api.rs 26 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::*;
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;
18
use crate::scenechange::SceneChangeDetector;
Raphaël Zumer's avatar
Raphaël Zumer committed
19
use self::EncoderStatus::*;
Luca Barbato's avatar
Luca Barbato committed
20

Raphaël Zumer's avatar
Raphaël Zumer committed
21
use std::{cmp, fmt, io};
fbossen's avatar
fbossen committed
22
use std::collections::BTreeMap;
Luca Barbato's avatar
Luca Barbato committed
23
use std::sync::Arc;
Josh Holmer's avatar
Josh Holmer committed
24
use std::collections::BTreeSet;
25
use std::path::PathBuf;
Josh Holmer's avatar
Josh Holmer committed
26 27

const LOOKAHEAD_FRAMES: u64 = 10;
Luca Barbato's avatar
Luca Barbato committed
28 29 30

// TODO: use the num crate?
#[derive(Clone, Copy, Debug)]
31
#[repr(C)]
Luca Barbato's avatar
Luca Barbato committed
32 33 34
pub struct Rational {
  pub num: u64,
  pub den: u64
Luca Barbato's avatar
Luca Barbato committed
35 36
}

Luca Barbato's avatar
Luca Barbato committed
37 38 39
impl Rational {
  pub fn new(num: u64, den: u64) -> Self {
    Rational { num, den }
Luca Barbato's avatar
Luca Barbato committed
40 41 42
  }
}

43 44 45 46 47 48 49
#[derive(Clone, Copy, Debug)]
#[repr(C)]
pub struct Point {
  pub x: u16,
  pub y: u16
}

50
#[derive(Clone, Debug)]
51
pub struct EncoderConfig {
52
  // output size
53 54
  pub width: usize,
  pub height: usize,
55 56

  // data format and ancillary color information
57 58 59
  pub bit_depth: usize,
  pub chroma_sampling: ChromaSampling,
  pub chroma_sample_position: ChromaSamplePosition,
60 61 62 63 64 65
  pub pixel_range: PixelRange,
  pub color_description: Option<ColorDescription>,
  pub mastering_display: Option<MasteringDisplay>,
  pub content_light: Option<ContentLight>,

  // encoder configuration
66
  pub time_base: Rational,
Josh Holmer's avatar
Josh Holmer committed
67 68 69 70
  /// The *minimum* interval between two keyframes
  pub min_key_frame_interval: u64,
  /// The *maximum* interval between two keyframes
  pub max_key_frame_interval: u64,
71
  pub low_latency: bool,
72
  pub quantizer: usize,
73
  pub bitrate: i32,
74 75
  pub tune: Tune,
  pub speed_settings: SpeedSettings,
76 77
  /// `None` for one-pass encode. `Some(1)` or `Some(2)` for two-pass encoding.
  pub pass: Option<u8>,
78
  pub show_psnr: bool,
79
  pub stats_file: Option<PathBuf>,
80 81 82 83
}

impl Default for EncoderConfig {
  fn default() -> Self {
84 85
    const DEFAULT_SPEED: usize = 3;
    Self::with_speed_preset(DEFAULT_SPEED)
86 87 88
  }
}

89 90 91
impl EncoderConfig {
  pub fn with_speed_preset(speed: usize) -> Self {
    EncoderConfig {
92 93
      width: 640,
      height: 480,
94

95 96 97
      bit_depth: 8,
      chroma_sampling: ChromaSampling::Cs420,
      chroma_sample_position: ChromaSamplePosition::Unknown,
98 99 100 101 102
      pixel_range: PixelRange::Unspecified,
      color_description: None,
      mastering_display: None,
      content_light: None,

103
      time_base: Rational { num: 30, den: 1 },
Josh Holmer's avatar
Josh Holmer committed
104 105
      min_key_frame_interval: 12,
      max_key_frame_interval: 240,
106
      low_latency: false,
107
      quantizer: 100,
108
      bitrate: 0,
109
      tune: Tune::Psnr,
110
      speed_settings: SpeedSettings::from_preset(speed),
111
      pass: None,
112
      show_psnr: false,
113
      stats_file: None,
114 115 116 117 118 119 120 121 122 123 124 125 126 127 128
    }
  }
}

