api.rs 26.8 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.

10
use arg_enum_proc_macro::ArgEnum;
Raphaël Zumer's avatar
Raphaël Zumer committed
11
use bitstream_io::*;
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;
19
use crate::scenechange::SceneChangeDetector;
20
use crate::util::Pixel;
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,
134 135
  pub diamond_me: bool,
  pub cdef: bool
136 137 138 139 140
}

impl Default for SpeedSettings {
  fn default() -> Self {
    SpeedSettings {
141
      min_block_size: BlockSize::BLOCK_16X16,
142 143 144 145
      multiref: false,
      fast_deblock: false,
      reduced_tx_set: false,
      tx_domain_distortion: false,
146
      tx_domain_rate: false,
147 148 149 150 151
      encode_bottomup: false,
      rdo_tx_decision: false,
      prediction_modes: PredictionModesSetting::Simple,
      include_near_mvs: false,
      no_scene_detection: false,
152
      diamond_me: false,
153
      cdef: false,
154 155
    }
  }
156 157 158 159 160 161 162 163 164 165
}

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),
166
      tx_domain_rate: Self::tx_domain_rate_preset(speed),
167 168 169 170
      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),
171
      no_scene_detection: Self::no_scene_detection_preset(speed),
172
      diamond_me: Self::diamond_me_preset(speed),
173
      cdef: Self::cdef_preset(speed),
174 175 176
    }
  }

177 178
  /// 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).
179
  fn min_block_size_preset(speed: usize) -> BlockSize {
180
    if speed == 0 {
181
      BlockSize::BLOCK_4X4
182
    } else if speed <= 8 {
183 184 185 186 187 188
      BlockSize::BLOCK_8X8
    } else {
      BlockSize::BLOCK_64X64
    }
  }

189 190 191 192
  /// 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.
193
  fn multiref_preset(speed: usize) -> bool {
194
    speed <= 1
195 196 197
  }

  fn fast_deblock_preset(speed: usize) -> bool {
198
    speed >= 8
199 200 201
  }

  fn reduced_tx_set_preset(speed: usize) -> bool {
202
    speed >= 5
203 204
  }

205 206 207
  /// TX domain distortion is always faster, with no significant quality change
  fn tx_domain_distortion_preset(_speed: usize) -> bool {
    true
208 209
  }

210 211 212 213
  fn tx_domain_rate_preset(_speed: usize) -> bool {
    false
  }

214 215 216 217 218 219 220 221 222
  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 {
223
    if speed <= 1 {
224
      PredictionModesSetting::ComplexAll
225
    } else if speed <= 5 {
226 227 228 229 230 231 232 233 234
      PredictionModesSetting::ComplexKeyframes
    } else {
      PredictionModesSetting::Simple
    }
  }

  fn include_near_mvs_preset(speed: usize) -> bool {
    speed <= 2
  }
235 236 237 238

  fn no_scene_detection_preset(speed: usize) -> bool {
    speed == 10
  }
239

240 241 242 243 244 245 246
  /// 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
247
  }
248 249 250 251

  fn cdef_preset(_speed: usize) -> bool {
    true
  }
252 253
}

254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275
#[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"),
    }
  }
}

276 277 278 279 280 281 282
#[derive(Clone, Copy, Debug, PartialOrd, PartialEq)]
pub enum PredictionModesSetting {
  Simple,
  ComplexKeyframes,
  ComplexAll,
}

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
#[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
  }
}

325 326 327 328 329 330
#[derive(ArgEnum, Debug, Clone, Copy, PartialEq)]
#[repr(C)]
pub enum PixelRange {
    Unspecified = 0,
    Limited,
    Full,
331 332 333 334 335 336 337 338
}

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

339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
#[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,
356 357 358 359 360 361 362 363
}

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

364 365 366 367 368 369 370 371 372 373 374 375 376 377 378
#[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,
379 380 381 382 383 384 385 386
}

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

387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406
#[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,
407 408 409 410 411 412 413 414
}

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

415 416 417 418 419 420 421
#[derive(Copy, Clone, Debug)]
pub struct ColorDescription {
    pub color_primaries: ColorPrimaries,
    pub transfer_characteristics: TransferCharacteristics,
    pub matrix_coefficients: MatrixCoefficients
}

422 423 424 425 426 427 428 429 430 431 432 433 434 435
#[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,
}

436
/// Contain all the encoder configuration
437
#[derive(Clone, Debug)]
438
pub struct Config {
Luca Barbato's avatar
Luca Barbato committed
439 440 441
  pub enc: EncoderConfig,
  /// The number of threads in the threadpool.
  pub threads: usize
Luca Barbato's avatar
Luca Barbato committed
442 443
}

