1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364
// Copyright (c) 2020, 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. use num_derive::*; use crate::partition::BlockSize; use crate::serialize::{Deserialize, Serialize}; use std::fmt; // NOTE: Add Structures at the end. /// Contains the speed settings. #[derive(Clone, Copy, Debug, Serialize, Deserialize)] pub struct SpeedSettings { /// Enables inter-frames to have multiple reference frames. /// /// Enabled is slower. pub multiref: bool, /// Enables fast deblocking filter. pub fast_deblock: bool, /// Enables reduced transform set. /// /// Enabled is faster. pub reduced_tx_set: bool, /// Enables using transform-domain distortion instead of pixel-domain. /// /// Enabled is faster. pub tx_domain_distortion: bool, /// Enables using transform-domain rate estimation. /// /// Enabled is faster. pub tx_domain_rate: bool, /// Enables bottom-up encoding, rather than top-down. /// /// Enabled is slower. pub encode_bottomup: bool, /// Enables searching transform size and type with RDO. /// /// Enabled is slower. pub rdo_tx_decision: bool, /// Prediction modes to search. /// /// Complex settings are slower. pub prediction_modes: PredictionModesSetting, /// Enables searching near motion vectors during RDO. /// /// Enabled is slower. pub include_near_mvs: bool, /// Disables scene-cut detection. /// /// Enabled is faster. pub no_scene_detection: bool, /// Fast scene detection mode, uses simple SAD instead of encoder cost estimates. pub fast_scene_detection: bool, /// Enables CDEF. pub cdef: bool, /// Enables LRF. pub lrf: bool, /// The amount of search done for self guided restoration. pub sgr_complexity: SGRComplexityLevel, /// Use SATD instead of SAD for subpixel search. /// /// Enabled is slower. pub use_satd_subpel: bool, /// Use non-square partition type everywhere /// /// Enabled is slower. pub non_square_partition: bool, /// Use segmentation. pub enable_segmentation: bool, /// Enable tx split for inter mode block. pub enable_inter_tx_split: bool, /// Use fine directional intra prediction pub fine_directional_intra: bool, // NOTE: put enums and basic type fields above /// Range of partition sizes that can be used. Larger ranges are slower. /// /// Must be based on square block sizes, so e.g. 8×4 isn't allowed here. pub partition_range: PartitionRange, } impl Default for SpeedSettings { /// This is currently used exclusively for feature testing and comparison. /// It is set to the slowest settings possible. fn default() -> Self { SpeedSettings { partition_range: PartitionRange::new( BlockSize::BLOCK_4X4, BlockSize::BLOCK_64X64, ), multiref: true, fast_deblock: false, reduced_tx_set: false, tx_domain_distortion: false, tx_domain_rate: false, encode_bottomup: true, rdo_tx_decision: true, prediction_modes: PredictionModesSetting::ComplexAll, include_near_mvs: true, no_scene_detection: false, fast_scene_detection: false, cdef: true, lrf: false, sgr_complexity: SGRComplexityLevel::Full, use_satd_subpel: true, non_square_partition: true, enable_segmentation: true, enable_inter_tx_split: false, fine_directional_intra: false, } } } impl SpeedSettings { /// Set the speed setting according to a numeric speed preset. /// /// The speed settings vary depending on speed value from 0 to 10. /// - 10 (fastest): min block size 64x64, reduced TX set, fast deblock, fast scenechange detection. /// - 9: min block size 32x32, reduced TX set, fast deblock. /// - 8: min block size 8x8, reduced TX set, fast deblock. /// - 7: min block size 8x8, reduced TX set. /// - 6 (default): min block size 8x8, reduced TX set, complex pred modes for keyframes. /// - 5: min block size 8x8, complex pred modes for keyframes, RDO TX decision. /// - 4: min block size 8x8, complex pred modes for keyframes, RDO TX decision, full SGR search. /// - 3: min block size 8x8, complex pred modes for keyframes, RDO TX decision, include near MVs, /// full SGR search. /// - 2: min block size 4x4, complex pred modes, RDO TX decision, include near MVs, /// full SGR search. /// - 1: min block size 4x4, complex pred modes, RDO TX decision, include near MVs, /// bottom-up encoding, full SGR search. /// - 0 (slowest): min block size 4x4, complex pred modes, RDO TX decision, include near MVs, /// bottom-up encoding with non-square partitions everywhere, full SGR search. pub fn from_preset(speed: usize) -> Self { SpeedSettings { partition_range: Self::partition_range_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), tx_domain_rate: Self::tx_domain_rate_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), no_scene_detection: Self::no_scene_detection_preset(speed), fast_scene_detection: Self::fast_scene_detection_preset(speed), cdef: Self::cdef_preset(speed), lrf: Self::lrf_preset(speed), sgr_complexity: