api.rs 27.8 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.

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

Raphaël Zumer's avatar
Raphaël Zumer committed
22
use std::{cmp, fmt, io};
fbossen's avatar
fbossen committed
23
use std::collections::BTreeMap;
Luca Barbato's avatar
Luca Barbato committed
24
use std::sync::Arc;
Josh Holmer's avatar
Josh Holmer committed
25
use std::collections::BTreeSet;
26
use std::path::PathBuf;
Josh Holmer's avatar
Josh Holmer committed
27 28

const LOOKAHEAD_FRAMES: u64 = 10;
Luca Barbato's avatar
Luca Barbato committed
29 30 31

// TODO: use the num crate?
#[derive(Clone, Copy, Debug)]
32
#[repr(C)]
Luca Barbato's avatar
Luca Barbato committed
33 34 35
pub struct Rational {
  pub num: u64,
  pub den: u64
Luca Barbato's avatar
Luca Barbato committed
36 37
}

Luca Barbato's avatar
Luca Barbato committed
38 39 40
impl Rational {
  pub fn new(num: u64, den: u64) -> Self {
    Rational { num, den }
Luca Barbato's avatar
Luca Barbato committed
41 42 43
  }
}

44 45 46 47 48 49 50
#[derive(Clone, Copy, Debug)]
#[repr(C)]
pub struct Point {
  pub x: u16,
  pub y: u16
}

51
#[derive(Clone, Debug)]
52
pub struct EncoderConfig {
53
  // output size
54 55
  pub width: usize,
  pub height: usize,
56 57

  // data format and ancillary color information
58 59 60
  pub bit_depth: usize,
  pub chroma_sampling: ChromaSampling,
  pub chroma_sample_position: ChromaSamplePosition,
61 62 63 64 65 66
  pub pixel_range: PixelRange,
  pub color_description: Option<ColorDescription>,
  pub mastering_display: Option<MasteringDisplay>,
  pub content_light: Option<ContentLight>,

  // encoder configuration
67
  pub time_base: Rational,
Josh Holmer's avatar
Josh Holmer committed
68 69 70 71
  /// The *minimum* interval between two keyframes
  pub min_key_frame_interval: u64,
  /// The *maximum* interval between two keyframes
  pub max_key_frame_interval: u64,
72
  pub low_latency: bool,
73
  pub quantizer: usize,
74
  pub bitrate: i32,
75 76
  pub tune: Tune,
  pub speed_settings: SpeedSettings,
77 78
  /// `None` for one-pass encode. `Some(1)` or `Some(2)` for two-pass encoding.
  pub pass: Option<u8>,
79
  pub show_psnr: bool,
80
  pub stats_file: Option<PathBuf>,
81
  pub train_rdo: bool,
82 83 84 85
}

impl Default for EncoderConfig {
  fn default() -> Self {
86
    const DEFAULT_SPEED: usize = 5;
87
    Self::with_speed_preset(DEFAULT_SPEED)
88 89 90
  }
}

91 92 93
impl EncoderConfig {
  pub fn with_speed_preset(speed: usize) -> Self {
    EncoderConfig {
94 95
      width: 640,
      height: 480,
96

97 98 99
      bit_depth: 8,
      chroma_sampling: ChromaSampling::Cs420,
      chroma_sample_position: ChromaSamplePosition::Unknown,
100 101 102 103 104
      pixel_range: PixelRange::Unspecified,
      color_description: None,
      mastering_display: None,
      content_light: None,

105
      time_base: Rational { num: 30, den: 1 },
Josh Holmer's avatar
Josh Holmer committed
106 107
      min_key_frame_interval: 12,
      max_key_frame_interval: 240,
108
      low_latency: false,
109
      quantizer: 100,
110
      bitrate: 0,
111
      tune: Tune::default(),
112
      speed_settings: SpeedSettings::from_preset(speed),
113
      pass: None,
114
      show_psnr: false,
115
      stats_file: None,
116
      train_rdo: false
117 118 119 120 121 122 123 124 125 126 127
    }
  }
}

