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 <hardware/audio_effect.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 36 namespace android { 37 38 // ---------------------------------------------------------------------------- 39 40 class AudioMixer 41 { 42 public: 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 8 channels for output. 55 static const uint32_t MAX_NUM_CHANNELS = 8; 56 static const uint32_t MAX_NUM_VOLUMES = 2; // stereo volume only 57 // maximum number of channels supported for the content 58 static const uint32_t MAX_NUM_CHANNELS_TO_DOWNMIX = AUDIO_CHANNEL_COUNT_MAX; 59 60 static const uint16_t UNITY_GAIN_INT = 0x1000; 61 static const float UNITY_GAIN_FLOAT = 1.0f; 62 63 enum { // names 64 65 // track names (MAX_NUM_TRACKS units) 66 TRACK0 = 0x1000, 67 68 // 0x2000 is unused 69 70 // setParameter targets 71 TRACK = 0x3000, 72 RESAMPLE = 0x3001, 73 RAMP_VOLUME = 0x3002, // ramp to new volume 74 VOLUME = 0x3003, // don't ramp 75 76 // set Parameter names 77 // for target TRACK 78 CHANNEL_MASK = 0x4000, 79 FORMAT = 0x4001, 80 MAIN_BUFFER = 0x4002, 81 AUX_BUFFER = 0x4003, 82 DOWNMIX_TYPE = 0X4004, 83 MIXER_FORMAT = 0x4005, // AUDIO_FORMAT_PCM_(FLOAT|16_BIT) 84 MIXER_CHANNEL_MASK = 0x4006, // Channel mask for mixer output 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 trackNames()125 uint32_t trackNames() const { return mTrackNames; } 126 127 size_t getUnreleasedFrames(int name) const; 128 isValidPcmTrackFormat(audio_format_t format)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 } 135 136 private: 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_VOLUMES]; // U4.12 fixed point (top bit should be zero) 168 int32_t volumeRL; 169 }; 170 171 int32_t prevVolume[MAX_NUM_VOLUMES]; 172 173 // 16-byte boundary 174 175 int32_t volumeInc[MAX_NUM_VOLUMES]; 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_VOLUMES]; // floating point set volume 221 float mPrevVolume[MAX_NUM_VOLUMES]; // floating point previous volume 222 float mVolumeInc[MAX_NUM_VOLUMES]; // 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 audio_channel_mask_t mMixerChannelMask; 230 uint32_t mMixerChannelCount; 231 needsRamptrack_t232 bool needsRamp() { return (volumeInc[0] | volumeInc[1] | auxInc) != 0; } 233 bool setResampler(uint32_t trackSampleRate, uint32_t devSampleRate); doesResampletrack_t234 bool doesResample() const { return resampler != NULL; } resetResamplertrack_t235 void resetResampler() { if (resampler != NULL) resampler->reset(); } 236 void adjustVolumeRamp(bool aux, bool useFloat = false); getUnreleasedFramestrack_t237 size_t getUnreleasedFrames() const { return resampler != NULL ? 238 resampler->getUnreleasedFrames() : 0; }; 239 }; 240 241 typedef void (*process_hook_t)(state_t* state, int64_t pts); 242 243 // pad to 32-bytes to fill cache line 244 struct state_t { 245 uint32_t enabledTracks; 246 uint32_t needsChanged; 247 size_t frameCount; 248 process_hook_t hook; // one of process__*, never NULL 249 int32_t *outputTemp; 250 int32_t *resampleTemp; 251 NBLog::Writer* mLog; 252 int32_t reserved[1]; 253 // FIXME allocate dynamically to save some memory when maxNumTracks < MAX_NUM_TRACKS 254 track_t tracks[MAX_NUM_TRACKS] __attribute__((aligned(32))); 255 }; 256 257 // Base AudioBufferProvider class used for DownMixerBufferProvider, RemixBufferProvider, 258 // and ReformatBufferProvider. 259 // It handles a private buffer for use in converting format or channel masks from the 260 // input data to a form acceptable by the mixer. 261 // TODO: Make a ResamplerBufferProvider when integers are entirely removed from the 262 // processing pipeline. 263 class CopyBufferProvider : public AudioBufferProvider { 264 public: 265 // Use a private buffer of bufferFrameCount frames (each frame is outputFrameSize bytes). 266 // If bufferFrameCount is 0, no private buffer is created and in-place modification of 267 // the upstream buffer provider's buffers is performed by copyFrames(). 