media: simplify timestamp computation

Audio time bases are always the inverse of the sample rate, meaning the
increment is always the number of samples in a frame. Video usually has
a time base inverse of its framerate, but this is not always the case.

Change-Id: I50d2d84d073052f8b3a832e8b99725b9d66b12a8

src/media/audio/audio_input.cpp

@@ -52,16 +52,6 @@ AudioInput::~AudioInput()
     loop_.join();
 }
 
-// seq: frame number for video, sent samples audio
-// sampleFreq: fps for video, sample rate for audio
-// clock: stream time base (packetization interval times)
-// FIXME duplicate code from media encoder
-int64_t
-getNextTimestamp(int64_t seq, rational<int64_t> sampleFreq, rational<int64_t> clock)
-{
-    return (seq / (sampleFreq * clock)).real<int64_t>();
-}
-
 void
 AudioInput::process()
 {
@@ -99,7 +89,7 @@ AudioInput::process()
     auto audioFrame = resampled.toAVFrame();
     auto frame = audioFrame->pointer();
     auto ms = MediaStream("a:local", format_);
-    frame->pts = getNextTimestamp(sent_samples, ms.sampleRate, static_cast<rational<int64_t>>(ms.timeBase));
+    frame->pts = sent_samples;
     sent_samples += frame->nb_samples;
 
     {

src/media/audio/audio_rtp_session.cpp

@@ -156,23 +156,13 @@ AudioSender::setup(SocketPair& socketPair)
     return true;
 }
 
-// seq: frame number for video, sent samples audio
-// sampleFreq: fps for video, sample rate for audio
-// clock: stream time base (packetization interval times)
-// FIXME duplicate code from media_encoder
-static int64_t
-getNextTimestamp(int64_t seq, rational<int64_t> sampleFreq, rational<int64_t> clock)
-{
-    return (seq / (sampleFreq * clock)).real<int64_t>();
-}
-
 void
 AudioSender::update(Observable<std::shared_ptr<ring::AudioFrame>>* /*obs*/, const std::shared_ptr<ring::AudioFrame>& framePtr)
 {
     auto frame = framePtr->pointer();
     auto ms = MediaStream("a:local", frame->format, rational<int>(1, frame->sample_rate),
                           frame->sample_rate, frame->channels);
-    frame->pts = getNextTimestamp(sent_samples, ms.sampleRate, static_cast<rational<int64_t>>(ms.timeBase));
+    frame->pts = sent_samples;
     ms.firstTimestamp = frame->pts;
     sent_samples += frame->nb_samples;
 

src/media/media_encoder.cpp

@@ -343,15 +343,6 @@ MediaEncoder::startIO()
 #endif
 }
 
-// seq: frame number for video, sent samples audio
-// sampleFreq: fps for video, sample rate for audio
-// clock: stream time base (packetization interval times)
-static int64_t
-getNextTimestamp(int64_t seq, rational<int64_t> sampleFreq, rational<int64_t> clock)
-{
-    return (seq / (sampleFreq * clock)).real<int64_t>();
-}
-
 #ifdef RING_VIDEO
 int
 MediaEncoder::encode(VideoFrame& input, bool is_keyframe,
@@ -366,7 +357,11 @@ MediaEncoder::encode(VideoFrame& input, bool is_keyframe,
 
     auto frame = scaledFrame_.pointer();
     AVCodecContext* enc = encoders_[currentStreamIdx_];
-    frame->pts = getNextTimestamp(frame_number, enc->framerate, enc->time_base);
+    // ideally, time base is the inverse of framerate, but this may not always be the case
+    if (enc->framerate.num == enc->time_base.den && enc->framerate.den == enc->time_base.num)
+        frame->pts = frame_number;
+    else
+        frame->pts = (frame_number / (rational<int64_t>(enc->framerate) * rational<int64_t>(enc->time_base))).real<int64_t>();
 
     if (is_keyframe) {
         frame->pict_type = AV_PICTURE_TYPE_I;
@@ -382,8 +377,7 @@ MediaEncoder::encode(VideoFrame& input, bool is_keyframe,
 
 int MediaEncoder::encodeAudio(AudioFrame& frame)
 {
-    auto enc = encoders_[currentStreamIdx_];
-    frame.pointer()->pts = getNextTimestamp(sent_samples, enc->sample_rate, enc->time_base);
+    frame.pointer()->pts = sent_samples;
     sent_samples += frame.pointer()->nb_samples;
     encode(frame.pointer(), currentStreamIdx_);
     return 0;