#[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,
  pub encode_bottomup: bool,
  pub rdo_tx_decision: bool,
  pub prediction_modes: PredictionModesSetting,
  pub include_near_mvs: bool,
129
  pub no_scene_detection: bool,
130
  pub diamond_me: bool
131 132 133 134 135
}

impl Default for SpeedSettings {
  fn default() -> Self {
    SpeedSettings {
136
      min_block_size: BlockSize::BLOCK_16X16,
137 138 139 140 141 142 143 144 145
      multiref: false,
      fast_deblock: false,
      reduced_tx_set: false,
      tx_domain_distortion: false,
      encode_bottomup: false,
      rdo_tx_decision: false,
      prediction_modes: PredictionModesSetting::Simple,
      include_near_mvs: false,
      no_scene_detection: false,
146
      diamond_me: false,
147 148
    }
  }
149 150 151 152 153 154 155 156 157 158 159 160 161 162
}

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),
      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),
163
      no_scene_detection: Self::no_scene_detection_preset(speed),
164
      diamond_me: Self::diamond_me_preset(speed),
165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194
    }
  }

  fn min_block_size_preset(speed: usize) -> BlockSize {
    if speed <= 1 {
      BlockSize::BLOCK_4X4
    } else if speed <= 2 {
      BlockSize::BLOCK_8X8
    } else if speed <= 3 {
      BlockSize::BLOCK_16X16
    } else if speed <= 4 {
      BlockSize::BLOCK_32X32
    } else {
      BlockSize::BLOCK_64X64
    }
  }

  fn multiref_preset(speed: usize) -> bool {
    speed <= 2
  }

  fn fast_deblock_preset(speed: usize) -> bool {
    speed >= 4
  }

  fn reduced_tx_set_preset(speed: usize) -> bool {
    speed >= 2
  }

  fn tx_domain_distortion_preset(speed: usize) -> bool {
195
    speed >= 4
196 197 198 199 200 201 202 203 204 205 206
  }

  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 {
207
    if speed <= 1 {
208
      PredictionModesSetting::ComplexAll
209
    } else if speed <= 3 {
210 211 212 213 214 215 216 217 218
      PredictionModesSetting::ComplexKeyframes
    } else {
      PredictionModesSetting::Simple
    }
  }

  fn include_near_mvs_preset(speed: usize) -> bool {
    speed <= 2
  }
219 220 221 222

  fn no_scene_detection_preset(speed: usize) -> bool {
    speed == 10
  }
223 224 225 226

  fn diamond_me_preset(speed: usize) -> bool {
    speed >= 3
  }
227 228
}

229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250
#[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"),
    }
  }
}

251 252 253 254 255 256 257
#[derive(Clone, Copy, Debug, PartialOrd, PartialEq)]
pub enum PredictionModesSetting {
  Simple,
  ComplexKeyframes,
  ComplexAll,
}

258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299
#[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
  }
}

300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315
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
    }
}

316 317
arg_enum!{
  #[derive(Debug, Clone, Copy, PartialEq)]
318
  #[repr(C)]
319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334
  pub enum MatrixCoefficients {
      Identity = 0,
      BT709,
      Unspecified,
      BT470M = 4,
      BT470BG,
      ST170M,
      ST240M,
      YCgCo,
      BT2020NonConstantLuminance,
      BT2020ConstantLuminance,
      ST2085,
      ChromaticityDerivedNonConstantLuminance,
      ChromaticityDerivedConstantLuminance,
      ICtCp,
  }
335 336 337 338 339 340 341 342
}

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

343
arg_enum!{
344
  #[derive(Debug, Clone, Copy, PartialEq)]
345
  #[repr(C)]
346 347 348 349 350 351 352 353 354 355 356 357 358 359
  pub enum ColorPrimaries {
      BT709 = 1,
      Unspecified,
      BT470M = 4,
      BT470BG,
      ST170M,
      ST240M,
      Film,
      BT2020,
      ST428,
      P3DCI,
      P3Display,
      Tech3213 = 22,
  }
360 361 362 363 364 365 366 367
}

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

368
arg_enum!{
369
  #[derive(Debug, Clone, Copy, PartialEq)]
370
  #[repr(C)]
371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389
  pub enum TransferCharacteristics {
      BT1886 = 1,
      Unspecified,
      BT470M = 4,
      BT470BG,
      ST170M,
      ST240M,
      Linear,
      Logarithmic100,
      Logarithmic316,
      XVYCC,
      BT1361E,
      SRGB,
      BT2020Ten,
      BT2020Twelve,
      PerceptualQuantizer,
      ST428,
      HybridLogGamma,
  }
390 391 392 393 394 395 396 397
}