444 445
const MAX_USABLE_THREADS: usize = 4;

Luca Barbato's avatar
Luca Barbato committed
446
impl Config {
447
  pub fn new_context<T: Pixel>(&self) -> Context<T> {
448 449 450
    // initialize with temporal delimiter
    let packet_data = TEMPORAL_DELIMITER.to_vec();

451 452 453 454 455 456
    let maybe_ac_qi_max = if self.enc.quantizer < 255 {
      Some(self.enc.quantizer as u8)
    } else {
      None
    };

457 458 459 460 461 462 463
    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
464

Josh Holmer's avatar
Josh Holmer committed
465 466
    Context {
      frame_count: 0,
467
      limit: 0,
Josh Holmer's avatar
Josh Holmer committed
468
      idx: 0,
Josh Holmer's avatar
Josh Holmer committed
469
      frames_processed: 0,
Josh Holmer's avatar
Josh Holmer committed
470
      frame_q: BTreeMap::new(),
Josh Holmer's avatar
Josh Holmer committed
471 472
      frame_data: BTreeMap::new(),
      keyframes: BTreeSet::new(),
473
      packet_data,
Luca Barbato's avatar
Luca Barbato committed
474
      pool,
Josh Holmer's avatar
Josh Holmer committed
475 476
      segment_start_idx: 0,
      segment_start_frame: 0,
477
      keyframe_detector: SceneChangeDetector::new(self.enc.bit_depth),
478
      config: self.clone(),
479 480 481 482 483 484 485 486 487 488
      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,
489 490 491
      first_pass_data: FirstPassData {
        frames: Vec::new(),
      },
Josh Holmer's avatar
Josh Holmer committed
492
    }
Luca Barbato's avatar
Luca Barbato committed
493 494 495
  }
}

496
pub struct Context<T: Pixel> {
Luca Barbato's avatar
Luca Barbato committed
497
  //    timebase: Rational,
fbossen's avatar
fbossen committed
498
  frame_count: u64,
499
  limit: u64,
500
  pub(crate) idx: u64,
Josh Holmer's avatar
Josh Holmer committed
501 502
  frames_processed: u64,
  /// Maps frame *number* to frames
503
  frame_q: BTreeMap<u64, Option<Arc<Frame<T>>>>, //    packet_q: VecDeque<Packet>
Josh Holmer's avatar
Josh Holmer committed
504
  /// Maps frame *idx* to frame data
505
  frame_data: BTreeMap<u64, FrameInvariants<T>>,
Josh Holmer's avatar
Josh Holmer committed
506 507 508
  /// 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>,
509
  /// A storage space for reordered frames.
Josh Holmer's avatar
Josh Holmer committed
510 511 512
  packet_data: Vec<u8>,
  segment_start_idx: u64,
  segment_start_frame: u64,
513
  keyframe_detector: SceneChangeDetector<T>,
514
  pub config: Config,
515
  rc_state: RCState,
516
  maybe_prev_log_base_q: Option<i64>,
517
  pub first_pass_data: FirstPassData,
Luca Barbato's avatar
Luca Barbato committed
518
  pool: rayon::ThreadPool,
Luca Barbato's avatar
Luca Barbato committed
519 520 521 522
}

#[derive(Clone, Copy, Debug)]
pub enum EncoderStatus {
523
  /// The encoder needs more data to produce an output Packet--used with frame reordering
Luca Barbato's avatar
Luca Barbato committed
524
  NeedMoreData,
525 526
  /// The encoder needs more Frames to analyze lookahead
  NeedMoreFrames,
Luca Barbato's avatar
Luca Barbato committed
527 528 529
  /// There are enough Frames queue
  EnoughData,
  ///
Luca Barbato's avatar
Luca Barbato committed
530 531 532
  Failure,
  InvalidKey,
  ParseError
Luca Barbato's avatar
Luca Barbato committed
533 534
}

535
pub struct Packet<T: Pixel> {
Luca Barbato's avatar
Luca Barbato committed
536
  pub data: Vec<u8>,
537
  pub rec: Option<Frame<T>>,
fbossen's avatar
fbossen committed
538
  pub number: u64,
539 540 541
  pub frame_type: FrameType,
  /// PSNR for Y, U, and V planes
  pub psnr: Option<(f64, f64, f64)>,
542 543
}

544
impl<T: Pixel> fmt::Display for Packet<T> {
545
  fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
Luca Barbato's avatar
Luca Barbato committed
546 547 548 549 550 551 552
    write!(
      f,
      "Frame {} - {} - {} bytes",
      self.number,
      self.frame_type,
      self.data.len()
    )
553
  }
Luca Barbato's avatar
Luca Barbato committed
554 555
}