Self::sgr_complexity_preset(speed), use_satd_subpel: Self::use_satd_subpel(speed), non_square_partition: Self::non_square_partition_preset(speed), enable_segmentation: Self::enable_segmentation_preset(speed), enable_inter_tx_split: Self::enable_inter_tx_split_preset(speed), fine_directional_intra: Self::fine_directional_intra_preset(speed), } } /// This preset is set this way because 8x8 with reduced TX set is faster but with equivalent /// or better quality compared to 16x16 (to which reduced TX set does not apply). fn partition_range_preset(speed: usize) -> PartitionRange { if speed <= 2 { PartitionRange::new(BlockSize::BLOCK_4X4, BlockSize::BLOCK_64X64) } else if speed <= 8 { PartitionRange::new(BlockSize::BLOCK_8X8, BlockSize::BLOCK_64X64) } else if speed <= 9 { PartitionRange::new(BlockSize::BLOCK_32X32, BlockSize::BLOCK_64X64) } else { PartitionRange::new(BlockSize::BLOCK_32X32, BlockSize::BLOCK_32X32) } } /// Multiref is enabled automatically if low_latency is false. /// /// If low_latency is true, enabling multiref allows using multiple /// backwards references. low_latency false enables both forward and /// backwards references. const fn multiref_preset(speed: usize) -> bool { speed <= 7 } const fn fast_deblock_preset(speed: usize) -> bool { speed >= 8 } const fn reduced_tx_set_preset(speed: usize) -> bool { speed >= 6 } /// TX domain distortion is always faster, with no significant quality change const fn tx_domain_distortion_preset(_speed: usize) -> bool { true } const fn tx_domain_rate_preset(_speed: usize) -> bool { false } const fn encode_bottomup_preset(speed: usize) -> bool { speed <= 1 } const fn rdo_tx_decision_preset(speed: usize) -> bool { speed <= 5 } fn prediction_modes_preset(speed: usize) -> PredictionModesSetting { if speed <= 2 { PredictionModesSetting::ComplexAll } else if speed <= 6 { PredictionModesSetting::ComplexKeyframes } else { PredictionModesSetting::Simple } } const fn include_near_mvs_preset(speed: usize) -> bool { speed <= 3 } const fn no_scene_detection_preset(_speed: usize) -> bool { false } const fn fast_scene_detection_preset(speed: usize) -> bool { speed == 10 } const fn cdef_preset(_speed: usize) -> bool { true } const fn lrf_preset(speed: usize) -> bool { speed <= 9 } fn sgr_complexity_preset(speed: usize) -> SGRComplexityLevel { if speed <= 4 { SGRComplexityLevel::Full } else { SGRComplexityLevel::Reduced } } const fn use_satd_subpel(speed: usize) -> bool { speed <= 9 } const fn non_square_partition_preset(speed: usize) -> bool { speed == 0 } // FIXME: this is currently only enabled at speed 0 because choosing a segment // requires doing RDO, but once that is replaced by a less bruteforce // solution we should be able to enable segmentation at all speeds. const fn enable_segmentation_preset(speed: usize) -> bool { speed == 0 } // FIXME: With unknown reasons, inter_tx_split does not work if reduced_tx_set is false const fn enable_inter_tx_split_preset(speed: usize) -> bool { speed >= 9 } fn fine_directional_intra_preset(_speed: usize) -> bool { true } } /// Range of block sizes to use. #[derive(Clone, Copy, Debug, Serialize, Deserialize)] pub struct PartitionRange { pub(crate) min: BlockSize, pub(crate) max: BlockSize, } impl PartitionRange { /// Creates a new partition range with min and max partition sizes. pub fn new(min: BlockSize, max: BlockSize) -> Self { assert!(max >= min); // Topdown search checks the min block size for PARTITION_SPLIT only, so // the min block size must be square. assert!(min.is_sqr()); // Rectangular max partition sizes have not been tested. assert!(max.is_sqr()); Self { min, max } } } /// Prediction modes to search. #[derive( Clone, Copy, Debug, PartialOrd, PartialEq, FromPrimitive, Serialize, Deserialize, )] pub enum PredictionModesSetting { /// Only simple prediction modes. Simple, /// Search all prediction modes on key frames and simple modes on other /// frames. ComplexKeyframes, /// Search all prediction modes on all frames. ComplexAll, } impl fmt::Display for PredictionModesSetting { fn fmt(&self, f: &mut fmt::Formatter) -> Result<(), fmt::Error> { write!( f, "{}", match self { PredictionModesSetting::Simple => "Simple", PredictionModesSetting::ComplexKeyframes => "Complex-KFs", PredictionModesSetting::ComplexAll => "Complex-All", } ) } } /// Search level for self guided restoration #[derive( Clone, Copy, Debug, PartialOrd, PartialEq, FromPrimitive, Serialize, Deserialize, )] pub enum SGRComplexityLevel { /// Search all sgr parameters Full, /// Search a reduced set of sgr parameters Reduced, } impl fmt::Display for SGRComplexityLevel { fn fmt(&self, f: &mut fmt::Formatter) -> Result<(), fmt::Error> { write!( f, "{}", match self { SGRComplexityLevel::Full => "Full", SGRComplexityLevel::Reduced => "Reduced", } ) } }