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

398 399 400 401 402 403 404
#[derive(Copy, Clone, Debug)]
pub struct ColorDescription {
    pub color_primaries: ColorPrimaries,
    pub transfer_characteristics: TransferCharacteristics,
    pub matrix_coefficients: MatrixCoefficients
}

405 406 407 408 409 410 411 412 413 414 415 416 417 418
#[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,
}

419
/// Contain all the encoder configuration
420
#[derive(Clone, Debug)]
421
pub struct Config {
Luca Barbato's avatar
Luca Barbato committed
422 423 424 425
  pub enc: EncoderConfig
}

impl Config {
Luca Barbato's avatar
Luca Barbato committed
426 427
  pub fn parse(&mut self, key: &str, value: &str) -> Result<(), EncoderStatus> {
    match key {
428
      "low_latency" => self.enc.low_latency = value.parse().map_err(|_e| ParseError)?,
Josh Holmer's avatar
Josh Holmer committed
429 430
      "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)?,
431
      "quantizer" => self.enc.quantizer = value.parse().map_err(|_e| ParseError)?,
432
      "speed" => self.enc.speed_settings = SpeedSettings::from_preset(value.parse().map_err(|_e| ParseError)?),
433 434
      "tune" => self.enc.tune = value.parse().map_err(|_e| ParseError)?,
      _ => return Err(InvalidKey)
Luca Barbato's avatar
Luca Barbato committed
435 436 437 438 439
    }

    Ok(())
  }

Luca Barbato's avatar
Luca Barbato committed
440
  pub fn new_context(&self) -> Context {
441
    #[cfg(feature = "aom")]
Luca Barbato's avatar
Luca Barbato committed
442
    unsafe {
Luca Barbato's avatar
Luca Barbato committed
443 444
      av1_rtcd();
      aom_dsp_rtcd();
Luca Barbato's avatar
Luca Barbato committed
445 446
    }

447 448 449
    // initialize with temporal delimiter
    let packet_data = TEMPORAL_DELIMITER.to_vec();

450 451 452 453 454 455
    let maybe_ac_qi_max = if self.enc.quantizer < 255 {
      Some(self.enc.quantizer as u8)
    } else {
      None
    };

Josh Holmer's avatar
Josh Holmer committed
456 457
    Context {
      frame_count: 0,
458
      limit: 0,
Josh Holmer's avatar
Josh Holmer committed
459
      idx: 0,
Josh Holmer's avatar
Josh Holmer committed
460
      frames_processed: 0,
Josh Holmer's avatar
Josh Holmer committed
461
      frame_q: BTreeMap::new(),
Josh Holmer's avatar
Josh Holmer committed
462 463
      frame_data: BTreeMap::new(),
      keyframes: BTreeSet::new(),
464
      packet_data,
Josh Holmer's avatar
Josh Holmer committed
465 466
      segment_start_idx: 0,
      segment_start_frame: 0,
467
      keyframe_detector: SceneChangeDetector::new(self.enc.bit_depth),
468
      config: self.clone(),
469 470 471 472 473 474 475 476 477 478
      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,
479 480 481
      first_pass_data: FirstPassData {
        frames: Vec::new(),
      },
Josh Holmer's avatar
Josh Holmer committed
482
    }
Luca Barbato's avatar
Luca Barbato committed
483 484 485 486
  }
}

pub struct Context {
Luca Barbato's avatar
Luca Barbato committed
487
  //    timebase: Rational,
fbossen's avatar
fbossen committed
488
  frame_count: u64,
489
  limit: u64,
490
  pub(crate) idx: u64,
Josh Holmer's avatar
Josh Holmer committed
491 492
  frames_processed: u64,
  /// Maps frame *number* to frames
493
  frame_q: BTreeMap<u64, Option<Arc<Frame>>>, //    packet_q: VecDeque<Packet>
Josh Holmer's avatar
Josh Holmer committed
494 495 496 497 498
  /// Maps frame *idx* to frame data
  frame_data: BTreeMap<u64, FrameInvariants>,
  /// 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>,
499
  /// A storage space for reordered frames.
Josh Holmer's avatar
Josh Holmer committed
500 501 502 503
  packet_data: Vec<u8>,
  segment_start_idx: u64,
  segment_start_frame: u64,
  keyframe_detector: SceneChangeDetector,
504
  pub config: Config,
505
  rc_state: RCState,
506
  maybe_prev_log_base_q: Option<i64>,
507
  pub first_pass_data: FirstPassData,
Luca Barbato's avatar
Luca Barbato committed
508 509 510 511 512 513 514 515 516
}