#[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,
128
  pub tx_domain_rate: bool,
129 130 131 132
  pub encode_bottomup: bool,
  pub rdo_tx_decision: bool,
  pub prediction_modes: PredictionModesSetting,
  pub include_near_mvs: bool,
133
  pub no_scene_detection: bool,
134
  pub diamond_me: bool
135 136 137 138 139
}

impl Default for SpeedSettings {
  fn default() -> Self {
    SpeedSettings {
140
      min_block_size: BlockSize::BLOCK_16X16,
141 142 143 144
      multiref: false,
      fast_deblock: false,
      reduced_tx_set: false,
      tx_domain_distortion: false,
145
      tx_domain_rate: false,
146 147 148 149 150
      encode_bottomup: false,
      rdo_tx_decision: false,
      prediction_modes: PredictionModesSetting::Simple,
      include_near_mvs: false,
      no_scene_detection: false,
151
      diamond_me: false,
152 153
    }
  }
154 155 156 157 158 159 160 161 162 163
}

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),
164
      tx_domain_rate: Self::tx_domain_rate_preset(speed),
165 166 167 168
      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),
169
      no_scene_detection: Self::no_scene_detection_preset(speed),
170
      diamond_me: Self::diamond_me_preset(speed),
171 172 173
    }
  }

174 175
  /// 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).
176
  fn min_block_size_preset(speed: usize) -> BlockSize {
177
    if speed == 0 {
178
      BlockSize::BLOCK_4X4
179
    } else if speed <= 8 {
180 181 182 183 184 185
      BlockSize::BLOCK_8X8
    } else {
      BlockSize::BLOCK_64X64
    }
  }

186 187 188 189
  /// 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.
190
  fn multiref_preset(speed: usize) -> bool {
191
    speed <= 1
192 193 194
  }

  fn fast_deblock_preset(speed: usize) -> bool {
195
    speed >= 8
196 197 198
  }

  fn reduced_tx_set_preset(speed: usize) -> bool {
199
    speed >= 5
200 201
  }

202 203 204
  /// TX domain distortion is always faster, with no significant quality change
  fn tx_domain_distortion_preset(_speed: usize) -> bool {
    true
205 206
  }

207 208 209 210
  fn tx_domain_rate_preset(_speed: usize) -> bool {
    false
  }

211 212 213 214 215 216 217 218 219
  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 {
220
    if speed <= 1 {
221
      PredictionModesSetting::ComplexAll
222
    } else if speed <= 5 {
223 224 225 226 227 228 229 230 231
      PredictionModesSetting::ComplexKeyframes
    } else {
      PredictionModesSetting::Simple
    }
  }

  fn include_near_mvs_preset(speed: usize) -> bool {
    speed <= 2
  }
232 233 234 235

  fn no_scene_detection_preset(speed: usize) -> bool {
    speed == 10
  }
236

237 238 239 240 241 242 243
  /// 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
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
#[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"),
    }
  }
}

269 270 271 272 273 274 275
#[derive(Clone, Copy, Debug, PartialOrd, PartialEq)]
pub enum PredictionModesSetting {
  Simple,
  ComplexKeyframes,
  ComplexAll,
}

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
#[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
  }
}

318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333
arg_enum!{
  #[derive(Debug, Clone, Copy, PartialEq)]
  #[repr(C)]
  pub enum PixelRange {
      Unspecified = 0,
      Limited,
      Full,
  }
}

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

334 335
arg_enum!{
  #[derive(Debug, Clone, Copy, PartialEq)]
336
  #[repr(C)]
337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352
  pub enum MatrixCoefficients {
      Identity = 0,
      BT709,
      Unspecified,
      BT470M = 4,
      BT470BG,
      ST170M,
      ST240M,
      YCgCo,
      BT2020NonConstantLuminance,
      BT2020ConstantLuminance,
      ST2085,
      ChromaticityDerivedNonConstantLuminance,
      ChromaticityDerivedConstantLuminance,
      ICtCp,
  }
353 354 355 356 357 358 359 360
}

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

361
arg_enum!{
362
  #[derive(Debug, Clone, Copy, PartialEq)]
363
  #[repr(C)]
364 365 366 367 368 369 370 371 372 373 374 375 376 377
  pub enum ColorPrimaries {
      BT709 = 1,
      Unspecified,
      BT470M = 4,
      BT470BG,
      ST170M,
      ST240M,
      Film,
      BT2020,
      ST428,
      P3DCI,
      P3Display,
      Tech3213 = 22,
  }
378 379 380 381 382 383 384 385
}

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

386
arg_enum!{
387
  #[derive(Debug, Clone, Copy, PartialEq)]
388
  #[repr(C)]
389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407
  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
      _ => return Err(InvalidKey)
Luca Barbato's avatar
Luca Barbato committed
458 459 460 461 462
    }

