Accumulate remainder in buffer duration calculations.

When dropping the remainder, the decoder and encoder timestamps start diverging after a few buffers when no speed changes are supposed to occur. Tracking the remainder keeps them in sync.

PiperOrigin-RevId: 408341074

library/transformer/src/main/java/com/google/android/exoplayer2/transformer/AudioSamplePipeline.java

@@ -31,7 +31,9 @@ import com.google.android.exoplayer2.audio.AudioProcessor;
 import com.google.android.exoplayer2.audio.AudioProcessor.AudioFormat;
 import com.google.android.exoplayer2.audio.SonicAudioProcessor;
 import com.google.android.exoplayer2.decoder.DecoderInputBuffer;
+import com.google.common.math.LongMath;
 import java.io.IOException;
+import java.math.RoundingMode;
 import java.nio.ByteBuffer;
 import org.checkerframework.checker.nullness.qual.EnsuresNonNullIf;
 import org.checkerframework.checker.nullness.qual.MonotonicNonNull;
@@ -62,6 +64,7 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
   private @MonotonicNonNull AudioFormat encoderInputAudioFormat;
   private @MonotonicNonNull MediaCodecAdapterWrapper encoder;
   private long nextEncoderInputBufferTimeUs;
+  private long encoderBufferDurationRemainder;
 
   private ByteBuffer sonicOutputBuffer;
   private boolean drainingSonicForSpeedChange;
@@ -82,6 +85,7 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
     sonicAudioProcessor = new SonicAudioProcessor();
     sonicOutputBuffer = AudioProcessor.EMPTY_BUFFER;
     nextEncoderInputBufferTimeUs = 0;
+    encoderBufferDurationRemainder = 0;
     speedProvider = new SegmentSpeedProvider(decoderInputFormat);
     currentSpeed = speedProvider.getSpeed(0);
     try {
@@ -269,7 +273,7 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
     encoderInputBufferData.put(inputBuffer);
     encoderInputBuffer.timeUs = nextEncoderInputBufferTimeUs;
     nextEncoderInputBufferTimeUs +=
-        getBufferDurationUs(
+        getEncoderBufferDurationUs(
             /* bytesWritten= */ encoderInputBufferData.position(),
             encoderInputAudioFormat.bytesPerFrame,
             encoderInputAudioFormat.sampleRate);
@@ -366,9 +370,17 @@ import org.checkerframework.checker.nullness.qual.RequiresNonNull;
         errorCode);
   }
 
-  // TODO(internal b/204978301): Ensure encoder and decoder timestamps match when no speed change.
-  private static long getBufferDurationUs(long bytesWritten, int bytesPerFrame, int sampleRate) {
-    long framesWritten = bytesWritten / bytesPerFrame;
-    return framesWritten * C.MICROS_PER_SECOND / sampleRate;
+  private long getEncoderBufferDurationUs(long bytesWritten, int bytesPerFrame, int sampleRate) {
+    // The calculation below accounts for remainders and rounding. Without that it corresponds to
+    // the following:
+    // bufferDurationUs = numberOfFramesInBuffer * sampleDurationUs
+    //     where numberOfFramesInBuffer = bytesWritten / bytesPerFrame
+    //     and   sampleDurationUs       = C.MICROS_PER_SECOND / sampleRate
+    long framesWrittenMicrosPerSecond =
+        bytesWritten * C.MICROS_PER_SECOND / bytesPerFrame + encoderBufferDurationRemainder;
+    long bufferDurationUs =
+        LongMath.divide(framesWrittenMicrosPerSecond, sampleRate, RoundingMode.CEILING);
+    encoderBufferDurationRemainder = framesWrittenMicrosPerSecond - bufferDurationUs * sampleRate;
+    return bufferDurationUs;
   }
 }

