AudioMixer.h revision a08810b2feafeec88870c7c1f01efc39ee8e0d78
1/* 2** 3** Copyright 2007, The Android Open Source Project 4** 5** Licensed under the Apache License, Version 2.0 (the "License"); 6** you may not use this file except in compliance with the License. 7** You may obtain a copy of the License at 8** 9** http://www.apache.org/licenses/LICENSE-2.0 10** 11** Unless required by applicable law or agreed to in writing, software 12** distributed under the License is distributed on an "AS IS" BASIS, 13** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14** See the License for the specific language governing permissions and 15** limitations under the License. 16*/ 17 18#ifndef ANDROID_AUDIO_MIXER_H 19#define ANDROID_AUDIO_MIXER_H 20 21#include <stdint.h> 22#include <sys/types.h> 23 24#include <utils/threads.h> 25 26#include <media/AudioBufferProvider.h> 27#include "AudioResampler.h" 28 29#include <audio_effects/effect_downmix.h> 30#include <system/audio.h> 31#include <media/nbaio/NBLog.h> 32 33// FIXME This is actually unity gain, which might not be max in future, expressed in U.12 34#define MAX_GAIN_INT AudioMixer::UNITY_GAIN_INT 35 36namespace android { 37 38// ---------------------------------------------------------------------------- 39 40class AudioMixer 41{ 42public: 43 AudioMixer(size_t frameCount, uint32_t sampleRate, 44 uint32_t maxNumTracks = MAX_NUM_TRACKS); 45 46 /*virtual*/ ~AudioMixer(); // non-virtual saves a v-table, restore if sub-classed 47 48 49 // This mixer has a hard-coded upper limit of 32 active track inputs. 50 // Adding support for > 32 tracks would require more than simply changing this value. 51 static const uint32_t MAX_NUM_TRACKS = 32; 52 // maximum number of channels supported by the mixer 53 54 // This mixer has a hard-coded upper limit of 2 channels for output. 55 // There is support for > 2 channel tracks down-mixed to 2 channel output via a down-mix effect. 56 // Adding support for > 2 channel output would require more than simply changing this value. 57 static const uint32_t MAX_NUM_CHANNELS = 2; 58 // maximum number of channels supported for the content 59 static const uint32_t MAX_NUM_CHANNELS_TO_DOWNMIX = 8; 60 61 static const uint16_t UNITY_GAIN_INT = 0x1000; 62 static const float UNITY_GAIN_FLOAT = 1.0f; 63 64 enum { // names 65 66 // track names (MAX_NUM_TRACKS units) 67 TRACK0 = 0x1000, 68 69 // 0x2000 is unused 70 71 // setParameter targets 72 TRACK = 0x3000, 73 RESAMPLE = 0x3001, 74 RAMP_VOLUME = 0x3002, // ramp to new volume 75 VOLUME = 0x3003, // don't ramp 76 77 // set Parameter names 78 // for target TRACK 79 CHANNEL_MASK = 0x4000, 80 FORMAT = 0x4001, 81 MAIN_BUFFER = 0x4002, 82 AUX_BUFFER = 0x4003, 83 DOWNMIX_TYPE = 0X4004, 84 MIXER_FORMAT = 0x4005, // AUDIO_FORMAT_PCM_(FLOAT|16_BIT) 85 // for target RESAMPLE 86 SAMPLE_RATE = 0x4100, // Configure sample rate conversion on this track name; 87 // parameter 'value' is the new sample rate in Hz. 88 // Only creates a sample rate converter the first time that 89 // the track sample rate is different from the mix sample rate. 90 // If the new sample rate is the same as the mix sample rate, 91 // and a sample rate converter already exists, 92 // then the sample rate converter remains present but is a no-op. 93 RESET = 0x4101, // Reset sample rate converter without changing sample rate. 94 // This clears out the resampler's input buffer. 