    Ok(())
  }

463
  pub fn new_context<T: Pixel>(&self) -> Context<T> {
464 465 466
    // initialize with temporal delimiter
    let packet_data = TEMPORAL_DELIMITER.to_vec();

467 468 469 470 471 472
    let maybe_ac_qi_max = if self.enc.quantizer < 255 {
      Some(self.enc.quantizer as u8)
    } else {
      None
    };

473 474 475 476 477 478 479
    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
480

Josh Holmer's avatar
Josh Holmer committed
481 482
    Context {
      frame_count: 0,
483
      limit: 0,
Josh Holmer's avatar
Josh Holmer committed
484
      idx: 0,
Josh Holmer's avatar
Josh Holmer committed
485
      frames_processed: 0,
Josh Holmer's avatar
Josh Holmer committed
486
      frame_q: BTreeMap::new(),
Josh Holmer's avatar
Josh Holmer committed
487 488
      frame_data: BTreeMap::new(),
      keyframes: BTreeSet::new(),
489
      packet_data,
Luca Barbato's avatar
Luca Barbato committed
490
      pool,
Josh Holmer's avatar
Josh Holmer committed
491 492
      segment_start_idx: 0,
      segment_start_frame: 0,
493
      keyframe_detector: SceneChangeDetector::new(self.enc.bit_depth),
494
      config: self.clone(),
495 496 497 498 499 500 501 502 503 504
      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,
505 506 507
      first_pass_data: FirstPassData {
        frames: Vec::new(),
      },
Josh Holmer's avatar
Josh Holmer committed
508
    }
Luca Barbato's avatar
Luca Barbato committed
509 510 511
  }
}

512
pub struct Context<T: Pixel> {
Luca Barbato's avatar
Luca Barbato committed
513
  //    timebase: Rational,
fbossen's avatar
fbossen committed
514
  frame_count: u64,
515
  limit: u64,
516
  pub(crate) idx: u64,
Josh Holmer's avatar
Josh Holmer committed
517 518
  frames_processed: u64,
  /// Maps frame *number* to frames
519
  frame_q: BTreeMap<u64, Option<Arc<Frame<T>>>>, //    packet_q: VecDeque<Packet>
Josh Holmer's avatar
Josh Holmer committed
520
  /// Maps frame *idx* to frame data
521
  frame_data: BTreeMap<u64, FrameInvariants<T>>,
Josh Holmer's avatar
Josh Holmer committed
522 523 524
  /// 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>,
525
  /// A storage space for reordered frames.
Josh Holmer's avatar
Josh Holmer committed
526 527 528
  packet_data: Vec<u8>,
  segment_start_idx: u64,
  segment_start_frame: u64,
529
  keyframe_detector: SceneChangeDetector<T>,
530
  pub config: Config,
531
  rc_state: RCState,
532
  maybe_prev_log_base_q: Option<i64>,
533
  pub first_pass_data: FirstPassData,
Luca Barbato's avatar
Luca Barbato committed
534
  pool: rayon::ThreadPool,
Luca Barbato's avatar
Luca Barbato committed
535 536 537 538
}

#[derive(Clone, Copy, Debug)]
pub enum EncoderStatus {
539
  /// The encoder needs more data to produce an output Packet--used with frame reordering
Luca Barbato's avatar
Luca Barbato committed
540
  NeedMoreData,
541 542
  /// The encoder needs more Frames to analyze lookahead
  NeedMoreFrames,
Luca Barbato's avatar
Luca Barbato committed
543 544 545
  /// There are enough Frames queue
  EnoughData,
  ///
Luca Barbato's avatar
Luca Barbato committed
546 547 548
  Failure,
  InvalidKey,
  ParseError
Luca Barbato's avatar
Luca Barbato committed
549 550
}

551
pub struct Packet<T: Pixel> {
Luca Barbato's avatar
Luca Barbato committed
552
  pub data: Vec<u8>,
553
  pub rec: Option<Frame<T>>,
fbossen's avatar
fbossen committed
554
  pub number: u64,
555 556 557
  pub frame_type: FrameType,
  /// PSNR for Y, U, and V planes
  pub psnr: Option<(f64, f64, f64)>,
558 559
}

