Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit a007cc53 authored by Andy Hung's avatar Andy Hung Committed by Automerger Merge Worker
Browse files

Merge "libaudiopreprocessing: Cleanup PreProcessing.cpp" am: 85f5f276 am:... 

Merge "libaudiopreprocessing: Cleanup PreProcessing.cpp" am: 85f5f276 am: 7370827c am: 122e9ccf

Original change: https://android-review.googlesource.com/c/platform/frameworks/av/+/1641153

Change-Id: I5261205a2cb2ee72b0de80b84591742b5b9cc70f
parents fca42b22 122e9ccf

media/libeffects/preprocessing/PreProcessing.cpp

+28 −128
Original line number Diff line number Diff line
@@ -105,9 +105,8 @@ struct preproc_session_s { 
    webrtc::AudioProcessing* apm;  // handle on webRTC audio processing module (APM)

    // Audio Processing module builder

    webrtc::AudioProcessingBuilder ap_builder;

    size_t apmFrameCount;      // buffer size for webRTC process (10 ms)

    uint32_t apmSamplingRate;  // webRTC APM sampling rate (8/16 or 32 kHz)

    size_t frameCount;         // buffer size before input resampler ( <=> apmFrameCount)

    // frameCount represents the size of the buffers used for processing, and must represent 10ms.

    size_t frameCount;

    uint32_t samplingRate;     // sampling rate at effect process interface

    uint32_t inChannelCount;   // input channel count

    uint32_t outChannelCount;  // output channel count
@@ -119,21 +118,12 @@ struct preproc_session_s { 
    webrtc::AudioProcessing::Config config;

    webrtc::StreamConfig inputConfig;   // input stream configuration

    webrtc::StreamConfig outputConfig;  // output stream configuration

    int16_t* inBuf;    // input buffer used when resampling

    size_t inBufSize;  // input buffer size in frames

    size_t framesIn;   // number of frames in input buffer

    int16_t* outBuf;    // output buffer used when resampling

    size_t outBufSize;  // output buffer size in frames

    size_t framesOut;   // number of frames in output buffer

    uint32_t revChannelCount;  // number of channels on reverse stream

    uint32_t revEnabledMsk;    // bit field containing IDs of enabled pre processors

                               // with reverse channel

    uint32_t revProcessedMsk;  // bit field containing IDs of pre processors with reverse

                               // channel already processed in current round

    webrtc::StreamConfig revConfig;     // reverse stream configuration.

    int16_t* revBuf;    // reverse channel input buffer

    size_t revBufSize;  // reverse channel input buffer size

    size_t framesRev;   // number of frames in reverse channel input buffer

};



#ifdef DUAL_MIC_TEST
@@ -862,9 +852,7 @@ extern "C" int Session_CreateEffect(preproc_session_t* session, int32_t procId, 
            ALOGW("Session_CreateEffect could not get apm engine");

            goto error;

        }

        session->apmSamplingRate = kPreprocDefaultSr;

        session->apmFrameCount = (kPreprocDefaultSr) / 100;

        session->frameCount = session->apmFrameCount;

        session->frameCount = kPreprocDefaultSr / 100;

        session->samplingRate = kPreprocDefaultSr;

        session->inChannelCount = kPreProcDefaultCnl;

        session->outChannelCount = kPreProcDefaultCnl;
@@ -879,12 +867,6 @@ extern "C" int Session_CreateEffect(preproc_session_t* session, int32_t procId, 
        session->processedMsk = 0;

        session->revEnabledMsk = 0;

        session->revProcessedMsk = 0;

        session->inBuf = NULL;

        session->inBufSize = 0;

        session->outBuf = NULL;

        session->outBufSize = 0;

        session->revBuf = NULL;

        session->revBufSize = 0;

    }

    status = Effect_Create(&session->effects[procId], session, interface);

    if (status < 0) {
@@ -908,13 +890,6 @@ int Session_ReleaseEffect(preproc_session_t* session, preproc_effect_t* fx) { 
    if (session->createdMsk == 0) {

        delete session->apm;

        session->apm = NULL;

        delete session->inBuf;

        session->inBuf = NULL;

        free(session->outBuf);

        session->outBuf = NULL;

        delete session->revBuf;

        session->revBuf = NULL;



        session->id = 0;

    }
@@ -934,24 +909,8 @@ int Session_SetConfig(preproc_session_t* session, effect_config_t* config) { 
    ALOGV("Session_SetConfig sr %d cnl %08x", config->inputCfg.samplingRate,

          config->inputCfg.channels);



    // AEC implementation is limited to 16kHz

    if (config->inputCfg.samplingRate >= 32000 && !(session->createdMsk & (1 << PREPROC_AEC))) {

        session->apmSamplingRate = 32000;