#[derive(Clone, Copy, Debug)]
pub enum EncoderStatus {
  /// The encoder needs more Frames to produce an output Packet
  NeedMoreData,
  /// There are enough Frames queue
  EnoughData,
  ///
Luca Barbato's avatar
Luca Barbato committed
517 518 519
  Failure,
  InvalidKey,
  ParseError
Luca Barbato's avatar
Luca Barbato committed
520 521 522 523
}

pub struct Packet {
  pub data: Vec<u8>,
fbossen's avatar
fbossen committed
524 525
  pub rec: Option<Frame>,
  pub number: u64,
526 527 528
  pub frame_type: FrameType,
  /// PSNR for Y, U, and V planes
  pub psnr: Option<(f64, f64, f64)>,
529 530 531 532
}

impl fmt::Display for Packet {
  fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
Luca Barbato's avatar
Luca Barbato committed
533 534 535 536 537 538 539
    write!(
      f,
      "Frame {} - {} - {} bytes",
      self.number,
      self.frame_type,
      self.data.len()
    )
540
  }
Luca Barbato's avatar
Luca Barbato committed
541 542 543 544
}

impl Context {
  pub fn new_frame(&self) -> Arc<Frame> {
Josh Holmer's avatar
Josh Holmer committed
545
    Arc::new(Frame::new(
546 547
      self.config.enc.width,
      self.config.enc.height,
548
      self.config.enc.chroma_sampling
Josh Holmer's avatar
Josh Holmer committed
549
    ))
Luca Barbato's avatar
Luca Barbato committed
550 551 552 553 554 555
  }

  pub fn send_frame<F>(&mut self, frame: F) -> Result<(), EncoderStatus>
  where
    F: Into<Option<Arc<Frame>>>
  {
Josh Holmer's avatar
Josh Holmer committed
556 557
    let idx = self.frame_count;
    self.frame_q.insert(idx, frame.into());
558
    self.frame_count += 1;
Luca Barbato's avatar
Luca Barbato committed
559 560 561
    Ok(())
  }

562 563 564 565
  pub fn get_frame_count(&self) -> u64 {
    self.frame_count
  }

566 567
  pub fn set_limit(&mut self, limit: u64) {
    self.limit = limit;
568 569
  }

Josh Holmer's avatar
Josh Holmer committed
570 571 572 573
  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)
  }

574
  pub fn needs_more_frames(&self, frame_count: u64) -> bool {
575
    self.limit == 0 || frame_count < self.limit
576 577
  }

578 579 580
  pub fn container_sequence_header(&mut self) -> Vec<u8> {
    fn sequence_header_inner(seq: &Sequence) -> io::Result<Vec<u8>> {
      let mut buf = Vec::new();
581

582
      {
583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599
        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
600
      }
601

602 603 604
      Ok(buf)
    }

Josh Holmer's avatar
Josh Holmer committed
605
    sequence_header_inner(&self.frame_data[&0].sequence).unwrap()
606 607
  }

Josh Holmer's avatar
Josh Holmer committed
608
  fn next_keyframe(&self) -> u64 {
Josh Holmer's avatar
Josh Holmer committed
609 610 611 612
    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
613 614 615 616 617
    if next_detected.is_none() {
      return next_limit;
    }
    cmp::min(next_detected.unwrap(), next_limit)
  }
fbossen's avatar
fbossen committed
618

619
  fn set_frame_properties(&mut self, idx: u64) -> bool {
620
    let (fi, end_of_subgop) = self.build_frame_properties(idx);
621 622 623
    self.frame_data.insert(idx, fi);

    end_of_subgop
Josh Holmer's avatar
Josh Holmer committed
624 625
  }

626
  fn build_frame_properties(&mut self, idx: u64) -> (FrameInvariants, bool) {
Josh Holmer's avatar
Josh Holmer committed
627
    if idx == 0 {
628
      let seq = Sequence::new(&self.config.enc);
629

Josh Holmer's avatar
Josh Holmer committed
630
      // The first frame will always be a key frame
Josh Holmer's avatar
Josh Holmer committed
631 632
      let fi = FrameInvariants::new_key_frame(
        &FrameInvariants::new(
633
          self.config.enc.clone(),
634
          seq
Josh Holmer's avatar
Josh Holmer committed
635 636 637
        ),
        0
      );
638
      return (fi, true);
Josh Holmer's avatar
Josh Holmer committed
639 640
    }

Josh Holmer's avatar
Josh Holmer committed
641 642
    let mut fi = self.frame_data[&(idx - 1)].clone();