95 REMOVE = 0x4102, // Remove the sample rate converter on this track name; 96 // the track is restored to the mix sample rate. 97 // for target RAMP_VOLUME and VOLUME (8 channels max) 98 // FIXME use float for these 3 to improve the dynamic range 99 VOLUME0 = 0x4200, 100 VOLUME1 = 0x4201, 101 AUXLEVEL = 0x4210, 102 }; 103 104 105 // For all APIs with "name": TRACK0 <= name < TRACK0 + MAX_NUM_TRACKS 106 107 // Allocate a track name. Returns new track name if successful, -1 on failure. 108 // The failure could be because of an invalid channelMask or format, or that 109 // the track capacity of the mixer is exceeded. 110 int getTrackName(audio_channel_mask_t channelMask, 111 audio_format_t format, int sessionId); 112 113 // Free an allocated track by name 114 void deleteTrackName(int name); 115 116 // Enable or disable an allocated track by name 117 void enable(int name); 118 void disable(int name); 119 120 void setParameter(int name, int target, int param, void *value); 121 122 void setBufferProvider(int name, AudioBufferProvider* bufferProvider); 123 void process(int64_t pts); 124 125 uint32_t trackNames() const { return mTrackNames; } 126 127 size_t getUnreleasedFrames(int name) const; 128 129 static inline bool isValidPcmTrackFormat(audio_format_t format) { 130 return format == AUDIO_FORMAT_PCM_16_BIT || 131 format == AUDIO_FORMAT_PCM_24_BIT_PACKED || 132 format == AUDIO_FORMAT_PCM_32_BIT || 133 format == AUDIO_FORMAT_PCM_FLOAT; 134 } 135 136private: 137 138 enum { 139 // FIXME this representation permits up to 8 channels 140 NEEDS_CHANNEL_COUNT__MASK = 0x00000007, 141 }; 142 143 enum { 144 NEEDS_CHANNEL_1 = 0x00000000, // mono 145 NEEDS_CHANNEL_2 = 0x00000001, // stereo 146 147 // sample format is not explicitly specified, and is assumed to be AUDIO_FORMAT_PCM_16_BIT 148 149 NEEDS_MUTE = 0x00000100, 150 NEEDS_RESAMPLE = 0x00001000, 151 NEEDS_AUX = 0x00010000, 152 }; 153 154 struct state_t; 155 struct track_t; 156 class CopyBufferProvider; 157 158 typedef void (*hook_t)(track_t* t, int32_t* output, size_t numOutFrames, int32_t* temp, 159 int32_t* aux); 160 static const int BLOCKSIZE = 16; // 4 cache lines 161 162 struct track_t { 163 uint32_t needs; 164 165 // TODO: Eventually remove legacy integer volume settings 166 union { 167 int16_t volume[MAX_NUM_CHANNELS]; // U4.12 fixed point (top bit should be zero) 168 int32_t volumeRL; 169 }; 170 171 int32_t prevVolume[MAX_NUM_CHANNELS]; 172 173 // 16-byte boundary 174 175 int32_t volumeInc[MAX_NUM_CHANNELS]; 176 int32_t auxInc; 177 int32_t prevAuxLevel; 178 179 // 16-byte boundary 180 181 int16_t auxLevel; // 0 <= auxLevel <= MAX_GAIN_INT, but signed for mul performance 182 uint16_t frameCount; 183 184 uint8_t channelCount; // 1 or 2, redundant with (needs & NEEDS_CHANNEL_COUNT__MASK) 185 uint8_t unused_padding; // formerly format, was always 16 186 uint16_t enabled; // actually bool 187 audio_channel_mask_t channelMask; 188 189 // actual buffer provider used by the track hooks, see DownmixerBufferProvider below 190 // for how the Track buffer provider is wrapped by another one when dowmixing is required 191 AudioBufferProvider* bufferProvider; 192 193 // 16-byte boundary 194 195 mutable AudioBufferProvider::Buffer buffer; // 8 bytes 196 197 hook_t hook; 198 const void* in; // current location in buffer 199 200 // 16-byte boundary 201 202 AudioResampler* resampler; 203 uint32_t sampleRate; 204 int32_t* mainBuffer; 205 int32_t* auxBuffer; 206 207 // 16-byte boundary 208 AudioBufferProvider* mInputBufferProvider; // externally provided buffer provider. 209 CopyBufferProvider* mReformatBufferProvider; // provider wrapper for reformatting. 210 CopyBufferProvider* downmixerBufferProvider; // wrapper for channel conversion. 