testdata/src/test/assets/transformerdumps/mp4/sample_sef_slow_motion.mp4.dump

@@ -135,145 +135,145 @@ sample:
   dataHashCode = 1000136444
   size = 140
   isKeyFrame = true
-  presentationTimeUs = 416
+  presentationTimeUs = 417
 sample:
   trackIndex = 0
   dataHashCode = 217961709
   size = 172
   isKeyFrame = true
-  presentationTimeUs = 3332
+  presentationTimeUs = 3334
 sample:
   trackIndex = 0
   dataHashCode = -879376936
   size = 176
   isKeyFrame = true
-  presentationTimeUs = 6915
+  presentationTimeUs = 6917
 sample:
   trackIndex = 0
   dataHashCode = 1259979587
   size = 192
   isKeyFrame = true
-  presentationTimeUs = 10581
+  presentationTimeUs = 10584
 sample:
   trackIndex = 0
   dataHashCode = 907407225
   size = 188
   isKeyFrame = true
-  presentationTimeUs = 14581
+  presentationTimeUs = 14584
 sample:
   trackIndex = 0
   dataHashCode = -904354707
   size = 176
   isKeyFrame = true
-  presentationTimeUs = 18497
+  presentationTimeUs = 18500
 sample:
   trackIndex = 0
   dataHashCode = 1001385853
   size = 172
   isKeyFrame = true
-  presentationTimeUs = 22163
+  presentationTimeUs = 22167
 sample:
   trackIndex = 0
   dataHashCode = 1545716086
   size = 196
   isKeyFrame = true
-  presentationTimeUs = 25746
+  presentationTimeUs = 25750
 sample:
   trackIndex = 0
   dataHashCode = 358710839
   size = 180
   isKeyFrame = true
-  presentationTimeUs = 29829
+  presentationTimeUs = 29834
 sample:
   trackIndex = 0
   dataHashCode = -671124798
   size = 140
   isKeyFrame = true
-  presentationTimeUs = 33579
+  presentationTimeUs = 33584
 sample:
   trackIndex = 0
   dataHashCode = -945404910
   size = 120
   isKeyFrame = true
-  presentationTimeUs = 36495
+  presentationTimeUs = 36500
 sample:
   trackIndex = 0
   dataHashCode = 1881048379
   size = 88
   isKeyFrame = true
-  presentationTimeUs = 38995
+  presentationTimeUs = 39000
 sample:
   trackIndex = 0
   dataHashCode = 1059579897
   size = 88
   isKeyFrame = true
-  presentationTimeUs = 40828
+  presentationTimeUs = 40834
 sample:
   trackIndex = 0
   dataHashCode = 1496098648
   size = 84
   isKeyFrame = true
-  presentationTimeUs = 42661
+  presentationTimeUs = 42667
 sample:
   trackIndex = 0
   dataHashCode = 250093960
   size = 751
   isKeyFrame = true
-  presentationTimeUs = 44411
+  presentationTimeUs = 44417
 sample:
   trackIndex = 0
   dataHashCode = 1895536226
   size = 1045
   isKeyFrame = true
-  presentationTimeUs = 59994
+  presentationTimeUs = 60063
 sample:
   trackIndex = 0
   dataHashCode = 1723596464
   size = 947
   isKeyFrame = true
-  presentationTimeUs = 81744
+  presentationTimeUs = 81834
 sample:
   trackIndex = 0
   dataHashCode = -978803114
   size = 946
   isKeyFrame = true
-  presentationTimeUs = 101410
+  presentationTimeUs = 101563
 sample:
   trackIndex = 0
   dataHashCode = 387377078
   size = 946
   isKeyFrame = true
-  presentationTimeUs = 121076
+  presentationTimeUs = 121271
 sample:
   trackIndex = 0
   dataHashCode = -132658698
   size = 901
   isKeyFrame = true
-  presentationTimeUs = 140742
+  presentationTimeUs = 140980
 sample:
   trackIndex = 0
   dataHashCode = 1495036471
   size = 899
   isKeyFrame = true
-  presentationTimeUs = 159492
+  presentationTimeUs = 159750
 sample:
   trackIndex = 0
   dataHashCode = 304440590
   size = 878
   isKeyFrame = true
-  presentationTimeUs = 178158
+  presentationTimeUs = 178480
 sample:
   trackIndex = 0
   dataHashCode = -1955900344
   size = 112
   isKeyFrame = true
-  presentationTimeUs = 196408
+  presentationTimeUs = 196771
 sample:
   trackIndex = 0
   dataHashCode = 88896626
   size = 116
   isKeyFrame = true
-  presentationTimeUs = 198741
+  presentationTimeUs = 199105
 sample:
   trackIndex = 1
   dataHashCode = 2139021989