643 644 645 646 647
    // 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
648 649 650 651 652 653
    // 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 {
654
      let next_keyframe = if keyframe_only { self.segment_start_frame + 1 } else { self.next_keyframe() };
655
      let (fi_temp, end_of_subgop) = FrameInvariants::new_inter_frame(
Josh Holmer's avatar
Josh Holmer committed
656
        &fi,
657 658 659 660
        self.segment_start_frame,
        idx_in_segment,
        next_keyframe
      );
Josh Holmer's avatar
Josh Holmer committed
661
      fi = fi_temp;
662
      if !end_of_subgop {
Josh Holmer's avatar
Josh Holmer committed
663 664 665
        if !fi.inter_cfg.unwrap().reorder
          || ((idx_in_segment - 1) % fi.inter_cfg.unwrap().group_len == 0
          && fi.number == (next_keyframe - 1))
666
        {
Josh Holmer's avatar
Josh Holmer committed
667
          self.segment_start_idx = idx;
Josh Holmer's avatar
Josh Holmer committed
668
          self.segment_start_frame = next_keyframe;
Josh Holmer's avatar
Josh Holmer committed
669
          fi.number = next_keyframe;
Josh Holmer's avatar
Josh Holmer committed
670
        } else {
671
          return (fi, false);
Josh Holmer's avatar
Josh Holmer committed
672 673
        }
      }
fbossen's avatar
fbossen committed
674
    }
Luca Barbato's avatar
Luca Barbato committed
675

676 677
    match self.frame_q.get(&fi.number) {
      Some(Some(_)) => {},
Josh Holmer's avatar
Josh Holmer committed
678
      _ => { return (fi, false); }
679 680
    }

Josh Holmer's avatar
Josh Holmer committed
681
    // Now that we know the frame number, look up the correct frame type
Josh Holmer's avatar
Josh Holmer committed
682 683 684 685 686 687 688
    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
689

Josh Holmer's avatar
Josh Holmer committed
690 691 692 693 694
    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();
695
      let (fi_temp, end_of_subgop) = FrameInvariants::new_inter_frame(
Josh Holmer's avatar
Josh Holmer committed
696 697 698 699 700 701
        &fi,
        self.segment_start_frame,
        idx_in_segment,
        next_keyframe
      );
      fi = fi_temp;
702 703
      if !end_of_subgop {
        return (fi, false);
Josh Holmer's avatar
Josh Holmer committed
704 705
      }
    }
706
    (fi, true)
fbossen's avatar
fbossen committed
707
  }
Luca Barbato's avatar
Luca Barbato committed
708

fbossen's avatar
fbossen committed
709
  pub fn receive_packet(&mut self) -> Result<Packet, EncoderStatus> {
710 711 712 713 714 715
    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
716

717 718 719 720 721 722
      if !self.needs_more_frames(self.frame_data.get(&idx).unwrap().number) {
        self.idx += 1;
        return Err(EncoderStatus::EnoughData);
      }
      idx
    };
723

724
    let ret = {
Luca Barbato's avatar
Luca Barbato committed
725 726 727
      let fi = self.frame_data.get_mut(&idx).unwrap();
      if fi.show_existing_frame {
        self.idx += 1;
Luca Barbato's avatar
Luca Barbato committed
728

Luca Barbato's avatar
Luca Barbato committed
729
        let mut fs = FrameState::new(fi);
fbossen's avatar
fbossen committed
730

731 732
        // TODO: Record the bits spent here against the original frame for rate
        //  control purposes, or add a new frame subtype?
733 734
        let sef_data = encode_frame(fi, &mut fs);
        self.packet_data.extend(sef_data);
Luca Barbato's avatar
Luca Barbato committed
735

Luca Barbato's avatar
Luca Barbato committed
736
        let rec = if fi.show_frame { Some(fs.rec) } else { None };
737 738
        let fi = fi.clone();
        self.finalize_packet(&*fs.input, rec, &fi)
Luca Barbato's avatar
Luca Barbato committed
739 740 741
      } else {
        if let Some(f) = self.frame_q.get(&fi.number) {
          self.idx += 1;
fbossen's avatar
fbossen committed
742

743
          if let Some(frame) = f.clone() {
744
            let fti = fi.get_frame_subtype();
745 746 747
            let qps =
              self.rc_state.select_qi(self, fti, self.maybe_prev_log_base_q);
            let fi = self.frame_data.get_mut(&idx).unwrap();
748
            fi.set_quantizers(&qps);
Luca Barbato's avatar
Luca Barbato committed
749
            let mut fs = FrameState::new_with_frame(fi, frame.clone());
fbossen's avatar
fbossen committed
750

