Alex Loiko | 44c21f4 | 2019-04-25 15:09:32 +0200 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2019 The WebRTC project authors. All Rights Reserved. |
| 3 | * |
| 4 | * Use of this source code is governed by a BSD-style license |
| 5 | * that can be found in the LICENSE file in the root of the source |
| 6 | * tree. An additional intellectual property rights grant can be found |
| 7 | * in the file PATENTS. All contributing project authors may |
| 8 | * be found in the AUTHORS file in the root of the source tree. |
| 9 | */ |
| 10 | |
| 11 | /* |
| 12 | * LEFT TO DO: |
| 13 | * - WRITE TESTS for the stuff in this file. |
| 14 | * - Check the creation, maybe make it safer by returning an empty optional or |
| 15 | * unique_ptr. --- It looks OK, but RecreateEncoderInstance can perhaps crash |
| 16 | * on a valid config. Can run it in the fuzzer for some time. Should prbl also |
| 17 | * fuzz the config. |
| 18 | */ |
| 19 | |
| 20 | #include "modules/audio_coding/codecs/opus/audio_encoder_multi_channel_opus_impl.h" |
| 21 | |
| 22 | #include <algorithm> |
| 23 | #include <memory> |
| 24 | #include <string> |
| 25 | #include <vector> |
| 26 | |
| 27 | #include "absl/memory/memory.h" |
| 28 | #include "absl/strings/match.h" |
| 29 | #include "modules/audio_coding/codecs/opus/audio_coder_opus_common.h" |
| 30 | #include "rtc_base/arraysize.h" |
| 31 | #include "rtc_base/checks.h" |
| 32 | #include "rtc_base/logging.h" |
| 33 | #include "rtc_base/string_to_number.h" |
| 34 | |
| 35 | namespace webrtc { |
| 36 | |
| 37 | namespace { |
| 38 | |
| 39 | // Recommended bitrates for one channel: |
| 40 | // 8-12 kb/s for NB speech, |
| 41 | // 16-20 kb/s for WB speech, |
| 42 | // 28-40 kb/s for FB speech, |
| 43 | // 48-64 kb/s for FB mono music, and |
| 44 | // 64-128 kb/s for FB stereo music. |
| 45 | // The current implementation multiplies these values by the number of channels. |
| 46 | constexpr int kOpusBitrateNbBps = 12000; |
| 47 | constexpr int kOpusBitrateWbBps = 20000; |
| 48 | constexpr int kOpusBitrateFbBps = 32000; |
| 49 | |
| 50 | constexpr int kDefaultMaxPlaybackRate = 48000; |
| 51 | // These two lists must be sorted from low to high |
| 52 | #if WEBRTC_OPUS_SUPPORT_120MS_PTIME |
| 53 | constexpr int kOpusSupportedFrameLengths[] = {10, 20, 40, 60, 120}; |
| 54 | #else |
| 55 | constexpr int kOpusSupportedFrameLengths[] = {10, 20, 40, 60}; |
| 56 | #endif |
| 57 | |
| 58 | int GetBitrateBps(const AudioEncoderMultiChannelOpusConfig& config) { |
| 59 | RTC_DCHECK(config.IsOk()); |
| 60 | return config.bitrate_bps; |
| 61 | } |
| 62 | int GetMaxPlaybackRate(const SdpAudioFormat& format) { |
| 63 | const auto param = GetFormatParameter<int>(format, "maxplaybackrate"); |
| 64 | if (param && *param >= 8000) { |
| 65 | return std::min(*param, kDefaultMaxPlaybackRate); |
| 66 | } |
| 67 | return kDefaultMaxPlaybackRate; |
| 68 | } |
| 69 | |
| 70 | int GetFrameSizeMs(const SdpAudioFormat& format) { |
| 71 | const auto ptime = GetFormatParameter<int>(format, "ptime"); |