556 557
impl<T: Pixel> Context<T> {
  pub fn new_frame(&self) -> Arc<Frame<T>> {
Josh Holmer's avatar
Josh Holmer committed
558
    Arc::new(Frame::new(
559 560
      self.config.enc.width,
      self.config.enc.height,
561
      self.config.enc.chroma_sampling
Josh Holmer's avatar
Josh Holmer committed
562
    ))
Luca Barbato's avatar
Luca Barbato committed
563 564 565 566
  }

  pub fn send_frame<F>(&mut self, frame: F) -> Result<(), EncoderStatus>
  where
567 568
    F: Into<Option<Arc<Frame<T>>>>,
    T: Pixel,
Luca Barbato's avatar
Luca Barbato committed
569
  {
Josh Holmer's avatar
Josh Holmer committed
570 571
    let idx = self.frame_count;
    self.frame_q.insert(idx, frame.into());
572
    self.frame_count += 1;
Luca Barbato's avatar
Luca Barbato committed
573 574 575
    Ok(())
  }

576
  fn get_frame(&self, frame_number: u64) -> Arc<Frame<T>> {
577 578 579 580
    // Clones only the arc, so low cost overhead
    self.frame_q.get(&frame_number).as_ref().unwrap().as_ref().unwrap().clone()
  }

581 582 583 584
  pub fn get_frame_count(&self) -> u64 {
    self.frame_count
  }

585 586
  pub fn set_limit(&mut self, limit: u64) {
    self.limit = limit;
587 588
  }

589
  pub(crate) fn needs_more_lookahead(&self) -> bool {
Josh Holmer's avatar
Josh Holmer committed
590 591 592
    self.needs_more_frames(self.frame_count) && self.frames_processed + LOOKAHEAD_FRAMES > self.frame_q.keys().last().cloned().unwrap_or(0)
  }

593
  pub fn needs_more_frames(&self, frame_count: u64) -> bool {
594
    self.limit == 0 || frame_count < self.limit
595 596
  }

597 598 599
  pub fn container_sequence_header(&mut self) -> Vec<u8> {
    fn sequence_header_inner(seq: &Sequence) -> io::Result<Vec<u8>> {
      let mut buf = Vec::new();
600

601
      {
602 603 604 605
        let mut bw = BitWriter::endian(&mut buf, BigEndian);
        bw.write_bit(true)?; // marker
        bw.write(7, 1)?; // version
        bw.write(3, seq.profile)?;
606
        bw.write(5, 31)?; // level
607 608 609 610 611 612 613 614 615 616 617 618
        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
619
      }
620

621 622 623
      Ok(buf)
    }

624 625 626
    let seq = Sequence::new(&self.config.enc);

    sequence_header_inner(&seq).unwrap()
627 628
  }

Josh Holmer's avatar
Josh Holmer committed
629
  fn next_keyframe(&self) -> u64 {
Josh Holmer's avatar
Josh Holmer committed
630 631 632 633
    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
634 635 636 637 638
    if next_detected.is_none() {
      return next_limit;
    }
    cmp::min(next_detected.unwrap(), next_limit)
  }
fbossen's avatar
fbossen committed
639

640
  fn set_frame_properties(&mut self, idx: u64) -> bool {
641
    let (fi, end_of_subgop) = self.build_frame_properties(idx);
642 643 644
    self.frame_data.insert(idx, fi);

    end_of_subgop
Josh Holmer's avatar
Josh Holmer committed
645 646
  }

647
  fn build_frame_properties(&mut self, idx: u64) -> (FrameInvariants<T>, bool) {
Josh Holmer's avatar
Josh Holmer committed
648
    if idx == 0 {
649
      let seq = Sequence::new(&self.config.enc);
650

Josh Holmer's avatar
Josh Holmer committed
651
      // The first frame will always be a key frame
Josh Holmer's avatar
Josh Holmer committed
652 653
      let fi = FrameInvariants::new_key_frame(
        &FrameInvariants::new(
654
          self.config.enc.clone(),
655
          seq
Josh Holmer's avatar
Josh Holmer committed
656 657 658
        ),
        0
      );
659
      return (fi, true);
Josh Holmer's avatar
Josh Holmer committed
660 661
    }

Josh Holmer's avatar
Josh Holmer committed
662 663
    let mut fi = self.frame_data[&(idx - 1)].clone();