211 212 int32_t sessionId; 213 214 // 16-byte boundary 215 audio_format_t mMixerFormat; // output mix format: AUDIO_FORMAT_PCM_(FLOAT|16_BIT) 216 audio_format_t mFormat; // input track format 217 audio_format_t mMixerInFormat; // mix internal format AUDIO_FORMAT_PCM_(FLOAT|16_BIT) 218 // each track must be converted to this format. 219 220 float mVolume[MAX_NUM_CHANNELS]; // floating point set volume 221 float mPrevVolume[MAX_NUM_CHANNELS]; // floating point previous volume 222 float mVolumeInc[MAX_NUM_CHANNELS]; // floating point volume increment 223 224 float mAuxLevel; // floating point set aux level 225 float mPrevAuxLevel; // floating point prev aux level 226 float mAuxInc; // floating point aux increment 227 228 // 16-byte boundary 229 230 bool needsRamp() { return (volumeInc[0] | volumeInc[1] | auxInc) != 0; } 231 bool setResampler(uint32_t sampleRate, uint32_t devSampleRate); 232 bool doesResample() const { return resampler != NULL; } 233 void resetResampler() { if (resampler != NULL) resampler->reset(); } 234 void adjustVolumeRamp(bool aux, bool useFloat = false); 235 size_t getUnreleasedFrames() const { return resampler != NULL ? 236 resampler->getUnreleasedFrames() : 0; }; 237 }; 238 239 typedef void (*process_hook_t)(state_t* state, int64_t pts); 240 241 // pad to 32-bytes to fill cache line 242 struct state_t { 243 uint32_t enabledTracks; 244 uint32_t needsChanged; 245 size_t frameCount; 246 process_hook_t hook; // one of process__*, never NULL 247 int32_t *outputTemp; 248 int32_t *resampleTemp; 249 NBLog::Writer* mLog; 250 int32_t reserved[1]; 251 // FIXME allocate dynamically to save some memory when maxNumTracks < MAX_NUM_TRACKS 252 track_t tracks[MAX_NUM_TRACKS] __attribute__((aligned(32))); 253 }; 254 255 // Base AudioBufferProvider class used for DownMixerBufferProvider, RemixBufferProvider, 256 // and ReformatBufferProvider. 257 // It handles a private buffer for use in converting format or channel masks from the 258 // input data to a form acceptable by the mixer. 259 // TODO: Make a ResamplerBufferProvider when integers are entirely removed from the 260 // processing pipeline. 261 class CopyBufferProvider : public AudioBufferProvider { 262 public: 263 // Use a private buffer of bufferFrameCount frames (each frame is outputFrameSize bytes). 264 // If bufferFrameCount is 0, no private buffer is created and in-place modification of 265 // the upstream buffer provider's buffers is performed by copyFrames(). 266 CopyBufferProvider(size_t inputFrameSize, size_t outputFrameSize, 267 size_t bufferFrameCount); 268 virtual ~CopyBufferProvider(); 269 270 // Overrides AudioBufferProvider methods 271 virtual status_t getNextBuffer(Buffer* buffer, int64_t pts); 272 virtual void releaseBuffer(Buffer* buffer); 273 274 // Other public methods 275 276 // call this to release the buffer to the upstream provider. 277 // treat it as an audio discontinuity for future samples. 278 virtual void reset(); 279 280 // this function should be supplied by the derived class. It converts 281 // #frames in the *src pointer to the *dst pointer. It is public because 282 // some providers will allow this to work on arbitrary buffers outside 283 // of the internal buffers. 284 virtual void copyFrames(void *dst, const void *src, size_t frames) = 0; 285 286 // set the upstream buffer provider. Consider calling "reset" before this function. 287 void setBufferProvider(AudioBufferProvider *p) { 288 mTrackBufferProvider = p; 289 } 290 291 protected: 292 AudioBufferProvider* mTrackBufferProvider; 293 const size_t mInputFrameSize; 294 const size_t mOutputFrameSize; 295 private: 296 AudioBufferProvider::Buffer mBuffer; 297 const size_t mLocalBufferFrameCount; 298 void* mLocalBufferData; 299 size_t mConsumed; 300 }; 301 302 // DownmixerBufferProvider wraps a track AudioBufferProvider to provide 303 // position dependent downmixing by an Audio Effect. 