560
impl<T: Pixel> fmt::Display for Packet<T> {
561
  fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
Luca Barbato's avatar
Luca Barbato committed
562 563 564 565 566 567 568
    write!(
      f,
      "Frame {} - {} - {} bytes",
      self.number,
      self.frame_type,
      self.data.len()
    )
569
  }
Luca Barbato's avatar
Luca Barbato committed
570 571
}

572 573
impl<T: Pixel> Context<T> {
  pub fn new_frame(&self) -> Arc<Frame<T>> {
Josh Holmer's avatar
Josh Holmer committed
574
    Arc::new(Frame::new(
575 576
      self.config.enc.width,
      self.config.enc.height,
577
      self.config.enc.chroma_sampling
Josh Holmer's avatar
Josh Holmer committed
578
    ))
Luca Barbato's avatar
Luca Barbato committed
579 580 581 582
  }

  pub fn send_frame<F>(&mut self, frame: F) -> Result<(), EncoderStatus>
  where
583 584
    F: Into<Option<Arc<Frame<T>>>>,
    T: Pixel,
Luca Barbato's avatar
Luca Barbato committed
585
  {
Josh Holmer's avatar
Josh Holmer committed
586 587
    let idx = self.frame_count;
    self.frame_q.insert(idx, frame.into());
588
    self.frame_count += 1;
Luca Barbato's avatar
Luca Barbato committed
589 590 591
    Ok(())
  }

592
  pub fn get_frame(&self, frame_number: u64) -> Arc<Frame<T>> {
593 594 595 596
    // Clones only the arc, so low cost overhead
    self.frame_q.get(&frame_number).as_ref().unwrap().as_ref().unwrap().clone()
  }

597 598 599 600
  pub fn get_frame_count(&self) -> u64 {
    self.frame_count
  }

601 602
  pub fn set_limit(&mut self, limit: u64) {
    self.limit = limit;
603 604
  }

Josh Holmer's avatar
Josh Holmer committed
605 606 607 608
  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)
  }

609
  pub fn needs_more_frames(&self, frame_count: u64) -> bool {
610
    self.limit == 0 || frame_count < self.limit
611 612
  }

613 614 615
  pub fn container_sequence_header(&mut self) -> Vec<u8> {
    fn sequence_header_inner(seq: &Sequence) -> io::Result<Vec<u8>> {
      let mut buf = Vec::new();
616

617
      {
618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634
        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
635
      }
636

637 638 639
      Ok(buf)
    }

Josh Holmer's avatar
Josh Holmer committed
640
    sequence_header_inner(&self.frame_data[&0].sequence).unwrap()
641 642
  }

Josh Holmer's avatar
Josh Holmer committed
643
  fn next_keyframe(&self) -> u64 {
Josh Holmer's avatar
Josh Holmer committed
644 645 646 647
    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
648 649 650 651 652
    if next_detected.is_none() {
      return next_limit;
    }
    cmp::min(next_detected.unwrap(), next_limit)
  }
fbossen's avatar
fbossen committed
653

654
  fn set_frame_properties(&mut self, idx: u64) -> bool {
655
    let (fi, end_of_subgop) = self.build_frame_properties(idx);
656 657 658
    self.frame_data.insert(idx, fi);

    end_of_subgop
Josh Holmer's avatar
Josh Holmer committed
659 660
  }

661
  fn build_frame_properties(&mut self, idx: u64) -> (FrameInvariants<T>, bool) {
Josh Holmer's avatar
Josh Holmer committed
662
    if idx == 0 {
663
      let seq = Sequence::new(&self.config.enc);
664

Josh Holmer's avatar
Josh Holmer committed
665
      // The first frame will always be a key frame
Josh Holmer's avatar
Josh Holmer committed
666 667
      let fi = FrameInvariants::new_key_frame(
        &FrameInvariants::new(
668
          self.config.enc.clone(),
669
          seq
Josh Holmer's avatar
Josh Holmer committed
670 671 672
        ),
        0
      );
673
      return (fi, true);
Josh Holmer's avatar
Josh Holmer committed
674 675
    }

Josh Holmer's avatar
Josh Holmer committed
676 677
    let mut fi = self.frame_data[&(idx - 1)].clone();