| 72 | if (ptime.has_value()) { |
| 73 | // Pick the next highest supported frame length from |
| 74 | // kOpusSupportedFrameLengths. |
| 75 | for (const int supported_frame_length : kOpusSupportedFrameLengths) { |
| 76 | if (supported_frame_length >= *ptime) { |
| 77 | return supported_frame_length; |
| 78 | } |
| 79 | } |
| 80 | // If none was found, return the largest supported frame length. |
| 81 | return *(std::end(kOpusSupportedFrameLengths) - 1); |
| 82 | } |
| 83 | |
| 84 | return AudioEncoderOpusConfig::kDefaultFrameSizeMs; |
| 85 | } |
| 86 | |
| 87 | int CalculateDefaultBitrate(int max_playback_rate, size_t num_channels) { |
| 88 | const int bitrate = [&] { |
| 89 | if (max_playback_rate <= 8000) { |
| 90 | return kOpusBitrateNbBps * rtc::dchecked_cast<int>(num_channels); |
| 91 | } else if (max_playback_rate <= 16000) { |
| 92 | return kOpusBitrateWbBps * rtc::dchecked_cast<int>(num_channels); |
| 93 | } else { |
| 94 | return kOpusBitrateFbBps * rtc::dchecked_cast<int>(num_channels); |
| 95 | } |
| 96 | }(); |
| 97 | RTC_DCHECK_GE(bitrate, AudioEncoderMultiChannelOpusConfig::kMinBitrateBps); |
| 98 | return bitrate; |
| 99 | } |
| 100 | |
| 101 | // Get the maxaveragebitrate parameter in string-form, so we can properly figure |
| 102 | // out how invalid it is and accurately log invalid values. |
| 103 | int CalculateBitrate(int max_playback_rate_hz, |
| 104 | size_t num_channels, |
| 105 | absl::optional<std::string> bitrate_param) { |
| 106 | const int default_bitrate = |
| 107 | CalculateDefaultBitrate(max_playback_rate_hz, num_channels); |
| 108 | |
| 109 | if (bitrate_param) { |
| 110 | const auto bitrate = rtc::StringToNumber<int>(*bitrate_param); |
| 111 | if (bitrate) { |
| 112 | const int chosen_bitrate = |
| 113 | std::max(AudioEncoderOpusConfig::kMinBitrateBps, |
| 114 | std::min(*bitrate, AudioEncoderOpusConfig::kMaxBitrateBps)); |
| 115 | if (bitrate != chosen_bitrate) { |
| 116 | RTC_LOG(LS_WARNING) << "Invalid maxaveragebitrate " << *bitrate |
| 117 | << " clamped to " << chosen_bitrate; |
| 118 | } |
| 119 | return chosen_bitrate; |
| 120 | } |
| 121 | RTC_LOG(LS_WARNING) << "Invalid maxaveragebitrate \"" << *bitrate_param |
| 122 | << "\" replaced by default bitrate " << default_bitrate; |
| 123 | } |
| 124 | |
| 125 | return default_bitrate; |
| 126 | } |
| 127 | |
| 128 | } // namespace |
| 129 | |
| 130 | std::unique_ptr<AudioEncoder> |
| 131 | AudioEncoderMultiChannelOpusImpl::MakeAudioEncoder( |
| 132 | const AudioEncoderMultiChannelOpusConfig& config, |
| 133 | int payload_type) { |
| 134 | if (!config.IsOk()) { |
| 135 | return nullptr; |
| 136 | } |
| 137 | return absl::make_unique<AudioEncoderMultiChannelOpusImpl>(config, |
| 138 | payload_type); |
| 139 | } |
| 140 | |
| 141 | AudioEncoderMultiChannelOpusImpl::AudioEncoderMultiChannelOpusImpl( |
| 142 | const AudioEncoderMultiChannelOpusConfig& config, |
| 143 | int payload_type) |
| 144 | : payload_type_(payload_type), inst_(nullptr) { |
| 145 | RTC_DCHECK(0 <= payload_type && payload_type <= 127); |
| 146 | |