    } else if (config->inputCfg.samplingRate >= 16000) {

        session->apmSamplingRate = 16000;

    } else if (config->inputCfg.samplingRate >= 8000) {

        session->apmSamplingRate = 8000;

    }





    session->samplingRate = config->inputCfg.samplingRate;

    session->apmFrameCount = session->apmSamplingRate / 100;

    if (session->samplingRate == session->apmSamplingRate) {

        session->frameCount = session->apmFrameCount;

    } else {

        session->frameCount =

                (session->apmFrameCount * session->samplingRate) / session->apmSamplingRate;

    }

    session->frameCount = session->samplingRate / 100;

    session->inChannelCount = inCnl;

    session->outChannelCount = outCnl;

    session->inputConfig.set_sample_rate_hz(session->samplingRate);
@@ -963,13 +922,6 @@ int Session_SetConfig(preproc_session_t* session, effect_config_t* config) { 
    session->revConfig.set_sample_rate_hz(session->samplingRate);

    session->revConfig.set_num_channels(inCnl);



    // force process buffer reallocation

    session->inBufSize = 0;

    session->outBufSize = 0;

    session->framesIn = 0;

    session->framesOut = 0;





    session->state = PREPROC_SESSION_STATE_CONFIG;

    return 0;

}
@@ -1004,9 +956,6 @@ int Session_SetReverseConfig(preproc_session_t* session, effect_config_t* config 
    }

    uint32_t inCnl = audio_channel_count_from_out_mask(config->inputCfg.channels);

    session->revChannelCount = inCnl;

    // force process buffer reallocation

    session->revBufSize = 0;

    session->framesRev = 0;



    return 0;

}
@@ -1023,12 +972,8 @@ void Session_GetReverseConfig(preproc_session_t* session, effect_config_t* confi 


void Session_SetProcEnabled(preproc_session_t* session, uint32_t procId, bool enabled) {

    if (enabled) {

        if (session->enabledMsk == 0) {

            session->framesIn = 0;

        }

        session->enabledMsk |= (1 << procId);

        if (HasReverseStream(procId)) {

            session->framesRev = 0;

            session->revEnabledMsk |= (1 << procId);

        }

    } else {
@@ -1117,43 +1062,24 @@ int PreProcessingFx_Process(effect_handle_t self, audio_buffer_t* inBuffer, 
        return -EINVAL;

    }



    if (inBuffer->frameCount != outBuffer->frameCount) {

        ALOGW("inBuffer->frameCount %zu is not equal to outBuffer->frameCount %zu",

              inBuffer->frameCount, outBuffer->frameCount);

        return -EINVAL;

    }



    if (inBuffer->frameCount != session->frameCount) {

        ALOGW("inBuffer->frameCount %zu != %zu representing 10ms at sampling rate %d",

              inBuffer->frameCount, session->frameCount, session->samplingRate);

        return -EINVAL;

    }



    session->processedMsk |= (1 << effect->procId);



    //    ALOGV("PreProcessingFx_Process In %d frames enabledMsk %08x processedMsk %08x",

    //         inBuffer->frameCount, session->enabledMsk, session->processedMsk);



    if ((session->processedMsk & session->enabledMsk) == session->enabledMsk) {

        effect->session->processedMsk = 0;

        size_t framesRq = outBuffer->frameCount;

        size_t framesWr = 0;

        if (session->framesOut) {

            size_t fr = session->framesOut;

            if (outBuffer->frameCount < fr) {

                fr = outBuffer->frameCount;

            }

            memcpy(outBuffer->s16, session->outBuf,

                   fr * session->outChannelCount * sizeof(int16_t));

            memmove(session->outBuf, session->outBuf + fr * session->outChannelCount,

                    (session->framesOut - fr) * session->outChannelCount * sizeof(int16_t));

            session->framesOut -= fr;

            framesWr += fr;

        }

        outBuffer->frameCount = framesWr;

        if (framesWr == framesRq) {

            inBuffer->frameCount = 0;

            return 0;

        }



        size_t fr = session->frameCount - session->framesIn;

        if (inBuffer->frameCount < fr) {

            fr = inBuffer->frameCount;

        }

        session->framesIn += fr;

        inBuffer->frameCount = fr;

        if (session->framesIn < session->frameCount) {

            return 0;

        }

        session->framesIn = 0;

        if (int status = effect->session->apm->ProcessStream(

                    (const int16_t* const)inBuffer->s16,

                    (const webrtc::StreamConfig)effect->session->inputConfig,
@@ -1163,34 +1089,6 @@ int PreProcessingFx_Process(effect_handle_t self, audio_buffer_t* inBuffer, 
            ALOGE("Process Stream failed with error %d\n", status);

            return status;