268 CopyBufferProvider(size_t inputFrameSize, size_t outputFrameSize, 269 size_t bufferFrameCount); 270 virtual ~CopyBufferProvider(); 271 272 // Overrides AudioBufferProvider methods 273 virtual status_t getNextBuffer(Buffer* buffer, int64_t pts); 274 virtual void releaseBuffer(Buffer* buffer); 275 276 // Other public methods 277 278 // call this to release the buffer to the upstream provider. 279 // treat it as an audio discontinuity for future samples. 280 virtual void reset(); 281 282 // this function should be supplied by the derived class. It converts 283 // #frames in the *src pointer to the *dst pointer. It is public because 284 // some providers will allow this to work on arbitrary buffers outside 285 // of the internal buffers. 286 virtual void copyFrames(void *dst, const void *src, size_t frames) = 0; 287 288 // set the upstream buffer provider. Consider calling "reset" before this function. setBufferProvider(AudioBufferProvider * p)289 void setBufferProvider(AudioBufferProvider *p) { 290 mTrackBufferProvider = p; 291 } 292 293 protected: 294 AudioBufferProvider* mTrackBufferProvider; 295 const size_t mInputFrameSize; 296 const size_t mOutputFrameSize; 297 private: 298 AudioBufferProvider::Buffer mBuffer; 299 const size_t mLocalBufferFrameCount; 300 void* mLocalBufferData; 301 size_t mConsumed; 302 }; 303 304 // DownmixerBufferProvider wraps a track AudioBufferProvider to provide 305 // position dependent downmixing by an Audio Effect. 306 class DownmixerBufferProvider : public CopyBufferProvider { 307 public: 308 DownmixerBufferProvider(audio_channel_mask_t inputChannelMask, 309 audio_channel_mask_t outputChannelMask, audio_format_t format, 310 uint32_t sampleRate, int32_t sessionId, size_t bufferFrameCount); 311 virtual ~DownmixerBufferProvider(); 312 virtual void copyFrames(void *dst, const void *src, size_t frames); isValid()313 bool isValid() const { return mDownmixHandle != NULL; } 314 315 static status_t init(); isMultichannelCapable()316 static bool isMultichannelCapable() { return sIsMultichannelCapable; } 317 318 protected: 319 effect_handle_t mDownmixHandle; 320 effect_config_t mDownmixConfig; 321 322 // effect descriptor for the downmixer used by the mixer 323 static effect_descriptor_t sDwnmFxDesc; 324 // indicates whether a downmix effect has been found and is usable by this mixer 325 static bool sIsMultichannelCapable; 326 // FIXME: should we allow effects outside of the framework? 327 // We need to here. A special ioId that must be <= -2 so it does not map to a session. 328 static const int32_t SESSION_ID_INVALID_AND_IGNORED = -2; 329 }; 330 331 // RemixBufferProvider wraps a track AudioBufferProvider to perform an 332 // upmix or downmix to the proper channel count and mask. 333 class RemixBufferProvider : public CopyBufferProvider { 334 public: 335 RemixBufferProvider(audio_channel_mask_t inputChannelMask, 336 audio_channel_mask_t outputChannelMask, audio_format_t format, 337 size_t bufferFrameCount); 338 virtual void copyFrames(void *dst, const void *src, size_t frames); 339 340 protected: 341 const audio_format_t mFormat; 342 const size_t mSampleSize; 343 const size_t mInputChannels; 344 const size_t mOutputChannels; 345 int8_t mIdxAry[sizeof(uint32_t)*8]; // 32 bits => channel indices 346 }; 347 348 // ReformatBufferProvider wraps a track AudioBufferProvider to convert the input data 349 // to an acceptable mixer input format type. 350 class ReformatBufferProvider : public CopyBufferProvider { 351 public: 352 ReformatBufferProvider(int32_t channels, 353 audio_format_t inputFormat, audio_format_t outputFormat, 354 size_t bufferFrameCount); 355 virtual void copyFrames(void *dst, const void *src, size_t frames); 356 357 protected: 358 const int32_t mChannels; 359 const audio_format_t mInputFormat; 360 const audio_format_t mOutputFormat; 361 }; 362 363 // bitmask of allocated track names, where bit 0 corresponds to TRACK0 etc. 364 uint32_t mTrackNames; 365 366 // bitmask of configured track names; ~0 if maxNumTracks == MAX_NUM_TRACKS, 367 // but will have fewer bits set if maxNumTracks < MAX_NUM_TRACKS 368 const uint32_t mConfiguredNames; 369 370 const uint32_t mSampleRate; 371 372 NBLog::Writer mDummyLog; 373 public: 374 void setLog(NBLog::Writer* log); 375 private: 376 state_t mState __attribute__((aligned(32))); 377 378 // Call after changing either the enabled status of a track, or parameters of an enabled track. 