751
            // TODO: Trial encoding for first frame of each type.
Luca Barbato's avatar
Luca Barbato committed
752
            let data = encode_frame(fi, &mut fs);
753 754 755 756 757 758 759 760
            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
761
            self.packet_data.extend(data);
fbossen's avatar
fbossen committed
762

Luca Barbato's avatar
Luca Barbato committed
763
            fs.rec.pad(fi.width, fi.height);
fbossen's avatar
fbossen committed
764

Luca Barbato's avatar
Luca Barbato committed
765 766
            // 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
767

Luca Barbato's avatar
Luca Barbato committed
768
            update_rec_buffer(fi, fs);
fbossen's avatar
fbossen committed
769

Luca Barbato's avatar
Luca Barbato committed
770
            if fi.show_frame {
771 772
              let fi = fi.clone();
              self.finalize_packet(&*frame, rec, &fi)
Luca Barbato's avatar
Luca Barbato committed
773 774 775
            } else {
              Err(EncoderStatus::NeedMoreData)
            }
776 777 778
          } else {
            Err(EncoderStatus::NeedMoreData)
          }
fbossen's avatar
fbossen committed
779 780 781 782
        } else {
          Err(EncoderStatus::NeedMoreData)
        }
      }
783 784 785
    };

    if let Ok(ref pkt) = ret {
Luca Barbato's avatar
Luca Barbato committed
786
      self.garbage_collect(pkt.number);
787 788 789
    }

    ret
Luca Barbato's avatar
Luca Barbato committed
790 791
  }

792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809
  fn finalize_packet(&mut self, original_frame: &Frame, rec: Option<Frame>, fi: &FrameInvariants) -> Result<Packet, EncoderStatus> {
    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 {
        psnr = Some(calculate_frame_psnr(
          &*original_frame,
          rec,
          fi.sequence.bit_depth
        ));
      }
    }

810 811 812 813
    if self.config.enc.pass == Some(1) {
      self.first_pass_data.frames.push(FirstPassFrame::from(fi));
    }

814 815 816 817 818 819 820 821 822 823
    self.frames_processed += 1;
    Ok(Packet {
      data,
      rec,
      number: fi.number,
      frame_type: fi.frame_type,
      psnr
    })
  }

824
  fn garbage_collect(&mut self, cur_frame: u64) {
Josh Holmer's avatar
Josh Holmer committed
825 826 827 828 829 830 831 832 833 834 835 836 837 838
    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
839
  pub fn flush(&mut self) {
fbossen's avatar
fbossen committed
840
    self.frame_q.insert(self.frame_count, None);
841
    self.frame_count += 1;
Luca Barbato's avatar
Luca Barbato committed
842
  }
Josh Holmer's avatar
Josh Holmer committed
843

Josh Holmer's avatar
Josh Holmer committed
844 845
  fn determine_frame_type(&mut self, frame_number: u64) -> FrameType {
    if frame_number == 0 {
Josh Holmer's avatar
Josh Holmer committed
846 847
      return FrameType::KEY;
    }
848 849 850 851 852 853 854
    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
855

Josh Holmer's avatar
Josh Holmer committed
856 857 858 859 860 861 862 863
    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
864
    if let Some(frame) = frame {
Josh Holmer's avatar
Josh Holmer committed
865 866 867
      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
868 869
          // 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
870
          self.keyframe_detector.detect_scene_change(frame, frame_number as usize);
Josh Holmer's avatar
Josh Holmer committed
871 872 873
        }
        return FrameType::INTER;
      }
Josh Holmer's avatar
Josh Holmer committed
874
      if distance >= self.config.enc.max_key_frame_interval {
Josh Holmer's avatar
Josh Holmer committed
875 876
        return FrameType::KEY;
      }
Josh Holmer's avatar
Josh Holmer committed
877
      if self.keyframe_detector.detect_scene_change(frame, frame_number as usize) {
Josh Holmer's avatar
Josh Holmer committed
878 879 880 881 882
        return FrameType::KEY;
      }
    }
    FrameType::INTER
  }
883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955

  // 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
956
}
957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975

#[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,
}

impl From<&FrameInvariants> for FirstPassFrame {
  fn from(fi: &FrameInvariants) -> FirstPassFrame {
    FirstPassFrame {
      number: fi.number,
      frame_type: fi.frame_type,
    }
  }
976
}