        }

        outBuffer->frameCount = inBuffer->frameCount;



        if (session->outBufSize < session->framesOut + session->frameCount) {

            int16_t* buf;

            session->outBufSize = session->framesOut + session->frameCount;

            buf = (int16_t*)realloc(

                    session->outBuf,

                    session->outBufSize * session->outChannelCount * sizeof(int16_t));

            if (buf == NULL) {

                session->framesOut = 0;

                free(session->outBuf);

                session->outBuf = NULL;

                return -ENOMEM;

            }

            session->outBuf = buf;

        }



        fr = session->framesOut;

        if (framesRq - framesWr < fr) {

            fr = framesRq - framesWr;

        }

        memcpy(outBuffer->s16 + framesWr * session->outChannelCount, session->outBuf,

               fr * session->outChannelCount * sizeof(int16_t));

        memmove(session->outBuf, session->outBuf + fr * session->outChannelCount,

                (session->framesOut - fr) * session->outChannelCount * sizeof(int16_t));

        session->framesOut -= fr;

        outBuffer->frameCount += fr;



        return 0;

    } else {

        return -ENODATA;
@@ -1565,6 +1463,18 @@ int PreProcessingFx_ProcessReverse(effect_handle_t self, audio_buffer_t* inBuffe 
        return -EINVAL;

    }



    if (inBuffer->frameCount != outBuffer->frameCount) {

        ALOGW("inBuffer->frameCount %zu is not equal to outBuffer->frameCount %zu",

              inBuffer->frameCount, outBuffer->frameCount);

        return -EINVAL;

    }



    if (inBuffer->frameCount != session->frameCount) {

        ALOGW("inBuffer->frameCount %zu != %zu representing 10ms at sampling rate %d",

              inBuffer->frameCount, session->frameCount, session->samplingRate);

        return -EINVAL;

    }



    session->revProcessedMsk |= (1 << effect->procId);



    //    ALOGV("PreProcessingFx_ProcessReverse In %d frames revEnabledMsk %08x revProcessedMsk
@@ -1573,16 +1483,6 @@ int PreProcessingFx_ProcessReverse(effect_handle_t self, audio_buffer_t* inBuffe 


    if ((session->revProcessedMsk & session->revEnabledMsk) == session->revEnabledMsk) {

        effect->session->revProcessedMsk = 0;

        size_t fr = session->frameCount - session->framesRev;

        if (inBuffer->frameCount < fr) {

            fr = inBuffer->frameCount;

        }

        session->framesRev += fr;

        inBuffer->frameCount = fr;

        if (session->framesRev < session->frameCount) {

            return 0;

        }

        session->framesRev = 0;

        if (int status = effect->session->apm->ProcessReverseStream(

                    (const int16_t* const)inBuffer->s16,

                    (const webrtc::StreamConfig)effect->session->revConfig,

media/libeffects/preprocessing/README.md

0 → 100644
+7 −0
Original line number Diff line number Diff line 
# Preprocessing effects



## Limitations

- Preprocessing effects currently work on 10ms worth of data and do not support

  arbitrary frame counts. This limiation comes from the underlying effects in

  webrtc modules

- There is currently no api to communicate this requirement

media/libeffects/preprocessing/tests/PreProcessingTest.cpp

+4 −3
Original line number Diff line number Diff line
@@ -451,8 +451,8 @@ int main(int argc, const char* argv[]) { 
        }

        audio_buffer_t inputBuffer, outputBuffer;

        audio_buffer_t farInBuffer{};

        inputBuffer.frameCount = samplesRead;

        outputBuffer.frameCount = samplesRead;

        inputBuffer.frameCount = frameLength;

        outputBuffer.frameCount = frameLength;

        inputBuffer.s16 = in.data();

        outputBuffer.s16 = out.data();
@@ -472,7 +472,7 @@ int main(int argc, const char* argv[]) { 
                }

            }



            farInBuffer.frameCount = samplesRead;

            farInBuffer.frameCount = frameLength;

            farInBuffer.s16 = farIn.data();

        }
@@ -519,6 +519,7 @@ int main(int argc, const char* argv[]) { 
        }

        frameCounter += frameLength;

    }

    printf("frameCounter: [%d]\n", frameCounter);

    // Release all the effect handles created

    for (int i = 0; i < PREPROC_NUM_EFFECTS; i++) {

        if (int status = AUDIO_EFFECT_LIBRARY_INFO_SYM.release_effect(effectHandle[i]);