664 665 666 667 668
    // 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
669 670 671 672 673 674
    // 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 {
675
      let next_keyframe = if keyframe_only { self.segment_start_frame + 1 } else { self.next_keyframe() };
676
      let (fi_temp, end_of_subgop) = FrameInvariants::new_inter_frame(
Josh Holmer's avatar
Josh Holmer committed
677
        &fi,
678 679 680 681
        self.segment_start_frame,
        idx_in_segment,
        next_keyframe
      );
Josh Holmer's avatar
Josh Holmer committed
682
      fi = fi_temp;
683
      if !end_of_subgop {
Josh Holmer's avatar
Josh Holmer committed
684 685 686
        if !fi.inter_cfg.unwrap().reorder
          || ((idx_in_segment - 1) % fi.inter_cfg.unwrap().group_len == 0
          && fi.number == (next_keyframe - 1))
687
        {
Josh Holmer's avatar
Josh Holmer committed
688
          self.segment_start_idx = idx;
Josh Holmer's avatar
Josh Holmer committed
689
          self.segment_start_frame = next_keyframe;
Josh Holmer's avatar
Josh Holmer committed
690
          fi.number = next_keyframe;
Josh Holmer's avatar
Josh Holmer committed
691
        } else {
692
          return (fi, false);
Josh Holmer's avatar
Josh Holmer committed
693 694
        }
      }
fbossen's avatar
fbossen committed
695
    }
Luca Barbato's avatar
Luca Barbato committed
696

697 698
    match self.frame_q.get(&fi.number) {
      Some(Some(_)) => {},
Josh Holmer's avatar
Josh Holmer committed
699
      _ => { return (fi, false); }
700 701
    }

Josh Holmer's avatar
Josh Holmer committed
702
    // Now that we know the frame number, look up the correct frame type
Josh Holmer's avatar
Josh Holmer committed
703 704 705 706 707 708 709
    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
710

Josh Holmer's avatar
Josh Holmer committed
711 712 713 714 715
    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();
716
      let (fi_temp, end_of_subgop) = FrameInvariants::new_inter_frame(
Josh Holmer's avatar
Josh Holmer committed
717 718 719 720 721 722
        &fi,
        self.segment_start_frame,
        idx_in_segment,
        next_keyframe
      );
      fi = fi_temp;
723 724
      if !end_of_subgop {
        return (fi, false);
Josh Holmer's avatar
Josh Holmer committed
725 726
      }
    }
727
    (fi, true)
fbossen's avatar
fbossen committed
728
  }
Luca Barbato's avatar
Luca Barbato committed
729

730
  pub fn receive_packet(&mut self) -> Result<Packet<T>, EncoderStatus> {
731
    if self.needs_more_lookahead() {
732
      return Err(EncoderStatus::NeedMoreFrames);
733 734
    }

735 736 737 738 739 740
    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
741

742
      if !self.needs_more_frames(self.frame_data[&idx].number) {
743 744 745 746 747
        self.idx += 1;
        return Err(EncoderStatus::EnoughData);
      }
      idx
    };
748

749
    let ret = {
Luca Barbato's avatar
Luca Barbato committed
750 751 752
      let fi = self.frame_data.get_mut(&idx).unwrap();
      if fi.show_existing_frame {
        self.idx += 1;
Luca Barbato's avatar
Luca Barbato committed
753

Luca Barbato's avatar
Luca Barbato committed
754
        let mut fs = FrameState::new(fi);
fbossen's avatar
fbossen committed
755

756 757
        // TODO: Record the bits spent here against the original frame for rate
        //  control purposes, or add a new frame subtype?
758
        let sef_data = encode_show_existing_frame(fi, &mut fs);
759
        self.packet_data.extend(sef_data);
Luca Barbato's avatar
Luca Barbato committed
760

Luca Barbato's avatar
Luca Barbato committed
761
        let rec = if fi.show_frame { Some(fs.rec) } else { None };
762
        let fi = fi.clone();
763
        self.finalize_packet(rec, &fi)
Luca Barbato's avatar
Luca Barbato committed
764 765 766
      } else {
        if let Some(f) = self.frame_q.get(&fi.number) {
          self.idx += 1;
fbossen's avatar
fbossen committed
767

768
          if let Some(frame) = f.clone() {
769
            let fti = fi.get_frame_subtype();
770 771 772
            let qps =
              self.rc_state.select_qi(self, fti, self.maybe_prev_log_base_q);
            let fi = self.frame_data.get_mut(&idx).unwrap();
773
            fi.set_quantizers(&qps);
Luca Barbato's avatar
Luca Barbato committed
774
            let mut fs = FrameState::new_with_frame(fi, frame.clone());
fbossen's avatar
fbossen committed
775

