [i32; FRAME_NSUBTYPES] = [0; FRAME_NSUBTYPES];
    // Updates the frame counts with the accumulated values when we hit a
    //  keyframe.
    fn collect_counts(
      nframes: &mut [i32; FRAME_NSUBTYPES], acc: &mut [i32; FRAME_NSUBTYPES]
    ) {
      for fti in 0..FRAME_NSUBTYPES {
        nframes[fti] += acc[fti];
        acc[fti] = 0;
      }
      acc[FRAME_SUBTYPE_I] += 1;
    }
    for idx in self.idx..(self.idx + reservoir_frame_delay as u64) {
      if let Some(fd) = self.frame_data.get(&idx) {
        if fd.frame_type == FrameType::KEY {
          collect_counts(nframes, &mut acc);
          prev_keyframe = idx;
          continue;
        }
      } else if idx == 0
        || idx - prev_keyframe >= self.config.enc.max_key_frame_interval
      {
        collect_counts(nframes, &mut acc);
        prev_keyframe = idx;
        continue;
      }
      // TODO: Implement golden P-frames.
      let mut fti = FRAME_SUBTYPE_P;
      if !self.config.enc.low_latency {
        let pyramid_depth = 2;
        let group_src_len = 1 << pyramid_depth;
        let group_len = group_src_len + pyramid_depth;
        let idx_in_group = (idx - prev_keyframe - 1) % group_len;
        let lvl = if idx_in_group < pyramid_depth {
          idx_in_group
        } else {
          pos_to_lvl(idx_in_group - pyramid_depth + 1, pyramid_depth)
        };
        fti += lvl as usize;
      }
      acc[fti] += 1;
    }
    if prev_keyframe <= self.idx {
      // If there were no keyframes at all, or only the first frame was a
      //  keyframe, the accumulators never flushed and still contain counts for
      //  the entire buffer.
      // In both cases, we return these counts.
      collect_counts(nframes, &mut acc);
      reservoir_frame_delay
    } else {
      // Otherwise, we discard what remains in the accumulators as they contain
      //  the counts from and past the last keyframe.
      (prev_keyframe - self.idx) as i32
    }
  }
Luca Barbato's avatar
Luca Barbato committed
998
}
999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010

#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct FirstPassData {
  frames: Vec<FirstPassFrame>,
}

#[derive(Debug, Clone, Copy, Serialize, Deserialize)]
pub struct FirstPassFrame {
  number: u64,
  frame_type: FrameType,
}

1011 1012
impl<T: Pixel> From<&FrameInvariants<T>> for FirstPassFrame {
  fn from(fi: &FrameInvariants<T>) -> FirstPassFrame {
1013 1014 1015 1016 1017
    FirstPassFrame {
      number: fi.number,
      frame_type: fi.frame_type,
    }
  }
1018
}