304 class DownmixerBufferProvider : public CopyBufferProvider { 305 public: 306 DownmixerBufferProvider(audio_channel_mask_t inputChannelMask, 307 audio_channel_mask_t outputChannelMask, audio_format_t format, 308 uint32_t sampleRate, int32_t sessionId, size_t bufferFrameCount); 309 virtual ~DownmixerBufferProvider(); 310 virtual void copyFrames(void *dst, const void *src, size_t frames); 311 bool isValid() const { return mDownmixHandle != NULL; } 312 313 static status_t init(); 314 static bool isMultichannelCapable() { return sIsMultichannelCapable; } 315 316 protected: 317 effect_handle_t mDownmixHandle; 318 effect_config_t mDownmixConfig; 319 320 // effect descriptor for the downmixer used by the mixer 321 static effect_descriptor_t sDwnmFxDesc; 322 // indicates whether a downmix effect has been found and is usable by this mixer 323 static bool sIsMultichannelCapable; 324 // FIXME: should we allow effects outside of the framework? 325 // We need to here. A special ioId that must be <= -2 so it does not map to a session. 326 static const int32_t SESSION_ID_INVALID_AND_IGNORED = -2; 327 }; 328 329 // RemixBufferProvider wraps a track AudioBufferProvider to perform an 330 // upmix or downmix to the proper channel count and mask. 331 class RemixBufferProvider : public CopyBufferProvider { 332 public: 333 RemixBufferProvider(audio_channel_mask_t inputChannelMask, 334 audio_channel_mask_t outputChannelMask, audio_format_t format, 335 size_t bufferFrameCount); 336 virtual void copyFrames(void *dst, const void *src, size_t frames); 337 338 protected: 339 const audio_format_t mFormat; 340 const size_t mSampleSize; 341 const size_t mInputChannels; 342 const size_t mOutputChannels; 343 int8_t mIdxAry[sizeof(uint32_t)*8]; // 32 bits => channel indices 344 }; 345 346 // ReformatBufferProvider wraps a track AudioBufferProvider to convert the input data 347 // to an acceptable mixer input format type. 348 class ReformatBufferProvider : public CopyBufferProvider { 349 public: 350 ReformatBufferProvider(int32_t channels, 351 audio_format_t inputFormat, audio_format_t outputFormat, 352 size_t bufferFrameCount); 353 virtual void copyFrames(void *dst, const void *src, size_t frames); 354 355 protected: 356 const int32_t mChannels; 357 const audio_format_t mInputFormat; 358 const audio_format_t mOutputFormat; 359 }; 360 361 // bitmask of allocated track names, where bit 0 corresponds to TRACK0 etc. 362 uint32_t mTrackNames; 363 364 // bitmask of configured track names; ~0 if maxNumTracks == MAX_NUM_TRACKS, 365 // but will have fewer bits set if maxNumTracks < MAX_NUM_TRACKS 366 const uint32_t mConfiguredNames; 367 368 const uint32_t mSampleRate; 369 370 NBLog::Writer mDummyLog; 371public: 372 void setLog(NBLog::Writer* log); 373private: 374 state_t mState __attribute__((aligned(32))); 375 376 // Call after changing either the enabled status of a track, or parameters of an enabled track. 377 // OK to call more often than that, but unnecessary. 