776
            // TODO: Trial encoding for first frame of each type.
Luca Barbato's avatar
Luca Barbato committed
777
            let data = self.pool.install(||encode_frame(fi, &mut fs));
778 779 780 781 782 783 784 785
            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
786
            self.packet_data.extend(data);
fbossen's avatar
fbossen committed
787

Luca Barbato's avatar
Luca Barbato committed
788
            fs.rec.pad(fi.width, fi.height);
fbossen's avatar
fbossen committed
789

Luca Barbato's avatar
Luca Barbato committed
790 791
            // 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
792

Luca Barbato's avatar
Luca Barbato committed
793
            update_rec_buffer(fi, fs);
fbossen's avatar
fbossen committed
794

Luca Barbato's avatar
Luca Barbato committed
795
            if fi.show_frame {
796
              let fi = fi.clone();
797
              self.finalize_packet(rec, &fi)
Luca Barbato's avatar
Luca Barbato committed
798 799 800
            } else {
              Err(EncoderStatus::NeedMoreData)
            }
801 802 803
          } else {
            Err(EncoderStatus::NeedMoreData)
          }
fbossen's avatar
fbossen committed
804 805 806 807
        } else {
          Err(EncoderStatus::NeedMoreData)
        }
      }
808 809 810
    };

    if let Ok(ref pkt) = ret {
Luca Barbato's avatar
Luca Barbato committed
811
      self.garbage_collect(pkt.number);
812 813 814
    }

    ret
Luca Barbato's avatar
Luca Barbato committed
815 816
  }

817
  fn finalize_packet(&mut self, rec: Option<Frame<T>>, fi: &FrameInvariants<T>) -> Result<Packet<T>, EncoderStatus> {
818 819 820 821 822 823 824 825 826
    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 {
827
        let original_frame = self.get_frame(fi.number);
828 829 830 831 832 833 834 835
        psnr = Some(calculate_frame_psnr(
          &*original_frame,
          rec,
          fi.sequence.bit_depth
        ));
      }
    }

836 837 838 839
    if self.config.enc.pass == Some(1) {
      self.first_pass_data.frames.push(FirstPassFrame::from(fi));
    }

840 841 842 843 844 845 846 847 848 849
    self.frames_processed += 1;
    Ok(Packet {
      data,
      rec,
      number: fi.number,
      frame_type: fi.frame_type,
      psnr
    })
  }

850
  fn garbage_collect(&mut self, cur_frame: u64) {
Josh Holmer's avatar
Josh Holmer committed
851 852 853 854 855 856 857 858 859 860 861 862 863 864
    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
865
  pub fn flush(&mut self) {
fbossen's avatar
fbossen committed
866
    self.frame_q.insert(self.frame_count, None);
867
    self.frame_count += 1;
Luca Barbato's avatar
Luca Barbato committed
868
  }
Josh Holmer's avatar
Josh Holmer committed
869

Josh Holmer's avatar
Josh Holmer committed
870 871
  fn determine_frame_type(&mut self, frame_number: u64) -> FrameType {
    if frame_number == 0 {
Josh Holmer's avatar
Josh Holmer committed
872 873
      return FrameType::KEY;
    }
874 875 876 877 878 879 880
    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
881

Josh Holmer's avatar
Josh Holmer committed
882 883 884 885 886 887 888 889
    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
890
    if let Some(frame) = frame {
Josh Holmer's avatar
Josh Holmer committed
891 892 893
      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 {
894
          self.keyframe_detector.set_last_frame(frame, frame_number as usize);
Josh Holmer's avatar
Josh Holmer committed
895 896 897
        }
        return FrameType::INTER;
      }
Josh Holmer's avatar
Josh Holmer committed
898
      if distance >= self.config.enc.max_key_frame_interval {
Josh Holmer's avatar
Josh Holmer committed
899 900
        return FrameType::KEY;
      }
Josh Holmer's avatar
Josh Holmer committed
901
      if self.keyframe_detector.detect_scene_change(frame, frame_number as usize) {
Josh Holmer's avatar
Josh Holmer committed
902 903 904 905 906
        return FrameType::KEY;
      }
    }
    FrameType::INTER
  }
907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979

  // 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
980
}
981 982 983 984 985 986 987 988 989 990 991 992

#[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,
}

993 994
impl<T: Pixel> From<&FrameInvariants<T>> for FirstPassFrame {
  fn from(fi: &FrameInvariants<T>) -> FirstPassFrame {
995 996 997 998 999
    FirstPassFrame {
      number: fi.number,
      frame_type: fi.frame_type,
    }
  }
1000
}