379 // OK to call more often than that, but unnecessary. 380 void invalidateState(uint32_t mask); 381 382 bool setChannelMasks(int name, 383 audio_channel_mask_t trackChannelMask, audio_channel_mask_t mixerChannelMask); 384 385 // TODO: remove unused trackName/trackNum from functions below. 386 static status_t initTrackDownmix(track_t* pTrack, int trackName); 387 static status_t prepareTrackForDownmix(track_t* pTrack, int trackNum); 388 static void unprepareTrackForDownmix(track_t* pTrack, int trackName); 389 static status_t prepareTrackForReformat(track_t* pTrack, int trackNum); 390 static void unprepareTrackForReformat(track_t* pTrack, int trackName); 391 static void reconfigureBufferProviders(track_t* pTrack); 392 393 static void track__genericResample(track_t* t, int32_t* out, size_t numFrames, int32_t* temp, 394 int32_t* aux); 395 static void track__nop(track_t* t, int32_t* out, size_t numFrames, int32_t* temp, int32_t* aux); 396 static void track__16BitsStereo(track_t* t, int32_t* out, size_t numFrames, int32_t* temp, 397 int32_t* aux); 398 static void track__16BitsMono(track_t* t, int32_t* out, size_t numFrames, int32_t* temp, 399 int32_t* aux); 400 static void volumeRampStereo(track_t* t, int32_t* out, size_t frameCount, int32_t* temp, 401 int32_t* aux); 402 static void volumeStereo(track_t* t, int32_t* out, size_t frameCount, int32_t* temp, 403 int32_t* aux); 404 405 static void process__validate(state_t* state, int64_t pts); 406 static void process__nop(state_t* state, int64_t pts); 407 static void process__genericNoResampling(state_t* state, int64_t pts); 408 static void process__genericResampling(state_t* state, int64_t pts); 409 static void process__OneTrack16BitsStereoNoResampling(state_t* state, 410 int64_t pts); 411 412 static int64_t calculateOutputPTS(const track_t& t, int64_t basePTS, 413 int outputFrameIndex); 414 415 static uint64_t sLocalTimeFreq; 416 static pthread_once_t sOnceControl; 417 static void sInitRoutine(); 418 419 /* multi-format volume mixing function (calls template functions 420 * in AudioMixerOps.h). The template parameters are as follows: 421 * 422 * MIXTYPE (see AudioMixerOps.h MIXTYPE_* enumeration) 423 * USEFLOATVOL (set to true if float volume is used) 424 * ADJUSTVOL (set to true if volume ramp parameters needs adjustment afterwards) 425 * TO: int32_t (Q4.27) or float 426 * TI: int32_t (Q4.27) or int16_t (Q0.15) or float 427 * TA: int32_t (Q4.27) 428 */ 429 template <int MIXTYPE, bool USEFLOATVOL, bool ADJUSTVOL, 430 typename TO, typename TI, typename TA> 431 static void volumeMix(TO *out, size_t outFrames, 432 const TI *in, TA *aux, bool ramp, AudioMixer::track_t *t); 433 434 // multi-format process hooks 435 template <int MIXTYPE, typename TO, typename TI, typename TA> 436 static void process_NoResampleOneTrack(state_t* state, int64_t pts); 437 438 // multi-format track hooks 439 template <int MIXTYPE, typename TO, typename TI, typename TA> 440 static void track__Resample(track_t* t, TO* out, size_t frameCount, 441 TO* temp __unused, TA* aux); 442 template <int MIXTYPE, typename TO, typename TI, typename TA> 443 static void track__NoResample(track_t* t, TO* out, size_t frameCount, 444 TO* temp __unused, TA* aux); 445 446 static void convertMixerFormat(void *out, audio_format_t mixerOutFormat, 447 void *in, audio_format_t mixerInFormat, size_t sampleCount); 448 449 // hook types 450 enum { 451 PROCESSTYPE_NORESAMPLEONETRACK, 452 }; 453 enum { 454 TRACKTYPE_NOP, 455 TRACKTYPE_RESAMPLE, 456 TRACKTYPE_NORESAMPLE, 457 TRACKTYPE_NORESAMPLEMONO, 458 }; 459 460 // functions for determining the proper process and track hooks. 461 static process_hook_t getProcessHook(int processType, uint32_t channelCount, 462 audio_format_t mixerInFormat, audio_format_t mixerOutFormat); 463 static hook_t getTrackHook(int trackType, uint32_t channelCount, 464 audio_format_t mixerInFormat, audio_format_t mixerOutFormat); 465 }; 466 467 // ---------------------------------------------------------------------------- 468 }; // namespace android 469 470 #endif // ANDROID_AUDIO_MIXER_H 471