| 147 | RTC_CHECK(RecreateEncoderInstance(config)); |
| 148 | } |
| 149 | |
| 150 | AudioEncoderMultiChannelOpusImpl::~AudioEncoderMultiChannelOpusImpl() { |
| 151 | RTC_CHECK_EQ(0, WebRtcOpus_EncoderFree(inst_)); |
| 152 | } |
| 153 | |
| 154 | size_t AudioEncoderMultiChannelOpusImpl::SufficientOutputBufferSize() const { |
| 155 | // Calculate the number of bytes we expect the encoder to produce, |
| 156 | // then multiply by two to give a wide margin for error. |
| 157 | const size_t bytes_per_millisecond = |
| 158 | static_cast<size_t>(GetBitrateBps(config_) / (1000 * 8) + 1); |
| 159 | const size_t approx_encoded_bytes = |
| 160 | Num10msFramesPerPacket() * 10 * bytes_per_millisecond; |
| 161 | return 2 * approx_encoded_bytes; |
| 162 | } |
| 163 | |
| 164 | void AudioEncoderMultiChannelOpusImpl::Reset() { |
| 165 | RTC_CHECK(RecreateEncoderInstance(config_)); |
| 166 | } |
| 167 | |
| 168 | // If the given config is OK, recreate the Opus encoder instance with those |
| 169 | // settings, save the config, and return true. Otherwise, do nothing and return |
| 170 | // false. |
| 171 | bool AudioEncoderMultiChannelOpusImpl::RecreateEncoderInstance( |
| 172 | const AudioEncoderMultiChannelOpusConfig& config) { |
| 173 | if (!config.IsOk()) |
| 174 | return false; |
| 175 | config_ = config; |
| 176 | if (inst_) |
| 177 | RTC_CHECK_EQ(0, WebRtcOpus_EncoderFree(inst_)); |
| 178 | input_buffer_.clear(); |
| 179 | input_buffer_.reserve(Num10msFramesPerPacket() * SamplesPer10msFrame()); |
| 180 | RTC_CHECK_EQ( |
| 181 | 0, WebRtcOpus_MultistreamEncoderCreate( |
| 182 | &inst_, config.num_channels, |
| 183 | config.application == |
| 184 | AudioEncoderMultiChannelOpusConfig::ApplicationMode::kVoip |
| 185 | ? 0 |
| 186 | : 1, |
| 187 | config.num_streams, config.coupled_streams, |
| 188 | config.channel_mapping.data())); |
| 189 | const int bitrate = GetBitrateBps(config); |
| 190 | RTC_CHECK_EQ(0, WebRtcOpus_SetBitRate(inst_, bitrate)); |
| 191 | RTC_LOG(LS_VERBOSE) << "Set Opus bitrate to " << bitrate << " bps."; |
| 192 | if (config.fec_enabled) { |
| 193 | RTC_CHECK_EQ(0, WebRtcOpus_EnableFec(inst_)); |
| 194 | RTC_LOG(LS_VERBOSE) << "Opus enable FEC"; |
| 195 | } else { |
| 196 | RTC_CHECK_EQ(0, WebRtcOpus_DisableFec(inst_)); |
| 197 | RTC_LOG(LS_VERBOSE) << "Opus disable FEC"; |
| 198 | } |
| 199 | RTC_CHECK_EQ( |
| 200 | 0, WebRtcOpus_SetMaxPlaybackRate(inst_, config.max_playback_rate_hz)); |
| 201 | RTC_LOG(LS_VERBOSE) << "Set Opus playback rate to " |
| 202 | << config.max_playback_rate_hz << " hz."; |
| 203 | |
| 204 | // Use the DEFAULT complexity. |
| 205 | RTC_CHECK_EQ( |
| 206 | 0, WebRtcOpus_SetComplexity(inst_, AudioEncoderOpusConfig().complexity)); |
| 207 | RTC_LOG(LS_VERBOSE) << "Set Opus coding complexity to " |
| 208 | << AudioEncoderOpusConfig().complexity; |
| 209 | |
| 210 | if (config.dtx_enabled) { |
| 211 | RTC_CHECK_EQ(0, WebRtcOpus_EnableDtx(inst_)); |
| 212 | RTC_LOG(LS_VERBOSE) << "Opus enable DTX"; |
| 213 | } else { |
| 214 | RTC_CHECK_EQ(0, WebRtcOpus_DisableDtx(inst_)); |