378 void invalidateState(uint32_t mask); 379 380 static status_t initTrackDownmix(track_t* pTrack, int trackNum, audio_channel_mask_t mask); 381 static status_t prepareTrackForDownmix(track_t* pTrack, int trackNum); 382 static void unprepareTrackForDownmix(track_t* pTrack, int trackName); 383 static status_t prepareTrackForReformat(track_t* pTrack, int trackNum); 384 static void unprepareTrackForReformat(track_t* pTrack, int trackName); 385 static void reconfigureBufferProviders(track_t* pTrack); 386 387 static void track__genericResample(track_t* t, int32_t* out, size_t numFrames, int32_t* temp, 388 int32_t* aux); 389 static void track__nop(track_t* t, int32_t* out, size_t numFrames, int32_t* temp, int32_t* aux); 390 static void track__16BitsStereo(track_t* t, int32_t* out, size_t numFrames, int32_t* temp, 391 int32_t* aux); 392 static void track__16BitsMono(track_t* t, int32_t* out, size_t numFrames, int32_t* temp, 393 int32_t* aux); 394 static void volumeRampStereo(track_t* t, int32_t* out, size_t frameCount, int32_t* temp, 395 int32_t* aux); 396 static void volumeStereo(track_t* t, int32_t* out, size_t frameCount, int32_t* temp, 397 int32_t* aux); 398 399 static void process__validate(state_t* state, int64_t pts); 400 static void process__nop(state_t* state, int64_t pts); 401 static void process__genericNoResampling(state_t* state, int64_t pts); 402 static void process__genericResampling(state_t* state, int64_t pts); 403 static void process__OneTrack16BitsStereoNoResampling(state_t* state, 404 int64_t pts); 405#if 0 406 static void process__TwoTracks16BitsStereoNoResampling(state_t* state, 407 int64_t pts); 408#endif 409 410 static int64_t calculateOutputPTS(const track_t& t, int64_t basePTS, 411 int outputFrameIndex); 412 413 static uint64_t sLocalTimeFreq; 414 static pthread_once_t sOnceControl; 415 static void sInitRoutine(); 416 417 /* multi-format volume mixing function (calls template functions 418 * in AudioMixerOps.h). The template parameters are as follows: 419 * 420 * MIXTYPE (see AudioMixerOps.h MIXTYPE_* enumeration) 421 * NCHAN (number of channels, 2 for now) 422 * USEFLOATVOL (set to true if float volume is used) 423 * ADJUSTVOL (set to true if volume ramp parameters needs adjustment afterwards) 424 * TO: int32_t (Q4.27) or float 425 * TI: int32_t (Q4.27) or int16_t (Q0.15) or float 426 * TA: int32_t (Q4.27) 427 */ 428 template <int MIXTYPE, int NCHAN, bool USEFLOATVOL, bool ADJUSTVOL, 429 typename TO, typename TI, typename TA> 430 static void volumeMix(TO *out, size_t outFrames, 431 const TI *in, TA *aux, bool ramp, AudioMixer::track_t *t); 432 433 // multi-format process hooks 434 template <int MIXTYPE, int NCHAN, typename TO, typename TI, typename TA> 435 static void process_NoResampleOneTrack(state_t* state, int64_t pts); 436 437 // multi-format track hooks 438 template <int MIXTYPE, int NCHAN, typename TO, typename TI, typename TA> 439 static void track__Resample(track_t* t, TO* out, size_t frameCount, 440 TO* temp __unused, TA* aux); 441 template <int MIXTYPE, int NCHAN, typename TO, typename TI, typename TA> 442 static void track__NoResample(track_t* t, TO* out, size_t frameCount, 443 TO* temp __unused, TA* aux); 444 445 static void convertMixerFormat(void *out, audio_format_t mixerOutFormat, 446 void *in, audio_format_t mixerInFormat, size_t sampleCount); 447 448 // hook types 449 enum { 450 PROCESSTYPE_NORESAMPLEONETRACK, 451 }; 452 enum { 453 TRACKTYPE_NOP, 454 TRACKTYPE_RESAMPLE, 455 TRACKTYPE_NORESAMPLE, 456 TRACKTYPE_NORESAMPLEMONO, 457 }; 458 459 // functions for determining the proper process and track hooks. 460 static process_hook_t getProcessHook(int processType, int channels, 461 audio_format_t mixerInFormat, audio_format_t mixerOutFormat); 462 static hook_t getTrackHook(int trackType, int channels, 463 audio_format_t mixerInFormat, audio_format_t mixerOutFormat); 464}; 465 466// ---------------------------------------------------------------------------- 467}; // namespace android 468 469#endif // ANDROID_AUDIO_MIXER_H 470