678 679 680 681 682
    // 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
683 684 685 686 687 688
    // 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 {
689
      let next_keyframe = if keyframe_only { self.segment_start_frame + 1 } else { self.next_keyframe() };
690
      let (fi_temp, end_of_subgop) = FrameInvariants::new_inter_frame(
Josh Holmer's avatar
Josh Holmer committed
691
        &fi,
692 693 694 695
        self.segment_start_frame,
        idx_in_segment,
        next_keyframe
      );
Josh Holmer's avatar
Josh Holmer committed
696
      fi = fi_temp;
697
      if !end_of_subgop {
Josh Holmer's avatar
Josh Holmer committed
698 699 700
        if !fi.inter_cfg.unwrap().reorder
          || ((idx_in_segment - 1) % fi.inter_cfg.unwrap().group_len == 0
          && fi.number == (next_keyframe - 1))
701
        {
Josh Holmer's avatar
Josh Holmer committed
702
          self.segment_start_idx = idx;
Josh Holmer's avatar
Josh Holmer committed
703
          self.segment_start_frame = next_keyframe;
Josh Holmer's avatar
Josh Holmer committed
704
          fi.number = next_keyframe;
Josh Holmer's avatar
Josh Holmer committed
705
        } else {
706
          return (fi, false);
Josh Holmer's avatar
Josh Holmer committed
707 708
        }
      }
fbossen's avatar
fbossen committed
709
    }
Luca Barbato's avatar
Luca Barbato committed
710

711 712
    match self.frame_q.get(&fi.number) {
      Some(Some(_)) => {},
Josh Holmer's avatar
Josh Holmer committed
713
      _ => { return (fi, false); }
714 715
    }

Josh Holmer's avatar
Josh Holmer committed
716
    // Now that we know the frame number, look up the correct frame type
Josh Holmer's avatar
Josh Holmer committed
717 718 719 720 721 722 723
    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
724

Josh Holmer's avatar
Josh Holmer committed
725 726 727 728 729
    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();
730
      let (fi_temp, end_of_subgop) = FrameInvariants::new_inter_frame(
Josh Holmer's avatar
Josh Holmer committed
731 732 733 734 735 736
        &fi,
        self.segment_start_frame,
        idx_in_segment,
        next_keyframe
      );
      fi = fi_temp;
737 738
      if !end_of_subgop {
        return (fi, false);
Josh Holmer's avatar
Josh Holmer committed
739 740
      }
    }
741
    (fi, true)
fbossen's avatar
fbossen committed
742
  }
Luca Barbato's avatar
Luca Barbato committed
743

744
  pub fn receive_packet(&mut self) -> Result<Packet<T>, EncoderStatus> {
745
    if self.needs_more_lookahead() {
746
      return Err(EncoderStatus::NeedMoreFrames);
747 748
    }

749 750 751 752 753 754
    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
755

756
      if !self.needs_more_frames(self.frame_data[&idx].number) {
757 758 759 760 761
        self.idx += 1;
        return Err(EncoderStatus::EnoughData);
      }
      idx
    };
762

763
    let ret = {
Luca Barbato's avatar
Luca Barbato committed
764 765 766
      let fi = self.frame_data.get_mut(&idx).unwrap();
      if fi.show_existing_frame {
        self.idx += 1;
Luca Barbato's avatar
Luca Barbato committed
767

Luca Barbato's avatar
Luca Barbato committed
768
        let mut fs = FrameState::new(fi);
fbossen's avatar
fbossen committed
769

770 771
        // 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
772
        let sef_data = self.pool.install(|| encode_frame(fi, &mut fs));
773
        self.packet_data.extend(sef_data);
Luca Barbato's avatar
Luca Barbato committed
774

Luca Barbato's avatar
Luca Barbato committed
775
        let rec = if fi.show_frame { Some(fs.rec) } else { None };
776
        let fi = fi.clone();
777
        self.finalize_packet(rec, &fi)
Luca Barbato's avatar
Luca Barbato committed
778 779 780
      } else {
        if let Some(f) = self.frame_q.get(&fi.number) {
          self.idx += 1;
fbossen's avatar
fbossen committed
781

782
          if let Some(frame) = f.clone() {
783
            let fti = fi.get_frame_subtype();
784 785 786
            let qps =
              self.rc_state.select_qi(self, fti, self.maybe_prev_log_base_q);
            let fi = self.frame_data.get_mut(&idx).unwrap();
787
            fi.set_quantizers(&qps);
Luca Barbato's avatar
Luca Barbato committed
788
            let mut fs = FrameState::new_with_frame(fi, frame.clone());
fbossen's avatar
fbossen committed
789

790
            // TODO: Trial encoding for first frame of each type.
Luca Barbato's avatar
Luca Barbato committed
791
            let data = self.pool.install(||encode_frame(fi, &mut fs));
792 793 794 795 796 797 798 799
            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
800
            self.packet_data.extend(data);
fbossen's avatar
fbossen committed
801