| 215 | RTC_LOG(LS_VERBOSE) << "Opus disable DTX"; |
| 216 | } |
| 217 | |
| 218 | if (config.cbr_enabled) { |
| 219 | RTC_CHECK_EQ(0, WebRtcOpus_EnableCbr(inst_)); |
| 220 | RTC_LOG(LS_VERBOSE) << "Opus enable CBR"; |
| 221 | } else { |
| 222 | RTC_CHECK_EQ(0, WebRtcOpus_DisableCbr(inst_)); |
| 223 | RTC_LOG(LS_VERBOSE) << "Opus disable CBR"; |
| 224 | } |
| 225 | num_channels_to_encode_ = NumChannels(); |
| 226 | next_frame_length_ms_ = config_.frame_size_ms; |
| 227 | RTC_LOG(LS_VERBOSE) << "Set Opus frame length to " << config_.frame_size_ms |
| 228 | << " ms"; |
| 229 | return true; |
| 230 | } |
| 231 | |
| 232 | absl::optional<AudioEncoderMultiChannelOpusConfig> |
| 233 | AudioEncoderMultiChannelOpusImpl::SdpToConfig(const SdpAudioFormat& format) { |
| 234 | if (!absl::EqualsIgnoreCase(format.name, "multiopus") || |
| 235 | format.clockrate_hz != 48000) { |
| 236 | return absl::nullopt; |
| 237 | } |
| 238 | |
| 239 | AudioEncoderMultiChannelOpusConfig config; |
| 240 | config.num_channels = format.num_channels; |
| 241 | config.frame_size_ms = GetFrameSizeMs(format); |
| 242 | config.max_playback_rate_hz = GetMaxPlaybackRate(format); |
| 243 | config.fec_enabled = (GetFormatParameter(format, "useinbandfec") == "1"); |
| 244 | config.dtx_enabled = (GetFormatParameter(format, "usedtx") == "1"); |
| 245 | config.cbr_enabled = (GetFormatParameter(format, "cbr") == "1"); |
| 246 | config.bitrate_bps = |
| 247 | CalculateBitrate(config.max_playback_rate_hz, config.num_channels, |
| 248 | GetFormatParameter(format, "maxaveragebitrate")); |
| 249 | config.application = |
| 250 | config.num_channels == 1 |
| 251 | ? AudioEncoderMultiChannelOpusConfig::ApplicationMode::kVoip |
| 252 | : AudioEncoderMultiChannelOpusConfig::ApplicationMode::kAudio; |
| 253 | |
| 254 | config.supported_frame_lengths_ms.clear(); |
| 255 | std::copy(std::begin(kOpusSupportedFrameLengths), |
| 256 | std::end(kOpusSupportedFrameLengths), |
| 257 | std::back_inserter(config.supported_frame_lengths_ms)); |
| 258 | |
| 259 | auto num_streams = GetFormatParameter<int>(format, "num_streams"); |
| 260 | if (!num_streams.has_value()) { |
| 261 | return absl::nullopt; |
| 262 | } |
| 263 | config.num_streams = *num_streams; |
| 264 | |
| 265 | auto coupled_streams = GetFormatParameter<int>(format, "coupled_streams"); |
| 266 | if (!coupled_streams.has_value()) { |
| 267 | return absl::nullopt; |
| 268 | } |
| 269 | config.coupled_streams = *coupled_streams; |
| 270 | |
| 271 | auto channel_mapping = |
| 272 | GetFormatParameter<std::vector<unsigned char>>(format, "channel_mapping"); |
| 273 | if (!channel_mapping.has_value()) { |
| 274 | return absl::nullopt; |
| 275 | } |
| 276 | config.channel_mapping = *channel_mapping; |
| 277 | |
| 278 | return config; |
| 279 | } |
| 280 | |
| 281 | AudioCodecInfo AudioEncoderMultiChannelOpusImpl::QueryAudioEncoder( |
| 282 | const AudioEncoderMultiChannelOpusConfig& config) { |
| 283 | RTC_DCHECK(config.IsOk()); |
| 284 | AudioCodecInfo info(48000, config.num_channels, config.bitrate_bps, |