Luca Barbato's avatar
Luca Barbato committed
802
            fs.rec.pad(fi.width, fi.height);
fbossen's avatar
fbossen committed
803

Luca Barbato's avatar
Luca Barbato committed
804 805
            // 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
806

Luca Barbato's avatar
Luca Barbato committed
807
            update_rec_buffer(fi, fs);
fbossen's avatar
fbossen committed
808

Luca Barbato's avatar
Luca Barbato committed
809
            if fi.show_frame {
810
              let fi = fi.clone();
811
              self.finalize_packet(rec, &fi)
Luca Barbato's avatar
Luca Barbato committed
812 813 814
            } else {
              Err(EncoderStatus::NeedMoreData)
            }
815 816 817
          } else {
            Err(EncoderStatus::NeedMoreData)
          }
fbossen's avatar
fbossen committed
818 819 820 821
        } else {
          Err(EncoderStatus::NeedMoreData)
        }
      }
822 823 824
    };

    if let Ok(ref pkt) = ret {
Luca Barbato's avatar
Luca Barbato committed
825
      self.garbage_collect(pkt.number);
826 827 828
    }

    ret
Luca Barbato's avatar
Luca Barbato committed
829 830
  }

831
  fn finalize_packet(&mut self, rec: Option<Frame<T>>, fi: &FrameInvariants<T>) -> Result<Packet<T>, EncoderStatus> {
832 833 834 835 836 837 838 839 840
    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 {
841
        let original_frame = self.get_frame(fi.number);
842 843 844 845 846 847 848 849
        psnr = Some(calculate_frame_psnr(
          &*original_frame,
          rec,
          fi.sequence.bit_depth
        ));
      }
    }

850 851 852 853
    if self.config.enc.pass == Some(1) {
      self.first_pass_data.frames.push(FirstPassFrame::from(fi));
    }

854 855 856 857 858 859 860 861 862 863
    self.frames_processed += 1;
    Ok(Packet {
      data,
      rec,
      number: fi.number,
      frame_type: fi.frame_type,
      psnr
    })
  }

864
  fn garbage_collect(&mut self, cur_frame: u64) {
Josh Holmer's avatar
Josh Holmer committed
865 866 867 868 869 870 871 872 873 874 875 876 877 878
    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
879
  pub fn flush(&mut self) {
fbossen's avatar
fbossen committed
880
    self.frame_q.insert(self.frame_count, None);
881
    self.frame_count += 1;
Luca Barbato's avatar
Luca Barbato committed
882
  }
Josh Holmer's avatar
Josh Holmer committed
883

Josh Holmer's avatar
Josh Holmer committed
884 885
  fn determine_frame_type(&mut self, frame_number: u64) -> FrameType {
    if frame_number == 0 {
Josh Holmer's avatar
Josh Holmer committed
886 887
      return FrameType::KEY;
    }
888 889 890 891 892 893 894
    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
895

Josh Holmer's avatar
Josh Holmer committed
896 897 898 899 900 901 902 903
    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
904
    if let Some(frame) = frame {
Josh Holmer's avatar
Josh Holmer committed
905 906 907
      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
908 909
          // 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
910
          self.keyframe_detector.detect_scene_change(frame, frame_number as usize);
Josh Holmer's avatar
Josh Holmer committed
911 912 913
        }
        return FrameType::INTER;
      }
Josh Holmer's avatar
Josh Holmer committed
914
      if distance >= self.config.enc.max_key_frame_interval {
Josh Holmer's avatar
Josh Holmer committed
915 916
        return FrameType::KEY;
      }
Josh Holmer's avatar
Josh Holmer committed
917
      if self.keyframe_detector.detect_scene_change(frame, frame_number as usize) {
Josh Holmer's avatar
Josh Holmer committed
918 919 920 921 922
        return FrameType::KEY;
      }
    }
    FrameType::INTER
  }
923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995

  // 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
996
}
997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008

#[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,
}

1009 1010
impl<T: Pixel> From<&FrameInvariants<T>> for FirstPassFrame {
  fn from(fi: &FrameInvariants<T>) -> FirstPassFrame {
1011 1012 1013 1014 1015
    FirstPassFrame {
      number: fi.number,
      frame_type: fi.frame_type,
    }
  }
1016
}