| 285 | AudioEncoderOpusConfig::kMinBitrateBps, |
| 286 | AudioEncoderOpusConfig::kMaxBitrateBps); |
| 287 | info.allow_comfort_noise = false; |
| 288 | info.supports_network_adaption = false; |
| 289 | return info; |
| 290 | } |
| 291 | |
| 292 | size_t AudioEncoderMultiChannelOpusImpl::Num10msFramesPerPacket() const { |
| 293 | return static_cast<size_t>(rtc::CheckedDivExact(config_.frame_size_ms, 10)); |
| 294 | } |
| 295 | size_t AudioEncoderMultiChannelOpusImpl::SamplesPer10msFrame() const { |
| 296 | return rtc::CheckedDivExact(48000, 100) * config_.num_channels; |
| 297 | } |
| 298 | int AudioEncoderMultiChannelOpusImpl::SampleRateHz() const { |
| 299 | return 48000; |
| 300 | } |
| 301 | size_t AudioEncoderMultiChannelOpusImpl::NumChannels() const { |
| 302 | return config_.num_channels; |
| 303 | } |
| 304 | size_t AudioEncoderMultiChannelOpusImpl::Num10MsFramesInNextPacket() const { |
| 305 | return Num10msFramesPerPacket(); |
| 306 | } |
| 307 | size_t AudioEncoderMultiChannelOpusImpl::Max10MsFramesInAPacket() const { |
| 308 | return Num10msFramesPerPacket(); |
| 309 | } |
| 310 | int AudioEncoderMultiChannelOpusImpl::GetTargetBitrate() const { |
| 311 | return GetBitrateBps(config_); |
| 312 | } |
| 313 | |
| 314 | AudioEncoder::EncodedInfo AudioEncoderMultiChannelOpusImpl::EncodeImpl( |
| 315 | uint32_t rtp_timestamp, |
| 316 | rtc::ArrayView<const int16_t> audio, |
| 317 | rtc::Buffer* encoded) { |
| 318 | if (input_buffer_.empty()) |
| 319 | first_timestamp_in_buffer_ = rtp_timestamp; |
| 320 | |
| 321 | input_buffer_.insert(input_buffer_.end(), audio.cbegin(), audio.cend()); |
| 322 | if (input_buffer_.size() < |
| 323 | (Num10msFramesPerPacket() * SamplesPer10msFrame())) { |
| 324 | return EncodedInfo(); |
| 325 | } |
| 326 | RTC_CHECK_EQ(input_buffer_.size(), |
| 327 | Num10msFramesPerPacket() * SamplesPer10msFrame()); |
| 328 | |
| 329 | const size_t max_encoded_bytes = SufficientOutputBufferSize(); |
| 330 | EncodedInfo info; |
| 331 | info.encoded_bytes = encoded->AppendData( |
| 332 | max_encoded_bytes, [&](rtc::ArrayView<uint8_t> encoded) { |
| 333 | int status = WebRtcOpus_Encode( |
| 334 | inst_, &input_buffer_[0], |
| 335 | rtc::CheckedDivExact(input_buffer_.size(), config_.num_channels), |
| 336 | rtc::saturated_cast<int16_t>(max_encoded_bytes), encoded.data()); |
| 337 | |
| 338 | RTC_CHECK_GE(status, 0); // Fails only if fed invalid data. |
| 339 | |
| 340 | return static_cast<size_t>(status); |
| 341 | }); |
| 342 | input_buffer_.clear(); |
| 343 | |
| 344 | // Will use new packet size for next encoding. |
| 345 | config_.frame_size_ms = next_frame_length_ms_; |
| 346 | |
| 347 | info.encoded_timestamp = first_timestamp_in_buffer_; |
| 348 | info.payload_type = payload_type_; |
| 349 | info.send_even_if_empty = true; // Allows Opus to send empty packets. |
| 350 | |
| 351 | info.speech = true; |
| 352 | info.encoder_type = CodecType::kOther; |
| 353 | |
| 354 | return info; |
| 355 | } |
| 356 | |
| 357 | } // namespace webrtc |